Clarke Error Grid Analysis Research Articles

A Glucose Meter Accuracy and Precision Comparison: The FreeStyle Flash Versus the Accu-Chek Advantage, Accu-Chek Compact Plus, Ascensia Contour, and the BD Logic

This study compared the accuracy and precision of five blood glucose (BG) meters. Diabetes patients undergoing venipuncture for glucose testing were randomized to one of two groups consisting of three meters: FreeStyle Flash (Abbott Diabetes Care, Alameda, CA), Accu-Chek Advantage (Roche Diagnostics Corp., Indianapolis, IN), and Accu-Chek Compact Plus (Roche Diagnostics) or FreeStyle Flash, Ascensia Contour (Bayer Healthcare, Diagnostic Division, Tarrytown, NY), and BD Logic (BD Diabetes Care, Franklin Lake, NJ). Within 5 min following venipuncture, duplicate finger BG measurements from three ipsilateral fingers were taken. Finger glucose measurements were compared with laboratory reference values. Accuracy was assessed by a Clarke error grid analysis (EGA) and within 10% of the laboratory value criteria. Meter precision was determined by calculating the absolute mean differences in glucose values between duplicate samples. Finger sticks were obtained from 202 patients. Mean venipuncture BG was 148 mg/dL (SD +/- mg/64 dL; range 25-439 mg/dL). Accuracy by Clarke EGA (Zone A results) was demonstrated in 69% of Advantage samples, 75% of Compact Plus, and 96% of the first group of Flash versus 88% of the Contour, 67% of the Logic, and 91% of the second Flash samples (P < 0.05 for both Flash and Contour). Meter accuracy using the 10% criteria was demonstrated in 30%, 38%, 70%, 46%, 48%, and 68% of the samples, respectively (P < 0.05 for both Flash groups compared to each of the other meters). There were no differences in meter precision. No statistically significant differences in accuracy were evident using the Clarke EGA criteria (pooled results of Zone A and B), though the more strict 20% accuracy criteria (Zone A results only) found the Flash and Contour to have significantly greater accuracy compared to the Advantage, Compact Plus, and the Logic. Using the 10% accuracy criteria found the Flash to have significantly greater accuracy than each of the other four meters. All five meters demonstrated similar precision.

Noninvasive Transcutaneous Sampling of Glucose by Electroporation

In people with diabetes, blood glucose levels should be monitored regularly to prevent serious complications associated with diabetes. This involves the invasive method of withdrawing blood, which causes inconvenience to patients. The objective of this study was to investigate the efficiency of the noninvasive electroporation and transcutaneous sampling (ETS) technique for predicting blood glucose levels. In vitro studies were carried out in Franz diffusion cells using porcine epidermis to assess the feasibility of transcutaneous sampling of glucose. In vivo, the ETS technique was assessed in the diabetes-induced Sprague-Dawley rat model. Glucose was sampled following the application of 30 electrical pulses of 1 ms duration at 120 V/cm(2), 1 Hz. Clarke error grid analysis was carried out for the venous blood glucose levels that were determined by the ETS with reference to those measured by a glucose meter. The amount of glucose sampled by the ETS method both in vitro and in vivo was proportional to the dermal glucose concentration. All data points from in vivo studies were in A and B zones of Clarke error grid analysis, and the mean absolute relative error was 12.8%. Results of the present study demonstrate that ETS technique could be developed as a noninvasive method of predicting venous blood glucose levels in people with diabetes.

Evaluation of the VIA® Blood Chemistry Monitor for Glucose in Healthy and Diabetic Volunteers

Manual methods of blood glucose monitoring are labor-intensive, costly, prone to error, and expose the caregiver to blood. The VIA(R) blood chemistry monitor for glucose can automatically measure plasma glucose (PG) every 5 minutes for 72 hours using blood sampled from a peripheral vein/artery or a central vein. VIA performance was evaluated in eight normal and five type 1 diabetic (T1DM) subjects in 15 separate experiments. The VIA device was connected to a peripheral vein and reported a PG value every 5 minutes during each 510-minute experiment. Blood samples were collected manually every 10 minutes and assayed using a HemoCue(R) beta-glucose analyzer (HC). Whole blood HC measurements were corrected to PG values. Paired HC/VIA measurements (n = 717) were analyzed. Mean PG was 90 +/- 14 and 96 +/- 12 mg/dl in normal subjects and 194 +/- 64 and 173 +/- 48 mg/dl in T1DM subject as measured by the HC and VIA, respectively. Clark error grid analysis revealed 86% points in zone A, 11% points in zone B, and 2% points in zone D. Linear regression analysis yielded the following equation: VIA = 0.732 x HC + 30.5 (r(2) = 0.954). Residual analysis revealed a glucose-dependent bias between the HC and the VIA. VIA data were transformed using the linear regression equation to correct for bias. After the correction, the mean absolute relative difference between the VIA and the HC was less than 10%, and 99.6% of data were in zones A and B. The VIA was able to sample blood automatically every 5 minutes for more than 8 hours in the laboratory setting. On average, the VIA reported glucose values for 94% of the samples it attempted to obtain. This study demonstrated that the VIA blood chemistry monitor for glucose can reliably sample blood frequently for a prolonged period of time safely and effectively in diabetic and nondiabetic volunteers. Agreement between the two devices was the closest at normal glucose concentrations. After correcting for a glucose-dependent bias between the devices, the MARD was consistently less than 10% for all glucose ranges.

Clinical Validation of a New Control-Oriented Model of Insulin and Glucose Dynamics in Subjects with Type 1 Diabetes

The development of an artificial pancreas requires an accurate representation of diabetes pathophysiology to create effective and safe control systems for automatic insulin infusion regulation. The aim of the present study is the assessment of a previously developed mathematical model of insulin and glucose metabolism in type 1 diabetes and the evaluation of its effectiveness for the development and testing of control algorithms. Based on the already existing "minimal model" a new mathematical model was developed composed of glucose and insulin submodels. The glucose model includes the representation of peripheral uptake, hepatic uptake and release, and renal clearance. The insulin model describes the kinetics of exogenous insulin injected either subcutaneously or intravenously. The estimation of insulin sensitivity allows the model to personalize parameters to each subject. Data sets from two different clinical trials were used here for model validation through simulation studies. The first set had subcutaneous insulin injection, while the second set had intravenous insulin injection. The root mean square error between simulated and real blood glucose profiles (G(rms)) and the Clarke error grid analysis were used to evaluate the system efficacy. Results from our study demonstrated the model's capability in identifying individual characteristics even under different experimental conditions. This was reflected by an effective simulation as indicated by G(rms), and clinical acceptability by the Clarke error grid analysis, in both clinical data series. Simulation results confirmed the capacity of the model to faithfully represent the glucose-insulin relationship in type 1 diabetes in different circumstances.

Continuous Noninvasive Glucose Monitoring Technology Based on “Occlusion Spectroscopy”

A truly noninvasive glucose-sensing device could revolutionalize diabetes treatment by leading to improved compliance with recommended glucose levels, thus reducing the long-term complications and cost of diabetes. Herein, we present the technology and evaluate the efficacy of a truly noninvasive device for continuous blood glucose monitoring, the NBM (OrSense Ltd.). In vitro analysis was used to validate the technology and algorithms. A clinical study was performed to quantify the in vivo performance of the NBM device. A total of 23 patients with type 1 (n = 12) and type 2 (n = 11) diabetes were enrolled in the clinical study and participated in 111 sessions. Accuracy was assessed by comparing NBM data with paired self-monitoring of blood glucose meter readings. In vitro experiments showed a strong correlation between calculated and actual glucose concentrations. The clinical trial produced a total of 1690 paired glucose values (NBM vs reference). In the paired data set, the reference glucose range was 40-496 mg/dl. No systematic bias was found at any of the glucose levels examined (70, 100, 150, and 200 mg/dl). The mean relative absolute difference was 17.2%, and a Clarke error grid analysis showed that 95.5% of the measurements fall within the clinically acceptable A&B regions (zone A, 69.7%; and zone B, 25.7%). This study indicates the potential use of OrSense's NBM device as a noninvasive sensor for continuous blood glucose evaluation. The device was safe and well tolerated.

Clinical Experience of an Iontophoresis Based Glucose Measuring System

Currently finger pricking is the common method of blood glucose measurement in patients with diabetes mellitus. However, diabetes patients have proven to be reluctant to check their glucose profiles regularly because of the discomfort associated with this technique. Recently, a non-invasive and continuous Reverse Iontophoresis based Glucose Monitoring Device (RIGMD) was developed in Korea. The study was conducted during the period November 2003-January 2004 on 19 in-patients. Glucose measurements were performed using RIGMD between 10 a.m. and 4 p.m. Concurrent plasma glucose levels were checked hourly and subsequently compared with RIGMD data. The mean error of RIGMD measurements was -3.45±52.99 mg/dL with a mean absolute relative error of 20±15.16%. Measurements obtained by RIGMD were correlated with plasma glucose levels (correlation coefficient; 0.784 (p<0.05)) and this correlation was independent of time of data collection. However, after excluding confounding variables this correlation coefficient exhibited a tendency to increase. 98.9% of the results were clinically acceptable by Clarke error grid analysis. We concluded that RIGMD does not have the reliability and accuracy required to wholly replace conventional methods. However, further technical advancements that reduce its shortcomings would make this device useful for the management of diabetes.

Noninvasive Near-Infrared Blood Glucose Monitoring Using a Calibration Model Built by a Numerical Simulation Method: Trial Application to Patients in an Intensive Care Unit

We have applied a new methodology for noninvasive continuous blood glucose monitoring, proposed in our previous paper, to patients in ICU (intensive care unit), where strict controls of blood glucose levels are required. The new methodology can build calibration models essentially from numerical simulation, while the conventional methodology requires pre-experiments such as sugar tolerance tests, which are impossible to perform on ICU patients in most cases. The in vivo experiments in this study consisted of two stages, the first stage conducted on healthy subjects as preliminary experiments, and the second stage on ICU patients. The prediction performance of the first stage was obtained as a correlation coefficient (r) of 0.71 and standard error of prediction (SEP) of 28.7 mg/dL. Of the 323 total data, 71.5% were in the A zone, 28.5% were in the B zone, and none were in the C, D, and E zones for the Clarke error-grid analysis. The prediction performance of the second stage was obtained as an r of 0.97 and SEP of 27.2 mg/dL. Of the 304 total data, 80.3% were in the A zone, 19.7% were in the B zone, and none were in the C, D, and E zones. These prediction results suggest that the new methodology has the potential to realize a noninvasive blood glucose monitoring system using near-infrared spectroscopy (NIRS) in ICUs. Although the total performance of the present monitoring system has not yet reached a satisfactory level as a stand-alone system, it can be developed as a complementary system to the conventional one used in ICUs for routine blood glucose management, which checks the blood glucose levels of patients every few hours.

Microdialysis-Based 48-Hour Continuous Glucose Monitoring with GlucoDay™: Clinical Performance and Patients' Acceptance

This study was designed to assess clinical performance and patients' acceptance of the minimally invasive microdialysis-based continuous glucose monitoring system Gluco- Day() (Menarini Diagnostics, Florence, Italy) with a targeted monitoring time of 48 h. An inpatient sample of 28 patients with diabetes was studied. The analysis of clinical performance was performed using mean absolute differences (MAD) (in percent), Pearson correlations, the Bland-Altman analysis, and Clarke Error Grid Analysis (EGA). GlucoDay glucose values were compared with laboratory standard blood glucose measurements (glucohexokinase assay). The patients' acceptance of the monitoring device was assessed via two self-report scales (pain during application and discomfort while wearing device). A mean (+/- SD) monitoring time of 45.7 +/- 3.3 h with a total of 484 paired readings could be achieved. A correlation of r (average) = 0.91 and a MAD of 19.9% indicated satisfactory to good clinical performance. Of the paired readings, 95.5% fell into the acceptable A and B zones of the EGA. Rather wide 95% limits of agreement were revealed in the Bland-Altman analysis. Whereas virtually no pain was experienced during sensor application, discomfort associated with wearing the device was rather high. All of the participants, however, stated that they would wear the device again. Satisfactory to good performance of the GlucoDay monitor was observed, indicating the device to be suitable for routine clinical use. In particular, however, the discomfort experienced during wearing requires further improvements in its usability.

Minimally Invasive Extraction of Dermal Interstitial Fluid for Glucose Monitoring Using Microneedles

Compliance with glucose monitoring by patients with diabetes is poor because of the pain and inconvenience of conventional blood collection using lancets. To improve compliance, and thereby reduce morbidity and mortality associated with poor glucose control, this study sought to develop and test minimally invasive microneedles to extract dermal interstitial fluid (ISF) for glucose monitoring. We used a thermal puller to fabricate individual or multi-needle arrays of glass microneedles with tip radii of 15-40 microm to penetrate 700-1,500 microm deep into the skin of anesthetized hairless rats or conscious, normal, adult, human subjects. After applying a vacuum of 200-500 mm Hg for 2-10 min, we extracted ISF and measured glucose concentration. These measurements were compared with glucose levels in blood collected from the tail vein of rats or finger stick on humans. Using this procedure, 1-10 microL of ISF was extracted out of holes punctured in the skin using microneedles. Human subjects generally reported the procedure as painless. ISF glucose concentration correlated well with blood levels based on 140 measurements on 15 rats and six measurements on six human subjects, where 95% of rat data and 100% of human data fell within the clinically acceptable A + B region in Clarke Error Grid analysis. A linear calibration factor was needed to correlate ISF and blood glucose concentrations using our standard procedure. Modifying the procedure to prevent ISF evaporation during extraction provided a one-to-one correlation that eliminated the need for calibration. ISF glucose measurements tracked rapidly changing blood glucose levels following insulin injection with a time lag of less than 20 min. These results suggest that microneedle devices can be used to extract ISF for painless glucose monitoring.

Effect of hemoglobin concentration variation on the accuracy and precision of glucose analysis using tissue modulated, noninvasive, in vivo Raman spectroscopy of human blood: a small clinical study

Tissue modulated Raman spectroscopy was used noninvasively to measure blood glucose concentration in people with type I and type II diabetes with HemoCue fingerstick measurements being used as reference. Including all of the 49 measurements, a Clarke error grid analysis of the noninvasive measurements showed that 72% were A range, i.e., clinically accurate, 20% were B range, i.e., clinically benign, with the remaining 8% of measurements being essentially erroneous, i.e., C, D, or E range. Rejection of 11 outliers gave a correlation coefficient of 0.80, a standard deviation of 22 mg/dL with p<0.0001 for N=38 and places all but one of the measurements in the A and B ranges. The distribution of deviations of the noninvasive glucose measurements from the fingerstick glucose measurements is consistent with the suggestion that there are at least two systematic components in addition to the random noise associated with shot noise, charge coupled device spiking, and human factors. One component is consistent with the known variation of fingerstick glucose concentration measurements from laboratory reference measurements made using plasma or whole blood. A weak but significant correlation between the deviations of noninvasive measurements from fingerstick glucose measurements and the test subject's hemoglobin concentration was also observed.

Accuracy of continuous nocturnal glucose monitoring after 48 and 72 hours in type 2 diabetes patients on combined oral and insulin therapy

To evaluate the accuracy of nocturnal continuous glucose monitoring by CGMS. A CGMS was inserted in 14 type 2 diabetic patients on combined oral and insulin therapy at altogether 15 occasions. During the second (48 hours) and third (72 hours) night of registration each subject was observed and plasma glucose was analysed seven times during the night by venous sampling and compared to CGMS. These venous values were not used for calibration of the sensor. Pairs of data were analysed using correlation coefficient, Bland-Altman graphs and Clarke Error Grid analysis. 32% of CGMS data did not meet the quality criteria for optimal accuracy and were excluded. A correlation coefficient of 0.80 (p < 0.0001) was seen after 48 hours and a lower coefficient of 0.33 (p = 0.0082) after 72 hours. Bland-Altman analyses demonstrated that the dispersion of the values around the mean of differences was wider after 72 hours as compared to after 48 hours. In the Clark Error Grid analysis 100% of data were within zones A and B after 48 hours, with 60% in zone A. After 72 hours only 44% were in zone A and 7% of data were in the clinically unacceptable zone D. The MiniMed Medtronic CGMS has acceptable nocturnal accuracy after 48 hours of registration, but the accuracy thereafter deteriorates with time, implicating that the CGMS thereafter may prove less useful for nocturnal monitoring.

The Continuous Glucose Monitoring System During Pregnancy of Women with Type 1 Diabetes Mellitus: Accuracy Assessment

The Continuous Glucose Monitoring System (CGMS, Medtronic MiniMed, Northridge, CA) allows close monitoring of glucose patterns and might be helpful in explaining the persistence of high complication rates in pregnancies of women with type 1 diabetes. It was the aim of this study to determine whether the CGMS accurately reflects glucose levels in pregnant women with type 1 diabetes mellitus. Fifteen pregnant women with type 1 diabetes used the CGMS and were asked to determine at least seven fingerstick blood glucose levels each day, of which four were used for calibration. The patients were asked to keep a diary of the non-calibration blood glucose values. The accuracy of the CGMS was studied by comparing the non-calibration blood glucose values with simultaneously measured sensor glucose values using the Pearson correlation coefficient, the mean of absolute differences, and the Clarke error grid analysis. A total of 239 non-calibration blood glucose values were analyzed. The correlation coefficient between non-calibration blood glucose and sensor glucose value was 0.94 (P < 0.001). The mean of the absolute difference was 0.74 mmol/L. Of the non-calibration data 93.8% fell in the clinically acceptable zone of the Clarke error grid analysis. The CGMS is an accurate tool for additional glucose monitoring in pregnant women with type 1 diabetes mellitus.

Accuracy of the EasyTouch blood glucose self-monitoring system: a study of 516 cases

Background Self-monitoring blood glucose device is an important tool for diabetes patients to efficiently control their blood glucose concentrations. We evaluated the accuracy of EasyTouch glucose monitoring system. Methods Capillary blood glucose concentrations measured using EasyTouch and the reference values obtained from Yellow Springs Instruments (YSI) 2300 STAT were performed in the Department of Laboratory Medicine, Wei-Gong Memorial Hospital. Results were evaluated using (1) linear regression analysis, (2) Clarke Error Grid analysis, (3) percentage of readings within a defined range of deviation from the reference value, (4) bias plots, and (5) coefficients of variation (CVs) calculated from 60 measurements in series. Results The window of the 516 EasyTouch readings covered a range from 42 to 555 mg/dl. Linear regression analysis yielded a regression slope 0.9972, intercept 1.899 mg/dl, r 2 0.9571, and Syx 14.89 mg/dl. A Clarke Error Grid analysis showed 100% of the EasyTouch readings in clinically acceptable zones A and B. Of the EasyTouch readings, 98.3%, 91.9%, 78.3% and 46.9% were found within ±20%, ±15%, ±10%, and ±5%, respectively, of the reference values. Further analysis showed that the percentage of EasyTouch readings within the defined intervals was similar in three glucose ranges (≤100, 101–200, and ≥201 mg/dl). The CVs for the four lots of strips (lot 1 to lot 4) ranged from 3.5 to 5.5%, 2.1 to 4.8%, 1.8 to 3.6%, and 3.0 to 5.7%, respectively. Conclusions EasyTouch provides high accurate and precise glucose readings over a wide range of glucose concentrations.

Nonlinear model predictive control of glucose concentration in subjects with type 1 diabetes

A nonlinear model predictive controller has been developed to maintain normoglycemia in subjects with type 1 diabetes during fasting conditions such as during overnight fast. The controller employs a compartment model, which represents the glucoregulatory system and includes submodels representing absorption of subcutaneously administered short-acting insulin Lispro and gut absorption. The controller uses Bayesian parameter estimation to determine time-varying model parameters. Moving target trajectory facilitates slow, controlled normalization of elevated glucose levels and faster normalization of low glucose values. The predictive capabilities of the model have been evaluated using data from 15 clinical experiments in subjects with type 1 diabetes. The experiments employed intravenous glucose sampling (every 15 min) and subcutaneous infusion of insulin Lispro by insulin pump (modified also every 15 min). The model gave glucose predictions with a mean square error proportionally related to the prediction horizon with the value of 0.2 mmol L−1 per 15 min. The assessment of clinical utility of model-based glucose predictions using Clarke error grid analysis gave 95% of values in zone A and the remaining 5% of values in zone B for glucose predictions up to 60 min (n = 1674). In conclusion, adaptive nonlinear model predictive control is promising for the control of glucose concentration during fasting conditions in subjects with type 1 diabetes.

Experience with the continuous glucose monitoring system in a medical intensive care unit.

Strict glycemic control improves clinical outcomes in critically ill patients. However, practical tools for frequent monitoring of blood glucose (BG) levels in the intensive care unit (ICU) are limited. The Continuous Glucose Monitoring System (CGMS, Medtronic MiniMed, Northridge, CA) is currently approved for detecting glycemic excursions in outpatients with diabetes mellitus. The use of this device has never been carefully examined in the inpatient setting. This preliminary study was designed to investigate the accuracy of the CGMS in critically ill patients admitted to a medical ICU (MICU). Subjects at risk for hyperglycemia were recruited from among all patients admitted to our MICU. CGMS sensors were implanted for up to 72 h. Study subjects wore between one and five consecutive sensors. Four or more standard capillary BG readings were recorded per 24 h. All paired meter-sensor (M-S) readings were used both for CGMS calibration and for data analysis. Twenty-two MICU patients wore 41 CGMS sensors, yielding 546 M-S BG pairs. Overall, the Pearson correlation coefficient ( r ) was 0.88, with a mean M-S difference of 3.3 +/- 26.7 mg/dL (0.6 +/- 17.4%) and a mean absolute M-S difference of 19.7 +/- 18.3 mg/dL (12.8 +/- 11.9%). Clarke Error Grid analysis categorized 98.7% of the M-S pairs within "clinically acceptable" zones A and B. The CGMS is promising for potential use in critically ill patients. If validated in larger studies, the device could serve as a useful research tool for investigating the role of hyperglycemia (and strict glycemic control) in ICU patients. If further developed as a "real-time" glucose sensor, CGMS technology could ultimately prove clinically useful in the ICU, by decreasing nursing workload and/or by providing alarm signals for impending glycemic excursions.

Transcutaneous glucose measurement using near-infrared spectroscopy during hypoglycemia.

To analyze a transcutaneous near-infrared spectroscopy system as a technique for in vivo noninvasive blood glucose monitoring during euglycemia and hypoglycemia. Ten nondiabetic subjects and two patients with type 1 diabetes were examined in a total of 27 studies. In each study, the subject's plasma glucose was lowered to a hypoglycemia level (approximately 55 mg/dl) followed by recovery to a glycemic level of approximately 115 mg/dl using an intravenous infusion of insulin and 20% dextrose. Plasma glucose levels were determined at 5-min intervals by standard glucose oxidase method and simultaneously by a near-infrared spectroscopic system. The plasma glucose measured by the standard method was used to create a calibration model that could predict glucose levels from the near-infrared spectral data. The two data sets were correlated during the decline and recovery in plasma glucose, within 10 mg/dl plasma glucose ranges, and were examined using the Clarke Error Grid Analysis. Two sets of 1,704 plasma glucose determinations were examined. The near-infrared predictions during the fall and recovery in plasma glucose were highly correlated (r = 0.96 and 0.95, respectively). When analyzed during 10 mg/dl plasma glucose segments, the mean absolute difference between the near-infrared spectroscopy method and the chemometric reference ranged from 3.3 to 4.4 mg/dl in the nondiabetic subjects and from 2.6 to 3.8 mg/dl in the patients with type 1 diabetes. Using the Error Grid Analysis, 97.7% of all the near-infrared predictions were assigned to the A-zone. Our findings suggest that the near-infrared spectroscopy method can accurately predict plasma glucose levels during euglycemia and hypoglycemia in humans.

Preparation of ZnO thin film transducers by vapour transport: Belt, R.F., Florio, G.C. Journal of Applied Physics, Vol 39, No 11 (October 1968) pp 5215–5223

The Multisensor Glucose Monitoring System (MGMS) features non invasive sensors for dielectric characterisation of the skin and underlying tissue in a wide frequency range (1 kHz–100 MHz, 1 and 2 GHz) as well as optical characterisation. In this paper we describe the results of using an MGMS in a miniaturised housing with fully integrated sensors and battery.Six patients with Type I Diabetes Mellitus (age 44 ± 16 y; BMI 24.1 ± 1.3 kg/m2, duration of diabetes 27 ± 12 y; HbA1c 7.3 ± 1.0%) wore a single Multisensor at the upper arm position and performed a total of 45 in-clinic study days with 7 study days per patient on average (min. 5 and max. 10). Glucose changes were induced either orally or by i.v. glucose administration and the blood glucose was measured routinely. Several prospective data evaluation routines were applied to evaluate the data. The results are shown using one of the restrictive data evaluation routines, where measurements from the first 22 study days were used to train a linear regression model. The global model was then prospectively applied to the data of the remaining 23 study days to allow for an external validation of glucose prediction. The model application yielded a Mean Absolute Relative Difference of 40.8%, a Mean Absolute Difference of 51.9 mg dL−1, and a correlation of 0.84 on average per study day. The Clarke error grid analyses showed 89.0% in A + B, 4.5% in C, 4.6% in D and 1.9% in the E region.Prospective application of a global, purely statistical model, demonstrates that glucose variations can be tracked non invasively by the MGMS in most cases under these conditions.

Frequency response of an interdigital transducer for excitation of surface elastic waves: Tseng, C.C. IEEE Transactions on Electron Devices, Vol ED-15, No 8 (August 1968) pp 586–594 (958)

The Multisensor Glucose Monitoring System (MGMS) features non invasive sensors for dielectric characterisation of the skin and underlying tissue in a wide frequency range (1 kHz–100 MHz, 1 and 2 GHz) as well as optical characterisation. In this paper we describe the results of using an MGMS in a miniaturised housing with fully integrated sensors and battery.Six patients with Type I Diabetes Mellitus (age 44 ± 16 y; BMI 24.1 ± 1.3 kg/m2, duration of diabetes 27 ± 12 y; HbA1c 7.3 ± 1.0%) wore a single Multisensor at the upper arm position and performed a total of 45 in-clinic study days with 7 study days per patient on average (min. 5 and max. 10). Glucose changes were induced either orally or by i.v. glucose administration and the blood glucose was measured routinely. Several prospective data evaluation routines were applied to evaluate the data. The results are shown using one of the restrictive data evaluation routines, where measurements from the first 22 study days were used to train a linear regression model. The global model was then prospectively applied to the data of the remaining 23 study days to allow for an external validation of glucose prediction. The model application yielded a Mean Absolute Relative Difference of 40.8%, a Mean Absolute Difference of 51.9 mg dL−1, and a correlation of 0.84 on average per study day. The Clarke error grid analyses showed 89.0% in A + B, 4.5% in C, 4.6% in D and 1.9% in the E region.Prospective application of a global, purely statistical model, demonstrates that glucose variations can be tracked non invasively by the MGMS in most cases under these conditions.