Continuous Glucose Monitoring Traces Research Articles

Predicting the Risk of Developing Type 1 Diabetes Using a One-Week Continuous Glucose Monitoring Home Test With Classification Enhanced by Machine Learning: An Exploratory Study.

Detection of two or more autoantibodies (Ab) in the blood might describe those individuals at increased risk of developing type 1 diabetes (T1D) during the following years. The aim of this exploratory study is to propose a high versus low T1D risk classifier using machine learning technology based on continuous glucose monitoring (CGM) home data. Forty-two healthy relatives of people with T1D with mean ± SD age of 23.8 ± 10.5 years, HbA1c (glycated hemoglobin) of 5.3% ± 0.3%, and BMI (body mass index) of 23.2 ± 5.2 kg/m2 with zero (low risk; N = 21), and ≥2 (high risk; N = 21) Ab, were enrolled in an NIH (National Institutes of Health)-funded TrialNet ancillary study. Participants wore a CGM for a week and consumed three standardized liquid mixed meals (SLMM) instead of three breakfasts. Glycemic features were extracted from two-hour post-SLMM CGM traces, compared across groups, and used in four supervised machine learning Ab risk status classifiers. Recursive Feature Elimination (RFE) algorithm was used for feature selection; classifiers were evaluated through 10-fold cross-validation, using the receiver operating characteristic area under the curve (AUC-ROC) to select the best classification model. The percent time of glucose >180 mg/dL (T180), glucose range, and glucose CV (coefficient of variation) were the only significant differences between the glycemic features in the two groups with P values of .040, .035, and .028 respectively. The linear SVM (Support Vector Machine) model with RFE features achieved the best performance of classifying low-risk versus high-risk individuals with AUC-ROC = 0.88. A machine learning technology, combining a potentially self-administered one-week CGM home test, has the potential to reliably assess the T1D risk.

Predicting Immunological Risk for Stage 1 and Stage 2 Diabetes Using a 1-Week CGM Home Test, Nocturnal Glucose Increments, and Standardized Liquid Mixed Meal Breakfasts, with Classification Enhanced by Machine Learning.

Background: Predicting the risk for type 1 diabetes (T1D) is a significant challenge. We use a 1-week continuous glucose monitoring (CGM) home test to characterize differences in glycemia in at-risk healthy individuals based on autoantibody presence and develop a machine-learning technology for CGM-based islet autoantibody classification. Methods: Sixty healthy relatives of people with T1D with mean ± standard deviation age of 23.7 ± 10.7 years, HbA1c of 5.3% ± 0.3%, and body mass index of 23.8 ± 5.6 kg/m2 with zero (n = 21), one (n = 18), and ≥2 (n = 21) autoantibodies were enrolled in an National Institutes of Health TrialNet ancillary study. Participants wore a CGM for a week and consumed three standardized liquid mixed meals (SLMM) instead of three breakfasts. Glycemic outcomes were computed from weekly, overnight (12:00-06:00), and post-SLMM CGM traces, compared across groups, and used in four supervised machine-learning autoantibody status classifiers. Classifiers were evaluated through 10-fold cross-validation using the receiver operating characteristic area under the curve (AUC-ROC) to select the best classification model. Results: Among all computed glycemia metrics, only three were different across the autoantibodies groups: percent time >180 mg/dL (T180) weekly (P = 0.04), overnight CGM incremental AUC (P = 0.005), and T180 for 75 min post-SLMM CGM traces (P = 0.004). Once overnight and post-SLMM features are incorporated in machine-learning classifiers, a linear support vector machine model achieved the best performance of classifying autoantibody positive versus autoantibody negative participants with AUC-ROC ≥0.81. Conclusion: A new technology combining machine learning with a potentially self-administered 1-week CGM home test can help improve T1D risk detection without the need to visit a hospital or use a medical laboratory. Trial registration: ClinicalTrials.gov registration no. NCT02663661.

Evaluation of HbA1c from CGM traces in an Indian population.

The development of continuous glucose monitoring (CGM) over the last decade has provided access to many consecutive glucose concentration measurements from patients. A standard method for estimating glycated hemoglobin (HbA1c), already established in the literature, is based on its relationship with the average blood glucose concentration (aBG). We showed that the estimates obtained using the standard method were not sufficiently reliable for an Indian population and suggested two new methods for estimating HbA1c. Two datasets providing a total of 128 CGM and their corresponding HbA1c levels were received from two centers: Health Centre, Savitribai Phule Pune University, Pune and Joshi Hospital, Pune, from patients already diagnosed with diabetes, non-diabetes, and pre-diabetes. We filtered 112 data-sufficient CGM traces, of which 80 traces were used to construct two models using linear regression. The first model estimates HbA1c directly from the average interstitial fluid glucose concentration (aISF) of the CGM trace and the second model proceeds in two steps: first, aISF is scaled to aBG, and then aBG is converted to HbA1c via the Nathan model. Our models were tested on the remaining 32 data- sufficient traces. We also provided 95% confidence and prediction intervals for HbA1c estimates. The direct model (first model) for estimating HbA1c was HbA1cmmol/mol = 0.319 × aISFmg/dL + 16.73 and the adapted Nathan model (second model) for estimating HbA1c is HbA1cmmol/dL = 0.38 × (1.17 × ISFmg/dL) - 5.60. Our results show that the new equations are likely to provide better estimates of HbA1c levels than the standard model at the population level, which is especially suited for clinical epidemiology in Indian populations.

171-LB: Estimating the Immunological Risk for Type 1 Diabetes from a CGM-Based One Week Self-Administered Home Test

Background and aims: Detection of immunological risk (IR) for type 1 diabetes (T1D) requires testing for presence of islet autoantibodies (AB) . However, who and when to test is challenging. Our goal is to develop a (pre) screening technology to estimate the T1D IR from a simple continuous glucose monitoring (CGM) home test. Methods: Data from a NIH TrialNet ancillary study with 53 healthy relatives to T1D with (N=35) or without (N=18) AB, mean±SD age of 25.4±10.7, HbA1c of 5.1±0.3%, and BMI of 22.9±5.2 (kg/m2) were used. Subjects wear CGM for a week and consumed 3 caloric drinks (Boost) . Nine glycemia features were extracted from the 75 minutes post-Boost CGM traces and used for IR classification. K-fold cross-validation with eight different classification models (i.e., linear, nonlinear, advanced, etc.) was used to develop an IR classifier. Receiver operating characteristic (ROC) curves, sensitivity, and specificity were used to select the best performing classification model. Results: IR classifier based on a neural network (NN) model with upsampling and 10-fold cross-validation achieved the best performance between different classification models: an AUC-ROC of 0.93, a sensitivity of 89%, and a specificity of 75%. A random forest model-based classifier performed almost equally well. The Naive Bayes model achieved the best sensitivity of 100%, but had an AUC-ROC of 0.78, and specificity of 32%. On the other hand, the support vector machines models achieved the worst performance with the best being AUC-ROC of 0.63, a sensitivity of 65%, and a specificity of 54%. Conclusion: We have developed a new methodology, which combines a dedicated one-week CGM home test with a neural network-based classifier that reliably predicts the individual IR in terms of presence or absence of islet AB. The test is self-administered, does not require a visit to a hospital or a lab, and can be used as an alternative or in addition (e.g., for prescreening) to the standard test for presence of islet autoantibodies. Disclosure E. Montaser: None. O. Villard: None. C. Fabris: None. B. Lobo: Other Relationship; Dexcom, Inc., Research Support; Dexcom, Inc. S. A. Brown: Research Support; Dexcom, Inc., Insulet Corporation, Roche Diagnostics USA, Tandem Diabetes Care, Inc., Tolerion, Inc. M. D. Deboer: Research Support; Dexcom, Inc., Tandem Diabetes Care, Inc. B. Kovatchev: Other Relationship; Dexcom, Inc., Johnson & Johnson, Novo Nordisk, Sanofi, Research Support; Dexcom, Inc., Novo Nordisk, Tandem Diabetes Care, Inc., Speaker's Bureau; Dexcom, Inc., Tandem Diabetes Care, Inc. L. Farhy: Research Support; Dexcom, Inc., Novo Nordisk. Funding NIH DP3DK106907 (TrialNet ancillary study) and CRCF Award MF20-007-LS

218-OR: The Effect of Insulin Degludec vs. Insulin Glargine U100 on Continuous Glucose Monitoring (CGM) Recorded Glycemic Metrics in People with Type 1 Diabetes and Recurrent Nocturnal Severe Hypoglycemia

Background and Aims: Insulin Degludec (IDeg) , in comparison with Insulin Glargine U100 (IGla) , reduces the risk of hypoglycemic events in people with type 1 diabetes (T1D) . The impact on CGM assessed glycemic metrics: the coefficient of variation (CV) , time in range (TIR) , time below range (TBR) , and time-above range (TAR) is less known. We present CGM results from the HypoDeg trial comparing treatment with IDeg and IGla in people with T1D and recurrent nocturnal severe hypoglycemia. Materials and Methods: This is a pre-defined optional substudy of the HypoDeg trial: a 2-year investigator-initiated, randomized, cross-over trial comparing treatment with IDeg or IGla in 149 participants with T1D and at least one nocturnal severe hypoglycemic event within the last two years. Participants underwent 2 x 6 days of blinded CGM (Medtronic iPro) in each treatment arm after 6 and 12 months of treatment. Seventy-four participants completed at least one CGM period in each treatment arm. The endpoints were CV, TIR (3.9 - 10.0 mmol/L) , TBR at level 2 (< 3.0 mmol/L) and TAR (>10.0mmol/L, >13.9mmol/L) . Time spent in, below, or above range provided as percentage of readings. Results: We collected 261 CGM traces with a mean (SD) observation period of 5.9 (0.7) days. The all-day CV was lower with IDeg than IGla, with a mean (SE) CV of 40.5% (0.9) and 42.5% (0.9) , respectively (p=0.009) . A difference in CV during the night (23:00h to 07:00h) drove this, with a mean CV of 35.7% (1.1) for IDeg versus 39.6% (1.1) for IGla (p=0.001) . The all-day level 2 TBR was lower with IDeg than IGla, with a mean TBR of 1.8% (0.4) and 3.1% (0.4) , respectively (p=0.001) . The percentages of TIR and TAR were not different between treatments. Conclusion: In people with T1D prone to nocturnal severe hypoglycemia, treatment with IDeg results in a lower mean CV, reaching the definition of stable glucose levels during the night and a lower all-day TBR than IGla. Disclosure J.M.Brøsen: None. S.Lerche: None. K.Nørgaard: Advisory Panel; Medtronic, Novo Nordisk A/S, Consultant; Novo Nordisk A/S, Research Support; Dexcom, Inc., Medtronic, Novo Nordisk A/S, Zealand Pharma A/S, Speaker's Bureau; Medtronic, Stock/Shareholder; Novo Nordisk A/S. L.Tarnow: None. B.Thorsteinsson: None. U.Pedersen-bjergaard: Advisory Panel; Novo Nordisk A/S, Sanofi. R.Agesen: Employee; Novo Nordisk A/S. A.Alibegovic: Employee; Novo Nordisk A/S, Novo Nordisk A/S. H.U.Andersen: Advisory Panel; Abbott Diabetes, Stock/Shareholder; Novo Nordisk A/S. P.Gustenhoff: Advisory Panel; Abbott Diagnostics. T.K.Hansen: None. C.Hedetoft: None. C.R.Stolberg: None. C.B.Juhl: None. Funding Novo Nordisk

Measurement of Peak C-Peptide at Diagnosis Informs Glycemic Control but not Hypoglycemia in Adults With Type 1 Diabetes.

ContextHigh-residual C-peptide in longer-duration type 1 diabetes (T1D) is associated with fewer hypoglycemic events and reduced glycemic variability. Little is known about the impact of C-peptide close to diagnosis.ObjectiveUsing continuous glucose monitoring (CGM) data from a study of newly diagnosed adults with T1D, we aimed to explore if variation in C-peptide close to diagnosis influenced glycemic variability and risk of hypoglycemia.MethodsWe studied newly diagnosed adults with T1D who wore a Dexcom G4 CGM for 7 days as part of the Exercise in Type 1 Diabetes (EXTOD) study. We examined the relationship between peak stimulated C-peptide and glycemic metrics of variability and hypoglycemia for 36 CGM traces from 23 participants.ResultsFor every 100 pmol/L-increase in peak C-peptide, the percentage of time spent in the range 3.9 to 10 mmol/L increased by 2.4% (95% CI, 0.5-4.3), P = .01) with a reduction in time spent at level 1 hyperglycemia (> 10 mmol/L) and level 2 hyperglycemia (> 13.9 mmol/L) by 2.6% (95% CI, –4.9 to –0.4, P = .02) and 1.3% (95% CI, –2.7 to –0.006, P = .04), respectively. Glucose levels were on average lower by 0.19 mmol/L (95% CI, –0.4 to 0.02, P = .06) and SD reduced by 0.14 (95% CI, –0.3 to –0.02, P = .02). Hypoglycemia was not common in this group and no association was observed between time spent in hypoglycemia (P = .97) or hypoglycemic risk (P = .72). There was no association between peak C-peptide and insulin dose–adjusted glycated hemoglobin A1c (P = .45).ConclusionC-peptide is associated with time spent in the normal glucose range and with less hyperglycemia, but not risk of hypoglycemia in newly diagnosed people with T1D.

65-LB: Real-World Continuous Glucose Monitoring Data on Time-in-Range from a U.S. Population, 2015–2019

Aims: Improvements in technology and access have led to growing adoption of continuous glucose monitoring (CGM). Targets for glucose metrics including time in range (TIR) have been published. We aimed to investigate i) the proportion of people with ≥70% TIR and ii) the proportion with ≥70% TIR and Methods: Data were collected from 2015 to 2019 from the Cornerstone4Care (C4C) database, a freely available patient support program for people with both type 1 diabetes (T1D) and type 2 diabetes (T2D) on any treatment type. CGM traces were divided into 14-day periods according to the ambulatory glucose profile (AGP)-reporting system. Only profiles with data aligned with these standards were included. Patient data are presented using the first or the mean of all AGP profiles. Results: In total, 484 persons uploaded CGM-data to the database (4727 AGPs); 242 had T1D and 74 had T2D (self-reported), the rest were unknown (Table). People uploaded between 1 and 75 AGPs (mean=10). Average TIR (70-180 mg/dL) based on mean profiles was 63%, 68% and 64% for T1D, T2D and all, respectively. Conclusions: Less than half of the population achieved ≥70% TIR, ~30% with ≥70% TIR and Disclosure R. M. Bergenstal: Advisory Panel; Self; Abbott Diabetes, Eli Lilly and Company, Novo Nordisk, Onduo LLC., Roche, Sanofi, United Healthcare, Consultant; Self; Abbott Diabetes, Ascensia, Dexcom, Inc., Eli Lilly and Company, Hygieia, Johnson & Johnson, Medtronic, Novo Nordisk, Onduo LLC., Roche, Sanofi, United Healthcare, Other Relationship; Self; HealthPartners Institute, Research Support; Self; Abbott Diabetes, Dexcom, Inc., Eli Lilly and Company, Helmsley Charitable Trust, Hygieia, Johnson & Johnson, Medtronic, NIDDK, Novo Nordisk, Onduo LLC., Roche, Sanofi, United Healthcare. E. Hachmann-nielsen: Employee; Self; Novo Nordisk A/S. J. Tarp: Employee; Self; Novo Nordisk. K. Kvist: Employee; Self; Novo Nordisk A/S. J. B. Buse: Consultant; Self; Cirius Therapeutics, CSL Behring, Fortress Biotch, Mellitus Health, Moderna, Pendulum Therapeutics, Praetego Inc., Stability Health, Zealand Pharma A/S, Other Relationship; Self; AstraZeneca, Eli Lilly and Company, Novo Nordisk, vTv Therapeutics, Research Support; Self; NovaTarg Therapeutics, Novo Nordisk, Sanofi, Tolerion, Inc., vTv Therapeutics, Stock/Shareholder; Self; Mellitus Health, Pendulum Therapeutics, PhaseBio Pharmaceuticals, Inc., Stability Health. Funding Novo Nordisk A/S

153-OR: Measurement of C-Peptide at Diagnosis Informs Glycemic Control but Not Hypoglycemic Risk in Type 1 Diabetes

Background and Aims: High residual C-peptide in longer duration type 1 diabetes associates with fewer hypoglycemic events and reduced glycemic variability. However, little is known about the impact of C-peptide close to diagnosis. Using continuous glucose monitoring (CGM) data from a study of newly diagnosed adults with type 1 diabetes, we aimed to explore if variation in C-peptide close to diagnosis influenced glycemic variability and risk of hypoglycaemia. Materials and Methods: We studied newly diagnosed adults with type 1 diabetes who wore a Dexcom G4 CGM as part the EXTOD study. We examined the relationship between peak stimulated C-peptide in a mixed meal tolerance test and glycemic metrics of variability and hypoglycemia for 36 CGM traces from 23 participants. Results: Higher levels of C-peptide associated with lower glycemic variability, more time in range and less hyperglycemia. For every 100 pmol/L increase in peak C-peptide, percentage time spent in the range 3.9-10 mmol/L was increased by 2.39% [95% CI: 0.51,4.26], p=0.012) with a reduction in time spent in level 1 hyperglycaemia (> 10 mmol/L) and level 2 hyperglycemia (> 13.9 mmol/L) glucose levels by 2.64% [95% CI: -4.87, -0.41, p=0.018) and 1.33% [95% CI: -2.66, -0.0057], p= 0.041) respectively. Glucose levels were on average lower by 0.19 mmol/L ([95 % CI: -0.39,0.015], p=0.06) and standard deviation reduced by 0.14 [95% CI: -0.25, -0.023], p=0.017) for every 100 pmol/L increase in peak C-peptide. Hypoglycemia was not common in this group and no association was observed between time spent in hypoglycemia (p=0.967) or hypoglycemic risk (p=0.721). There was no association between peak C-peptide and insulin dose (p=0.707), HbA1C (p=0.364) or insulin dose adjusted HbA1c (IDAA1c, p=0.452). Conclusion: C-peptide associates with time spent in normal glucose range and with less hyperglycemia, but not risk of hypoglycemia in newly diagnosed people with type 1 diabetes. Disclosure A. Carr: None. R. A. Oram: Consultant; Self; Janssen Research & Development, LLC. P. Narendran: Speaker’s Bureau; Self; Novo Nordisk Inc., Sanofi. R. C. Andrews: None.

Forecasting of Glucose Levels and Hypoglycemic Events: Head-to-Head Comparison of Linear and Nonlinear Data-Driven Algorithms Based on Continuous Glucose Monitoring Data Only.

In type 1 diabetes management, the availability of algorithms capable of accurately forecasting future blood glucose (BG) concentrations and hypoglycemic episodes could enable proactive therapeutic actions, e.g., the consumption of carbohydrates to mitigate, or even avoid, an impending critical event. The only input of this kind of algorithm is often continuous glucose monitoring (CGM) sensor data, because other signals (such as injected insulin, ingested carbs, and physical activity) are frequently unavailable. Several predictive algorithms fed by CGM data only have been proposed in the literature, but they were assessed using datasets originated by different experimental protocols, making a comparison of their relative merits difficult. The aim of the present work was to perform a head-to-head comparison of thirty different linear and nonlinear predictive algorithms using the same dataset, given by 124 CGM traces collected over 10 days with the newest Dexcom G6 sensor available on the market and considering a 30-min prediction horizon. We considered the state-of-the art methods, investigating, in particular, linear black-box methods (autoregressive; autoregressive moving-average; and autoregressive integrated moving-average, ARIMA) and nonlinear machine-learning methods (support vector regression, SVR; regression random forest; feed-forward neural network, fNN; and long short-term memory neural network). For each method, the prediction accuracy and hypoglycemia detection capabilities were assessed using either population or individualized model parameters. As far as prediction accuracy is concerned, the results show that the best linear algorithm (individualized ARIMA) provides accuracy comparable to that of the best nonlinear algorithm (individualized fNN), with root mean square errors of 22.15 and 21.52 mg/dL, respectively. As far as hypoglycemia detection is concerned, the best linear algorithm (individualized ARIMA) provided precision = 64%, recall = 82%, and one false alarm/day, comparable to the best nonlinear technique (population SVR): precision = 63%, recall = 69%, and 0.5 false alarms/day. In general, the head-to-head comparison of the thirty algorithms fed by CGM data only made using a wide dataset shows that individualized linear models are more effective than population ones, while no significant advantages seem to emerge when employing nonlinear methodologies.

Effect of Insulin Analogs on Frequency of Non–Severe Hypoglycemia in Patients with Type 1 Diabetes Prone to Severe Hypoglycemia: Much Higher Rates Detected by Continuous Glucose Monitoring than by Self-Monitoring of Blood Glucose—The HypoAna Trial

Hypoglycemia is an increasingly important endpoint in clinical diabetes trials. The assessment of hypoglycemia should therefore be as complete as possible. Blinded continuous glucose monitoring (CGM) provides an improved opportunity to capture asymptomatic and nocturnal events. Here we report results from the HypoAna trial comparing all-analog-insulin therapy (aspart/detemir) with all-human-insulin therapy (neutral protamine Hagedorn/regular) on non-severe hypoglycemia (symptomatic and asymptomatic hypoglycemia) as assessed by blinded CGM and compared with data obtained by self-monitoring of blood glucose (SMBG) in patients with type 1 diabetes and recurrent severe hypoglycemia. Fifty-three patients completed a substudy of 4 × 3 days of blinded CGM. CGM traces were reviewed for hypoglycemic events lasting 15 min or longer. At the threshold ≤3.9 mmol/L, the per-protocol analysis demonstrated a 40% rate reduction (95% confidence interval [CI] 20%-60%; P = 0.002) in nocturnal non-severe hypoglycemia during analog treatment, mainly due to a 40% rate reduction (95% CI 0%-70%; P = 0.03) of nocturnal asymptomatic hypoglycemia. Similar nonsignificant trends were seen at the glucose threshold ≤3.0 mmol/L. Overall CGM-detected that nocturnal asymptomatic hypoglycemia ≤3.9 mmol/L was ∼17 times more frequent than SMBG-detected episodes (52 vs. 3 events/patient-year). This translates into a time needed to treat one patient with insulin analogs to prevent one episode that is 34 times shorter using CGM data than SMBG data (1.4 vs. 47 weeks). Capturing hypoglycemic events by the conventional method of SMBG in patients with impaired awareness reveals only a limited number of events. Blinded CGM can provide more complete data, particularly in terms of asymptomatic and nocturnal events.

Glycaemic State Analysis from Continuous Glucose Monitoring Measurements in Infants

Abstract Neonatal hypoglycaemia is common in at-risk infants and can cause adverse neurologic outcomes in later life. Continuous glucose monitoring (CGM) technology offers a way to continuously monitor patient condition, helping to detect hypoglycaemia as well as provide insight into the general glycaemic state of the patient. Characterising Glycaemic States can be easily done by eye, but no simple, clinically relevant algorithm exists to do this characterisation analytically or computationally. This paper presents such an algorithm to characterise Glycaemic States and detect State Changes. This algorithm was developed on a cohort of 366 infants, using a total of 12356 hours of CGM sensor data. State Changes were defined as an intersection between a 6-hour rolling average of the CGM trace and the average of the whole interstitial glucose CGM trace, with a 5 hour minimum crossover threshold defining a single State. The majority of infants were found to have experienced less than 2 State Changes in the first 48 hours of birth (279 of 366 patients, 76%). The median number of State Changes per day was 0.68 [IQR: 0.60, 1.14], while the median absolute change in IG over a State Change was 0.6 mmol/L [IQR: 0.4, 0.9 mmol/L]. Visually, the majority of algorithmically characterised State Changes matched CGM traces characterised by eye. Future use of the algorithm could associate the State Changes with clinical outcomes.

Glycemic Variability Percentage: A Novel Method for Assessing Glycemic Variability from Continuous Glucose Monitor Data.

Background: High levels of glycemic variability are still observed in most patients with diabetes with severe insulin deficiency. Glycemic variability may be an important risk factor for acute and chronic complications. Despite its clinical importance, there is no consensus on the optimum method for characterizing glycemic variability.Method: We developed a simple new metric, the glycemic variability percentage (GVP), to assess glycemic variability by analyzing the length of the continuous glucose monitoring (CGM) temporal trace normalized to the duration under evaluation. The GVP is similar to other recently proposed glycemic variability metrics, the distance traveled, and the mean absolute glucose (MAG) change. We compared results from distance traveled, MAG, GVP, standard deviation (SD), and coefficient of variation (CV) applied to simulated CGM traces accentuating the difference between amplitude and frequency of oscillations. The GVP metric was also applied to data from clinical studies for the Dexcom G4 Platinum CGM in subjects without diabetes, with type 2 diabetes, and with type 1 diabetes (adults, adolescents, and children).Results: In contrast to other metrics, such as CV and SD, the distance traveled, MAG, and GVP all captured both the amplitude and frequency of glucose oscillations. The GVP metric was also able to differentiate between diabetic and nondiabetic subjects and between subjects with diabetes with low, moderate, and high glycemic variability based on interquartile analysis.Conclusion: A new metric for the assessment of glycemic variability has been shown to capture glycemic variability due to fluctuations in both the amplitude and frequency of glucose given by CGM data.

Retrofitting Real-Life Dexcom G5 Data

We proposed in 2014 a retrofitting algorithm to retrospectively increase the accuracy of continuous glucose monitoring (CGM) data by using some blood glucose (BG) measurements. The method proved effective on Dexcom SEVEN Plus when about 10 highly accurate YSI measurements/session were available. In this study, we test the method on Dexcom G5 sensor in a more realistic setup, where only five capillary BG measurements (self-monitoring blood glucose [SMBG]) per 12 h-session are available. Furthermore, we investigate how accuracy is affected by the number of BG measurements. The algorithm was tested in 51 adults and 46 adolescents studied for 7 days with Dexcom G5. Each patient also underwent an ∼12-h hospital admission where frequent SMBG and YSI measurements were collected. First, five SMBGs per 12-h session were used to retrofit the CGM. Then, we varied the number of SMBGs provided to the method from 2 to 10 per 12-h session. Retrofitted CGM traces with five SMBGs per 12-h session have lower mean absolute difference than original CGM, reduced from 16.2 to 10.7 mg/dL (P < 0.001) in adults and from 17.6 to 11.5 mg/dL (P < 0.001) in adolescents, and mean absolute relative difference is reduced from 9.0% to 6.4% (P < 0.001) in adults and from 10.3% to 6.8% (P < 0.001) in adolescents. Reducing the number of BG measurements reduces improvement in the accuracy from >30% with 10 SMBGs per 12-h session to <16% with 2 SMBGs/day. The retrofitting method retrospectively improves the accuracy of CGM data, even if applied to one of the most accurate CGM sensors currently available on the market.

Interpreting the ambulatory glucose profile

A systematic approach to interpreting dense glucose data from continuous glucose monitoring (CGM) systems is important for dealing with the daunting amount of data that these systems provide. Moreover, a consensus approach would be useful to reduce differences in interpretation of similar data between physicians. The ambulatory glucose profile (AGP), now available for the FreeStyle Libre flash glucose monitoring system and some CGM systems, allows simpler and more confident identification of patterns, by averaging data for the modal day (itself derived from several daily CGM traces). Its output provides a single curve representing the average, with interquartile and interdecile ranges. Even scattered individual data values without discernible pattern can be averaged and produce useful clinically relevant information, especially where there is considerable glycaemic variability. Consensus recommendations on the use of the AGP build on an understanding of the user's daily routine, followed by examination of the AGP for periods of hypoglycaemic risk, i.e. when the 10th centile line for glucose approaches or enters the hypoglcycaemic range. These analyses are supported by the glucose pattern insights analysis available from the FreeStyle Libre flash glucose monitoring system, which flags up the likelihood of low glucose, the proximity of the median glucose to target, and the degree of variability below the median at various times of day. Where hypoglycaemia patterns are not apparent, the consensus guidance recommends assessing the periods around overnight, morning meal, midday meal, and evening meal in turn. In these ways, the consensus guideline provides the systematic approach to interpretation which is the cornerstone of using the AGP.

Poor Agreement of Computerized Calculators for Mean Amplitude of Glycemic Excursions

Glucose variability has been identified as a predictor of hypoglycemia and has been associated with mortality in critically ill patients without diabetes. A popular metric to quantify glucose variability is the mean amplitude of glycemic excursions (MAGE). The "ruler and pencil" approach to calculate MAGE is operator-dependent and time-consuming for analysis of continuous glucose monitoring data. Therefore, several computer software programs have been developed for the automated calculation of MAGE. The aim of our study was to evaluate the agreement of currently available MAGE calculators when applied to the same set of continuous glucose monitoring (CGM) traces. Four software programs for calculation of MAGE were identified and used to calculate MAGE of 21 CGM traces from seven patients with type 1 diabetes. Subsequently, the median MAGE per calculator was calculated. The correlation between the MAGE calculators was evaluated by Spearman's correlation analysis. Between-group comparison was performed using analysis of variance. The median MAGE (interquartile range) per calculator was 8.7 (7.1-10.7), 6.7 (5.5-8.6), 6.7 (5.2-8.6), and 5.8 (4.3-7.1), which was statistically different overall (P<0.001). The correlation coefficients between the calculators ranged from 0.787 to 0.999. Available computer programs developed to calculate MAGE show varying agreement. Although software programs for the calculation of MAGE would seem attractive to assess glucose variability, their use has limitations by different outcomes, in the absence of a gold standard.

257 Challenges in CFRD: The advantages of continuous glucose monitoring (CGM)

Introduction: With survival increasing in CF, CF-related diabetes (CFRD) and its assocated complications are more common. Achieving optimal glycaemic control can be particularly difficult despite traditional management methods. CGM provides greater insight into glycaemic trends through the day, and helps determine the benefits of treatment changes. We routinely use CGM for optimisation of diabetic control at our large CF centre (n = 282; 40% CFRD), and have explored its utility in the management of this CF complication. Method: We evaluated CGM traces in 29 CFRD patients over a 20 month period, along with any subsequent changes in weight, pulmonary function, HBA1C and antibiotic treatment. Food and exercise diaries completed during CGM were also evaluated. Results: See Table [Mean (SD)]: 69% gained weight, 72% improved their lung function (p< 0.0001) and 62% improved HBA1C (p< 0.0001). Overall, 45% required less IV antibiotic therapy (p< 0.0001) and 34% required fewer oral antibiotic courses. 90% of subjects had dietary adjustments following review of food and exercise diaries. 34% had a change in the type of insulin prescribed following CGM. Conclusion: This study demonstrates the utility of CGM to improve clinical status and optimize treatment including the dosage and type of insulin therapy. CGM used in conjunction with good education, support and the concurrent mapping of food and exercise diaries is of educational benefit allowing the fine-tuning of any adjustments of treatment.

Continuous glucose monitoring: the clinical picture. How to interpret and use the data

AbstractThere is a limited evidence base as to the benefits of continuous glucose monitoring (CGM) in clinical practice, but it is clear that in order to realise improvements in glycaemic control when using CGM there is a requirement for both health care professionals and patients to have the ability to interpret the data obtained from CGM.This article describes a personal approach to analysing CGM data using a structured approach and reporting tool, with examples to demonstrate how this system is implemented in practice. By viewing the daily overlay, then breaking the CGM traces into overnight, fasting/pre‐meal and post‐meal phases, and finally looking at the impact of other factors such as exercise, alcohol and work patterns, the user can be educated to make changes to both their insulin regimen and lifestyle to optimise glycaemic control. Those offered CGM as a real‐time adjunct to their intensive insulin regimen need to have such a structured approach to get positive re‐inforcement and thus use CGM sufficiently frequently to gain real benefit from it. Copyright © 2012 John Wiley &amp; Sons.

Using Stochastic modelling to identify unusual continuous glucose monitor measurements and behaviour, in newborn infants

BackgroundAbnormal blood glucose (BG) concentrations have been associated with increased morbidity and mortality in both critically ill adults and infants. Furthermore, hypoglycaemia and glycaemic variability have both been independently linked to mortality in these patients. Continuous Glucose Monitoring (CGM) devices have the potential to improve detection and diagnosis of these glycaemic abnormalities. However, sensor noise is a trade-off of the high measurement rate and must be managed effectively if CGMs are going to be used to monitor, diagnose and potentially help treat glycaemic abnormalities.AimTo develop a tool that will aid clinicians in identifying unusual CGM behaviour and highlight CGM data that potentially need to be interpreted with care.MethodsCGM data and BG measurements from 50 infants at risk of hypoglycaemia were used. Unusual CGM measurements were classified using a stochastic model based on the kernel density method and historical CGM measurements from the cohort. CGM traces were colour coded with very unusual measurements coloured red, highlighting areas to be interpreted with care. A 5-fold validation of the model was Monte Carlo simulated 25 times to ensure an adequate model fit.ResultsThe stochastic model was generated using ~67,000 CGM measurements, spread across the glycaemic range ~2-10 mmol/L. A 5-fold validation showed a good model fit: the model 80% confidence interval (CI) captured 83% of clinical CGM data, the model 90% CI captured 91% of clinical CGM data, and the model 99% CI captured 99% of clinical CGM data. Three patient examples show the stochastic classification method in use with 1) A stable, low variability patient which shows no unusual CGM measurements, 2) A patient with a very sudden, short hypoglycaemic event (classified as unusual), and, 3) A patient with very high, potentially un-physiological, glycaemic variability after day 3 of monitoring (classified as very unusual).ConclusionsThis study has produced a stochastic model and classification method capable of highlighting unusual CGM behaviour. This method has the potential to classify important glycaemic events (e.g. hypoglycaemia) as true clinical events or sensor noise, and to help identify possible sensor degradation. Colour coded CGM traces convey the information quickly and efficiently, while remaining computationally light enough to be used retrospectively or in real-time.

Response to the Need for Identifying Standardized Indices for Measuring Glucose Variability

We thank Picconi and colleagues1 for their valuable comments on our study.2 They observe a lack of standardization for measurement of glycemic variability (GV) and suggest that there might be different outcomes with different GV measures. Furthermore, the authors suggest a possible confounding role for insulin secretagogues on GV and oxidative stress parameters. The lack of standardization of GV measurement methods is a problem we had discussed.3 This lack limits comparison between trials and we would highly support development of a consensus on, ideally, one measure. We agree that mean amplitude of glycemic excursions has several limitations: not measuring excursions smaller than the standard deviation as well as not taking the frequency of excursions into account. Therefore, our group has developed a new measure that we believe overcomes these limitations and has proven robust in diabetes4 and intensive care unit populations5: mean absolute glucose change (MAG). This is a simple summation of all changes in glucose per unit of time and can be taken from self-monitored blood glucose profiles, blood glucoses, and continuous glucose monitoring traces. To investigate the influence of different measurement methods, we also measured the relation between oxidative stress and continuous net glycemic action (reflected as CONGA-1, CONGA-2, CONGA-4) as well as mean of daily differences originally in the subject study.2 As expected, all GV measures were highly correlated and, also for these additional measures, we were not able to show a correlation with oxidative stress in our population. To improve readability, we decided not to include them in the paper. Picconi and colleagues suggest that our outcomes might have been influenced by insulin secretagogues and other drugs used in our study population. It could be that the formation of oxidative stress is prohibited by these medicaments but, as mentioned, our population showed optimal glycemic control. The results of our study therefore reflect the current situation of diabetes regulation, which relies on polypharmacy. Moreover, it is plausible that mean glucose is the main driver of oxidative stress6 and not the way by which that mean glucose is reached. Noteworthy in this perspective is that the patients in the original Monnier article7 were allowed to use insulin secretagogues, which did not seem to influence the positive correlation found between GV and oxidative stress. In conclusion, we support that future studies use standardized GV measures, for which we propose the MAG, and, in case of isoprostane measurement, high-performance liquid chromatography tandem mass spectrometry to measure oxidative stress.

Normal Reference Range for Mean Tissue Glucose and Glycemic Variability Derived from Continuous Glucose Monitoring for Subjects Without Diabetes in Different Ethnic Groups

Glycemic variability has been proposed as a contributing factor in the development of diabetes complications. Multiple measures exist to calculate the magnitude of glycemic variability, but normative ranges for subjects without diabetes have not been described. For treatment targets and clinical research we present normative ranges for published measures of glycemic variability. Seventy-eight subjects without diabetes having a fasting plasma glucose of <120 mg/dL (6.7 mmol/L) underwent up to 72 h of continuous glucose monitoring (CGM) with a Medtronic Minimed (Northridge, CA) CGMS(®) Gold device. Glycemic variability was calculated using EasyGV(©) software (available free for non-commercial use at www.easygv.co.uk ), a custom program that calculates the SD, M-value, mean amplitude of glycemic excursions (MAGE), average daily risk ratio (ADRR), Lability Index (LI), J-Index, Low Blood Glucose Index (LBGI), High Blood Glucose Index (HBGI), continuous overlapping net glycemic action (CONGA), mean of daily differences (MODD), Glycemic Risk Assessment in Diabetes Equation (GRADE), and mean absolute glucose (MAG). Eight CGM traces were excluded because there were inadequate data. From the remaining 70 traces, normative reference ranges (mean±2 SD) for glycemic variability were calculated: SD, 0-3.0; CONGA, 3.6-5.5; LI, 0.0-4.7; J-Index, 4.7-23.6; LBGI, 0.0-6.9; HBGI, 0.0-7.7; GRADE, 0.0-4.7; MODD, 0.0-3.5; MAGE-CGM, 0.0-2.8; ADDR, 0.0-8.7; M-value, 0.0-12.5; and MAG, 0.5-2.2. We present normative ranges for measures of glycemic variability in adult subjects without diabetes for use in clinical care and academic research.