Background and Aims

Nocturnal hypoglycemia is most feared in children with diabetes, though often unrecognized. Heart rate variability (HRV) reflects activation of the autonomic nervous system. As hypoglycemia and declining glucose induce sympathoadrenal activation, changes of HRV could be assumed. Hence, the aim of this study was to evaluate, if HRV changes occur before the onset of nocturnal hypoglycemia in children with type 1 diabetes (T1D).

Methods

Children aged 1 to 18 y, with T1D for > 6 months, participated in an observational study using continuous glucose monitoring (CGM) during 5 days. Simultaneously, Holter ECG was recorded each night. HRV parameters such as time (SDNN) and frequency (VLF and total power) domain and sample entropy were calculated using a 6min moving window from 90mins before until the onset (=0mins) of nocturnal hypoglycemia (sensor glucose level <3.9mmol/l for ≥15mins). Mean HRV parameters at several timepoints were compared to timepoint 0 using a Kruskal-Wallis test.

Results

Twenty-five children (11f, mean (SD) age 13.5y (2.5) participated in the study. 33 hypoglycemic events with concomitant ECG recording occurred. Mean SDNN increased from 79ms at -70mins and 83ms at -30mins to 89ms at 0mins (p=0.002 and 0.01), as did VLF (from 3158ms and 4366ms to 5153ms; p=0.0001, 0.0002) and total power (from 7960ms and 9663ms to 10700ms; p=0.0002, 0.0081). Sample entropy decreased from 1.76 and 1.75 to 1.65 (p=0.001, 0.001).

Conclusions

Significant changes of multiple HRV variables as early as 70 mins before the occurence of nocturnal hypoglycemia were documented. This indicates that HRV changes may be used to help predict nocturnal hypoglycemia.

Background and Aims

While data on the efficacy and safety of CGM is evident across a broad age spectrum, there is a limited assessment in very young children with type 1 diabetes (T1D). This study aimed to assess real-world data in this high-risk population, focusing on glycemic variability and time in ranges.

Methods

The study adopted a prospective, multi-national, registry-based population cohort design to compare glycemia metrics over 12 months between very young children with T1D using real-time CGM and those using intermittent fingerstick blood glucose monitoring (BGM) alone. Major eligibility criteria included T1D diagnosed at least 6 months prior, age 1 to 7 years, insulin pump therapy for at least 3 months and at least 10 days of CGM/BGM data. The primary endpoint was assessment of glycemic variability as measured by the difference in coefficient of variation (CV) between the CGM users and BGM cohort and time in ranges as other pre-specified endpoints calculated for each group.

The trial is registered with Clinicaltrials.gov: NCT04558710

Results

Data from 229 individuals (44% were female, mean age 5.1±1.6 years) from 15 centers were analyzed.Results of the primary and pre-specified secondary efficacy outcomes are presented in Table 1 over the full 24-hour period.

Conclusions

The use of CGM was associated with reduced glucose fluctuations and increased time in range and decreased time above and below range. In our study, very young children with T1D using CGM were more likely to approach the targeted glycemia as measured by time in range.

Background and Aims

Background and Aims: GMI can estimate average HbA1c based on the mean glucose from at least 14 days of CGM data. Using CGM data, one can also assess glycemic variability with coefficient of variation for glucose (CV, glucose SD/mean x 100%). Given that glycemic levels fluctuate across the lifespan, we aimed to evaluate the accuracy of GMI in estimating average A1c based on CV and age.

Methods

Methods: GMI was calculated from over 300 hours of baseline, masked CGM data collected before starting RT-CGM in 3 US studies in persons with T1D (SENCE [N=143, 2-6y] / CITY [N=152, 14-24y] / WISDM [N=203, 60-86y]). GMI (%) was calculated using the formula: 3.31 + 0.02392 x [mean glucose in mg/dL]. We assessed associations between baseline HbA1c, from a central laboratory, and CGM metrics (mean glucose, TIR (70-180mg/dL) by study and stratified by CV (≤36,>36).

Results

Results: Across all studies, lab HbA1c was strongly correlated with CGM metrics (p<0.0001). When CV>36 vs. CV≤36, GMI underestimated average HbA1c in all studies; greatest differences were seen in adolescents/young adults (CITY). Correlations between HbA1c and mean glucose and TIR were stronger when CV>36 (Table).

Conclusions

Conclusions: These 3 studies affirm that CGM metrics, both mean glucose and TIR, can estimate laboratory HbA1c. Notably, glycemic variability (CV) alters these associations across the lifespan in persons with T1D. Caution should be used in interpreting GMI in those with high variability due to risk for under-approximation; changes in RBC kinetics likely also impact these associations.

Background and Aims

Cybersecurity in eHealthcare is a growing concern, and in particular, in diabetes care, where vast amounts of data are being generated by the recent surge in wearable devices, such as continuous glucose monitoring (CGM). The use of open-source platforms, for example, NightScout and OpenAPS, has increased the risk of data breaches and might represent a privacy concern for some users. In this work, we aim to demonstrate that it is possible to identify CGM data at an individual level by standard machine learning techniques.

Methods

The publically available REPLACE-BG dataset containing 226 adult participants with type 1 diabetes wearing CGM over 6 months was used. A support vector machine (SVM) binary classifier aiming to determine if a CGM data stream belongs to an individual was trained and tested for each subject in the dataset. Eleven standard glycaemic metrics were employed to generate the feature vector for the SVM. Data points in the training and testing datasets were generated by evaluating the selected glycaemic metrics over multiple incidences of one-month time windows. In order to increase the number of data points, a sliding time window was employed.

Results

The mean and standard deviation of sensitivity, specificity, and accuracy results on the testing data set were 0.80±0.24, 0.97±0.03, and 0.89±0.12, respectively.

Conclusions

This work demonstrates that it is possible to determine with relatively high accuracy if a CGM data stream belongs to an individual. The proposed approach can be used as a digital CGM ‘fingerprint’ or for detecting glycaemic changes within an individual (e.g. illness).

Background and Aims

The Wireless Innovation for Seniors with Diabetes Mellitus (WISDM) randomized clinical trial demonstrated reductions in hypoglycemia with CGM compared with blood glucose monitoring over 6 months among older adults with T1D. The aim of this analysis was to evaluate whether the treatment effect was influenced by the frequency of pre-study self-monitoring of blood glucose (SMBG).

Methods

203 older adults age ≥ 60 yrs were randomly assigned to use CGM or BGM for 26 weeks. CGM-measured outcomes stratified by baseline SMBG (<4x/d vs. >4x/d) were calculated at baseline (masked CGM) and during follow-up (masked CGM for BGM group).

Results

Among participants with baseline SMBG < 4 times/day, median percent time <70 mg/dL was reduced from 7.2% at baseline to 1.8% at 26 weeks in the CGM group and was 4.8% at baseline and 5.1% at 26 weeks in the BGM group. A positive treatment effect also was observed among participants with baseline SMBG ≥4 times/day (Table). After 26 weeks of CGM use, 78% and 85% of participants were using CGM six or more days/wk in those with baseline SMBG < 4 times/day and SMBG ≥4 times/day, respectively.

Conclusions

Prior to changing restrictions during the COVID-19 pandemic, public insurance (Medicare) for older adults in the US required SMBG ≥4 times/day for approval of CGM. These study results show CGM can be effectively used to reduce hypoglycemia irrespective of SMBG history.

Background and Aims

The prospective, multi-center PROMISE Study was conducted from 27Dec2018-08May2020 in 181 adults ≥18 years with diabetes at 8 US clinical sites to evaluate the safety and accuracy of the next generation Eversense CGM System for up to 180 days.

Methods

During 10 clinic visits between day 1-180 lasting up to 10 hours, accuracy between 40-400mg/dL was assessed comparing CGM and reference glucose values from Yellow Springs Instruments (YSI), including hyperglycemia and hypoglycemia challenges. A total of 279 sensors were placed (85 and 96 subjects had 1 and 2 sensors inserted in the arm, respectively, with 2 replaced sensors) for 558 insertion/removal procedures. Two calibrations/day to day 21 were prompted, after which it primarily prompts 1 calibration/day.

Results

The mean age was 48.6 years (range 18-77years). Accuracy analyses were based on 49,613 matched pairs over 180 days. Percent CGM readings within 15/15% and 20/20% of YSI values was 85.6% and 92.9%, respectively, and the overall Mean Absolute Relative Difference (MARD) was 9.1%. MARDs across different time periods over 180 days were <9.6% (Table), except for the beginning and end of sensor life. The confirmed hypoglycemia detection rate at 70mg/dL and 60mg/dL were 93% and 87%, respectively. Hyperglycemia at 180mg/dL was detected in 99%. There were no device or insertion/removal procedure-related serious adverse events. Two mild skin infections occurred for a rate of 0.36% per procedure.

Conclusions

These results indicate the next generation Eversense long-term implantable CGM System has sustained accuracy and safety up to 180 days with mainly one calibration/day.

Background and Aims

To investigate temporal trends in the use of insulin pumps (CSII) and continuous glucose monitoring (CGM) in children with type 1 diabetes (T1D) ≤6 years of age from 2000 to 2020. Furthermore, to study changes in HbA1c and event rates of severe hypoglycaemia over time in a prospective, multicentre diabetes patient follow-up (DPV) registry.

Methods

Children with T1D ≤6 years (≥ 6 months at onset), registered in DPV from 2000 to 2020 were included. Temporal trends in diabetes technology use were studied using repeated measurements logistic regression models considering sex, diabetes duration (≤1 year, >1 year), age (<2, 2-<4, ≥4 years), migration background and the interaction of age*year as covariates. Changes in HbA1c and event rates of severe hypoglycaemia were studied using linear and negative binomial regression models.

Results

Among 16,907 children, use of CSII increased consistently from 2000 (<1%) to 2020 (86%) with a most significant increase in very young children <2 years (2020: 96% (95%-CI:93-98%) vs 90% (88-92% 2-<4 years), 82% (80-84% 4-≤6 years)). Use of CGM increased from 2% in 2016 to 73% in 2020 (76% (69-81%) <2 years to 72% (70-74%) in 4-≤6 years). HbA1c was stable at 7.7%. Event rates of severe hypoglycaemia decreased significantly from 0.35 events/PY (0.30-0.40PY) in 2000 to 0.07 events/PY (0.06-0.09PY) in 2020.

Conclusions

These registry data show consistent increases in CSII and CGM use in T1D children ≤6 years during the last 2 decades. Increasing use of diabetes technology might be associated with the significant decrease in severe hypoglycaemia.

Background and Aims

This prospective, observational cohort study was designed to measure the change in HbA1c over 3-6 months in adults with T2DM on a basal-bolus insulin regimen using the FreeStyle Libre Flash Glucose Monitoring System^TM compared to self-monitoring blood glucose (SMBG), in a real-world setting.

Methods

A total of 322 patients (109 intervention, 213 control) with T2DM from 16 hospital sites in Italy were enrolled. To minimise selection bias, all eligible FreeStyle Libre users were included, matched to SMBG patients by HbA1c (within ±0.5%) and study site. The study population included adults on a basal-bolus insulin regimen for ≥1 year, with HbA1c 8.0-12.0% (64-108 mmol/mol), who were either new to FreeStyle Libre and planned to use it for ≥3 months (intervention) or planned to continue with SMBG (control) in a 1:2 ratio. On average, HbA1c was 8.9±0.8% (73.9±8.8 mmol/mol), age 67.2±10.0 years, BMI 30.5±6.5 kg/m² and average duration of insulin use 8.6±6.6 years (mean±SD), 56.2% were male.

Results

After 3-6 months, 234 complete case patients (83 intervention, 151 control) demonstrated significantly reduced HbA1c for FreeStyle Libre users compared to SMBG by 0.30% (95%CI: -0.53, -0.07), p=0.0113. Considering all 322 patients (109 intervention, 213 control), with imputed missing HbA1c values, HbA1c was also significantly reduced for FreeStyle Libre users vs. SMBG by 0.28% (95%CI: -0.50, -0.05), p=0.0199. The difference remains statistically significant after adjusting for the confounders.

Conclusions

This real-world, prospective cohort study concluded that people with T2DM on basal-bolus insulin, using FreeStyle Libre for 3-6 months significantly reduced HbA1c compared to SMBG.

Background and Aims

An international consensus proposed time in range 70–180 mg/dl (TIR), below (<70 mg/dl) and above (>180 mg/dl) range as compound metrics of glycemic control to complement HbA1c regardless of age in individuals with Type 1 Diabetes (T1D). The aim of this study was to evaluate real-world data from an international network for pediatric diabetes centers (SWEET).

Methods

We retrospectively analyzed data from four age groups: 1–6y (n=185), 7–13y (n=1,195), 14–17y (n=996) and 18-21y (n=347) with T1D using CGM. Linear regression models adjusted for gender, diabetes duration, age, BMI and insulin therapy modality were performed to identify potential predictors of TIR.

Results

CGM data from 2,723 individuals (mean 229 sensor-days/person) from 22 centers were analyzed. Overall median TIR was 54.5%. Time in/below/above ranges were 59.3/3.0/37.9% among 1–6y, 57.4/3.0/38.5% among 7–13y, 51.7/3.6/42.8% among 14–17y and 50.7/4.7/42.5% among 18-21y. We observed a significant positive association between TIR and insulin therapy modality, an inverse association between T1D duration and TIR, while there was no association with gender or BMI. Adjusted mean (95%CI) TIR was 57% (54;59) in insulin pump users and 51% (47;54) in non-users. Pearson correlation coefficients showed strong correlations between HbA1c and TIR (R=-0.746, P<0.0001).

Conclusions

Data from the SWEET registry demonstrate that only a minority of young individuals with T1D achieve recommended goals for TIR. We observed higher TIR at younger age groups, a significant decline in TIR with longer T1D duration and a positive association with insulin pump therapy.