Commercial and Scientific Solutions for Blood Glucose Monitoring—A Review

Diabetes is a chronic and, according to the state of the art, an incurable disease. Therefore, to treat diabetes, regular blood glucose monitoring is crucial since it is mandatory to mitigate the risk and incidence of hyperglycemia and hypoglycemia. Nowadays, it is common to use blood glucose meters or continuous glucose monitoring via stinging the skin, which is classified as invasive monitoring. In recent decades, non-invasive monitoring has been regarded as a dominant research field. In this paper, electrochemical and electromagnetic non-invasive blood glucose monitoring approaches will be discussed. Thereby, scientific sensor systems are compared to commercial devices by validating the sensor principle and investigating their performance utilizing the Clarke error grid. Additionally, the opportunities to enhance the overall accuracy and stability of non-invasive glucose sensing and even predict blood glucose development to avoid hyperglycemia and hypoglycemia using post-processing and sensor fusion are presented. Overall, the scientific approaches show a comparable accuracy in the Clarke error grid to that of the commercial ones. However, they are in different stages of development and, therefore, need improvement regarding parameter optimization, temperature dependency, or testing with blood under real conditions. Moreover, the size of scientific sensing solutions must be further reduced for a wearable monitoring system.

Quantification of stroke volume in a simulated healthy volunteer model of traumatic haemorrhage; a comparison of two non-invasive monitoring devices using error grid analysis alongside traditional measures of agreement

Introduction Haemorrhage is a leading cause of death following traumatic injury and the early detection of hypovolaemia is critical to effective management. However, accurate assessment of circulating blood volume is challenging when using traditional vital signs such as blood pressure. We conducted a study to compare the stroke volume (SV) recorded using two devices, trans-thoracic electrical bioimpedance (TEB) and supra-sternal Doppler (SSD), against a reference standard using trans- thoracic echocardiography (TTE). Methods A lower body negative pressure (LBNP) model was used to simulate hypovolaemia and in half of the study sessions lower limb tourniquets were applied as these are common in military practice and can potentially affect some haemodynamic monitoring systems. In order to provide a clinically relevant comparison we constructed an error grid alongside more traditional measures of agreement. Results 21 healthy volunteers aged 18–40 were enrolled and underwent 2 sessions of LBNP, with and without lower limb tourniquets. With respect to absolute SV values Bland Altman analysis showed significant bias in both non-tourniquet and tourniquet strands for TEB (-42.5 / -49.6 ml), rendering further analysis impossible. For SSD bias was minimal but percentage error was unacceptably high (35% / 48%). Degree of agreement for dynamic change in SV, assessed using 4 quadrant plots showed a seemingly acceptable concordance rate for both TEB (86% / 93%) and SSD (90% / 91%). However, when results were plotted on an error grid, constructed based on expert clinical opinion, a significant minority of measurement errors were identified that had potential to lead to moderate or severe patient harm. Conclusion Thoracic bioimpedance and suprasternal Doppler both demonstrated measurement errors that had the potential to lead to clinical harm and caution should be applied in interpreting the results in the detection of early hypovolaemia following traumatic injury.

Practical implementation of remote continuous glucose monitoring in hospitalized patients with diabetes

Disclaimer In an effort to expedite the publication of articles, AJHP is posting manuscripts online as soon as possible after acceptance. Accepted manuscripts have been peer-reviewed and copyedited, but are posted online before technical formatting and author proofing. These manuscripts are not the final version of record and will be replaced with the final article (formatted per AJHP style and proofed by the authors) at a later time. Purpose Inpatient diabetes management involves frequent assessment of glucose levels for treatment decisions. Here we describe a program for inpatient real-time continuous glucose monitoring (rtCGM) at a community hospital and the accuracy of rtCGM-based glucose estimates. Methods Adult inpatients with preexisting diabetes managed with intensive insulin therapy and a diagnosis of coronavirus disease 2019 (COVID-19) were monitored via rtCGM for safety. An rtCGM system transmitted glucose concentration and trending information at 5-minute intervals to nearby smartphones, which relayed the data to a centralized monitoring station. Hypoglycemia alerts were triggered by rtCGM values of ≤85 mg/dL, but rtCGM data were otherwise not used in management decisions; insulin dosing adjustments were based on blood glucose values measured via blood sampling. Accuracy was evaluated retrospectively by comparing rtCGM values to contemporaneous point-of-care (POC) blood glucose values. Results A total of 238 pairs of rtCGM and POC data points from 10 patients showed an overall mean absolute relative difference (MARD) of 10.3%. Clarke error grid analysis showed 99.2% of points in the clinically acceptable range, and surveillance error grid analysis showed 89.1% of points in the lowest risk category. It was determined that for 25% of the rtCGM values, discordances in rtCGM and POC values would likely have resulted in different insulin doses. Insulin dose recommendations based on rtCGM values differed by 1 to 3 units from POC-based recommendations. Conclusion rtCGM for inpatient diabetes monitoring is feasible. Evaluation of individual rtCGM-POC paired values suggested that using rtCGM data for management decisions poses minimal risks to patients. Further studies to establish the safety and cost implications of using rtCGM data for inpatient diabetes management decisions are warranted.

90% Accuracy for Photoplethysmography-Based Non-Invasive Blood Glucose Prediction by Deep Learning with Cohort Arrangement and Quarterly Measured HbA1c

Previously published photoplethysmography-(PPG) based non-invasive blood glucose (NIBG) measurements have not yet been validated over 500 subjects. As illustrated in this work, we increased the number subjects recruited to 2538 and found that the prediction accuracy (the ratio in zone A of Clarke’s error grid) reduced to undesirable 60.6%. We suspect the low prediction accuracy induced by larger sample size might arise from the physiological diversity of subjects, and one possibility is that the diversity might originate from medication. Therefore, we split the subjects into two cohorts for deep learning: with and without medication (1682 and 856 recruited subjects, respectively). In comparison, the cohort training for subjects without any medication had approximately 30% higher prediction accuracy over the cohort training for those with medication. Furthermore, by adding quarterly (every 3 months) measured glycohemoglobin (HbA1c), we were able to significantly boost the prediction accuracy by approximately 10%. For subjects without medication, the best performing model with quarterly measured HbA1c achieved 94.3% prediction accuracy, RMSE of 12.4 mg/dL, MAE of 8.9 mg/dL, and MAPE of 0.08, which demonstrates a very promising solution for NIBG prediction via deep learning. Regarding subjects with medication, a personalized model could be a viable means of further investigation.

Blood glucose prediction using multi-objective grammatical evolution: analysis of the “agnostic” and “what-if” scenarios

In this paper we investigate the benefits of applying a multi-objective approach for solving a symbolic regression problem by means of Grammatical Evolution. In particular, we extend previous work, obtaining mathematical expressions to model glucose levels in the blood of diabetic patients. Here we use a multi-objective Grammatical Evolution approach based on the NSGA-II algorithm, considering the root-mean-square error and an ad-hoc fitness function as objectives. This ad-hoc function is based on the Clarke Error Grid analysis, which is useful for showing the potential danger of mispredictions in diabetic patients. In this work, we use two datasets to analyse two different scenarios: What-if and Agnostic, the most common in daily clinical practice. In the What-if scenario, where future events are evaluated, results show that the multi-objective approach improves previous results in terms of Clarke Error Grid analysis by reducing the number of dangerous mispredictions. In the Agnostic situation, with no available information about future events, results suggest that we can obtain good predictions with only information from the previous hour for both Grammatical Evolution and Multi-Objective Grammatical Evolution.

Accuracy of Two CE-Marked Blood Glucose Monitoring System Based on EN ISO 15197:2015

Blood glucose monitoring systems (BGMS) are essential for the management of diabetic patients. Although International Organization for Standardization (ISO) 15197:2015 criteria require rigorous monitoring of BGMS devices before commercialization, manufacturing quality standards may decline after FDA or EU approval. This work aimed to demonstrate the accuracy and precision of two BGMS devices currently available on the market. A laboratory study was conducted from June to August 2021 using two BGMS devices. One hundred samples were collected and evaluated according to ISO 15197:2015 guidelines. Over 95% accuracy was achieved by both devices using stricter ISO criteria (at least 95% of values within ±10 mg/dL or ±10% of the results of the reference measurement procedure). Analysis of the error grid showed that 99.5% of the results were in zone A. Surveillance of the accuracy and precision of BGMS devices after FDA and EU approval is an essential procedure to help patients and physicians manage glycemia and determine an appropriate outcome and personalized approach to diabetes treatment.

Accuracy Assessment of the GlucoMen® Day CGM System in Individuals with Type 1 Diabetes: A Pilot Study

Aim of this study was to evaluate the accuracy and usability of a novel continuous glucose moni-toring (CGM) system designed for needle-free insertion and reduced environmental impact. We assessed sensor performance of two GlucoMen® Day CGM systems worn simultaneously in eight participants with type 1 diabetes. Self-monitoring of blood glucose (SMBG) was performed reg-ularly over 14 days at home. Participants underwent two standardized 5-hour meal challenges with frequent plasma glucose (PG) measurements using a laboratory reference instrument at the research center. When comparing CGM to PG the overall mean absolute relative difference (MARD) was 9.7 [2.6-14.6]%. The overall MARD of CGM vs SMBG was 13.1 [3.5-18.6]%. In the consensus error grid (CEG) analysis, 98% of both CGM/PG and CGM/SMBG pairs were in the clinically acceptable zones A and B. The analysis confirms that GlucoMen® Day CGM meets the clinical requirements for state-of-the-art CGM. The needle-free insertion technology is well toler-ated by users and reduces medical waste compared to conventional CGM systems.

Accuracy of Flash Glucose Monitoring During Postprandial Rest and Different Walking Conditions in Overweight or Obese Young Adults

Aims: To investigate the accuracy of FreeStyle LibreTM flash glucose monitoring (FGM) relevant to plasma glucose (PG) measurements during postprandial rest and different walking conditions in overweight/obese young adults.Methods: Data of 40 overweight/obese participants from two randomized crossover studies were pooled into four trials: (1) sitting (SIT, n = 40); (2) walking continuously for 30 min initiated 20 min before individual postprandial glucose peak (PPGP) (20iP + CONT, n = 40); (3) walking continuously for 30 min initiated at PPGP (iP + CONT, n = 20); and (4) accumulated walking for 30 min initiated 20 min before PPGP (20iP + ACCU, n = 20). Paired FGM and PG were measured 4 h following breakfast.Results: The overall mean absolute relative difference (MARD) between PG and FGM readings was 16.4 ± 8.6% for SIT, 16.2 ± 4.7% for 20iP + CONT, 16.7 ± 12.2% for iP + CONT, and 19.1 ± 6.8% for 20iP + ACCU. The Bland–Altman analysis showed a bias of −1.03 mmol⋅L–1 in SIT, −0.89 mmol⋅L–1 in 20iP + CONT, −0.82 mmol⋅L–1 in iP + CONT, and −1.23 mmol⋅L–1 in 20iP + ACCU. The Clarke error grid analysis showed that 99.6–100% of the values in all trials fell within zones A and B.Conclusion: Although FGM readings underestimated PG, the FGM accuracy was overall clinically acceptable during postprandial rest and walking in overweight/obese young adults.

Use of continuous glucose monitoring system in patients with type 2 mellitus diabetic during hemodialysis treatment

Background Studies highlight the inaccuracy of glycated hemoglobin (HbA1c) for the assessment of glycemic control in dialysis diabetics and suggest the use of continuous glucose monitoring (CGM) as an alternative. Of the CGMs, FreeStyle Libre® is the most used in worldwide, but there is still no consensus on its use in dialysis. Method A 3-week prospective study was performed with 12 patients comparing capillary and interstitial glucose during dialysis. Results Comparing capillary and interstitial measurements, similar values were observed in pre-dialysis in the 1st week (184.1 ± 69.5 mg/dl and 173.1 ± 78.9 mg/dl, respectively, p = 0.303), in patients with body mass index less than 24.9 kg/m2 (214.2 ± 72.2 mg/dl and 201.3 ± 77.0 mg/dl respectively, p = 0.466), in those dialysis fluid loss less than 2 l (185.5 ± 82.6 mg/dl and 183.1 ± 94.0 mg/dl respectively and p = 0.805) and in those with hemoglobin greater than 12 g/dl (152.0 ± 35, 5 mg/dl and 129.5 ± 47.4 mg/dl respectively, p = 0.016). In the correlation of the capillary measurement with the interstitial sensor, it was observed that the proportions in the Clarke Error Grid of zone A, zone B, zone C, zone D and zone E were 62.5%, 27.1%, 0.0%, 10.4% and 0.0% respectively and in the Parkes error grid in zone A, zone B, zone C, zone D and zone E were 80.6%, 9.7%, 9.7% 0.0% and 0.0%, respectively. Conclusion The mean absolute relative difference in dialysis patients is higher than the general population without end-stage renal disease. However, clinical decision-making based on the values measured by the system can be made with a good margin based on the correlation between interstitial and capillary measurements.

Incorporating Glucose Variability into Glucose Forecasting Accuracy Assessment Using the New Glucose Variability Impact Index and the Prediction Consistency Index: An LSTM Case Example

Background: In this work, we developed glucose forecasting algorithms trained and evaluated on a large dataset of free-living people with type 1 diabetes (T1D) using closed-loop (CL) and sensor-augmented pump (SAP) therapies; and we demonstrate how glucose variability impacts accuracy. We introduce the glucose variability impact index (GVII) and the glucose prediction consistency index (GPCI) to assess the accuracy of prediction algorithms. Methods: A long-short-term-memory (LSTM) neural network was designed to predict glucose up to 60 minutes in the future using continuous glucose measurements and insulin data collected from 175 people with T1D (41,318 days) and evaluated on 75 people (11,333 days) from the Tidepool Big Data Donation Dataset. LSTM was compared with two naïve forecasting algorithms as well as Ridge linear regression and a random forest using root-mean-square error (RMSE). Parkes error grid quantified clinical accuracy. Regression analysis was used to derive the GVII and GPCI. Results: The LSTM had highest accuracy and best GVII and GPCI. RMSE for CL was 19.8 ± 3.2 and 33.2 ± 5.4 mg/dL for 30- and 60-minute prediction horizons, respectively. RMSE for SAP was 19.6 ± 3.8 and 33.1 ± 7.3 mg/dL for 30- and 60-minute prediction horizons, respectively; 99.6% and 97.6% of predictions were within zones A+B of the Parkes error grid at 30- and 60-minute prediction horizons, respectively. Glucose variability was strongly correlated with RMSE (R≥0.64, P < 0.001); GVII and GPCI demonstrated a means to compare algorithms across datasets with different glucose variability. Conclusions: The LSTM model was accurate on a large real-world free-living dataset. Glucose variability should be considered when assessing prediction accuracy using indices such as GVII and GPCI.