bims-glumda 2025-04-13 papers

bims-glumda

Biomed News

on CGM data in management of diabetes

Issue of 2025–04–13
twenty papers selected by
Mott Given

Erratum for "Accuracy of the Freestyle Libre 3 Continuous Glucose Monitoring System in Hypo- and Euglycemic Cats".
A Comparative Analysis of Glycemic Metrics Derived From Three Continuous Glucose Monitoring Systems.
Optimal Alphabet for Encoding Daily Continuous Glucose Monitoring Time Series Into Compressed Text: Correspondence.
Future horizons in diabetes: integrating AI and personalized care.
Dialysis Nurses' Experiences with Continuous Glucose Monitoring in Hemodialysis Patients Treated with Insulin.
Advances in Continuous Glucose Monitoring: Clinical Applications.
Accounting for Hypoglycemia Treatments in Continuous Glucose Metrics.
Significant Hyperglycemia in Hospitalized Total Pancreatectomy Patients: Results of a Pilot Prospective Continuous Glucose Monitoring Study.
Chronobiologically-informed features from CGM data provide unique information for XGBoost prediction of longer-term glycemic dysregulation in 8,000 individuals with type-2 diabetes.
Circulating Polyunsaturated Fatty Acids and Glycemic Dynamics Assessed by Continuous Glucose Monitoring: A 13-Year Prospective Cohort Study.
Nanocomposite Hydrogel-Based Optical Fiber Probe for Continuous Glucose Sensing.
A Randomized Clinical Trial of a Culturally Tailored Diabetes Education Curriculum With and Without Real-Time Continuous Glucose Monitoring in a Latino Population With Type 2 Diabetes: The CUT-DM With Continuous Glucose Monitoring Study.
Early Detection of Elevated Ketone Bodies in Type 1 Diabetes Using Insulin and Glucose Dynamics Across Age Groups: Model Development Study.
Evidence for improved glucose metrics and perinatal outcomes with continuous glucose monitoring compared to self-monitoring in diabetes during pregnancy.
A pretrained transformer model for decoding individual glucose dynamics from continuous glucose monitoring data.
Evaluation of a Continuous Glucose Monitoring-Based Insulin Titration Protocol for Inpatients With Type 2 Diabetes in Nonintensive Care Unit Settings.
Personalized glucose forecasting for people with type 1 diabetes using large language models.
Evaluation of pharmacist-led management of type 2 diabetes using personal continuous glucose monitors across a large tertiary academic health system.
Screening for Impaired Glucose Homeostasis: A Novel Metric of Glycemic Control.
Exploring the potential of deep learning models integrating transformer and LSTM in predicting blood glucose levels for T1D patients.

J Vet Intern Med. 2025 May-Jun;39(3):39(3): e70085

Erratum for "Accuracy of the Freestyle Libre 3 Continuous Glucose Monitoring System in Hypo- and Euglycemic Cats".

DOI: https://doi.org/10.1111/jvim.70085
Diabetes Care. 2025 Apr 08. pii: dc250129. [Epub ahead of print]

A Comparative Analysis of Glycemic Metrics Derived From Three Continuous Glucose Monitoring Systems.

Guido Freckmann, Stephanie Wehrstedt, Manuel Eichenlaub, Stefan Pleus, Manuela Link, Nina Jendrike, Sükrü Öter, Derek Brandt, Cornelia Haug, Delia Waldenmaier.

OBJECTIVE: This study analyzed the differences in continuous glucose monitoring (CGM)-derived metrics among three current-generation systems and evaluated their impact on therapeutic decision-making.
RESEARCH DESIGN AND METHODS: Twenty-three participants wore the FreeStyle Libre 3, Dexcom G7, and Medtronic Simplera CGM systems for 14 days in parallel. CGM metrics were calculated for each participant and CGM system separately.
RESULTS: The apparent glucose profile was influenced by the used CGM system, resulting in substantially different glycemic metrics among the three systems. Agreement between FreeStyle Libre 3 and Dexcom G7 was higher than with Medtronic Simplera, which showed lower glucose levels, on average. There were marked intraparticipant discrepancies that would have resulted in different therapeutic recommendations.
CONCLUSIONS: The CGM systems indicated discordant glycemic metrics, which should be considered in diabetes therapy. Different CGM systems should provide the same glucose readings and CGM-derived metrics when used by the same person.

DOI: https://doi.org/10.2337/dc25-0129
J Diabetes Sci Technol. 2025 Apr 10. 19322968251334665

Optimal Alphabet for Encoding Daily Continuous Glucose Monitoring Time Series Into Compressed Text: Correspondence.

Hinpetch Daungsupawong, Viroj Wiwanitkit.



Keywords:  glucose; monitoring; time

DOI:  https://doi.org/10.1177/19322968251334665
Front Endocrinol (Lausanne). 2025 ;16 1583227

Future horizons in diabetes: integrating AI and personalized care.

Kaiqi Zhang, Yun Qi, Wenjun Wang, Xinyi Tian, Jiahui Wang, Lili Xu, Xu Zhai.

  Diabetes is a global health crisis with rising incidence, mortality, and economic burden. Traditional markers like HbA1c are insufficient for capturing short-term glycemic fluctuations, leading to the need for more precise metrics such as Glucose Variability (GV) and Time in Range (TIR). Continuous Glucose Monitoring (CGM) and AI integration offer real-time data analytics and personalized treatment plans, enhancing glycemic control and reducing complications. The combination of transcutaneous auricular vagus nerve stimulation (taVNS) with artificial Intelligence (AI) further optimizes glucose regulation and addresses comorbidities. Empowering patients through AI-driven self-management and community support is crucial for sustainable improvements. Future horizons in diabetes care must focus on overcoming challenges in data privacy, algorithmic bias, device interoperability, and equity in AI-driven care while integrating these innovations into healthcare systems to improve patient outcomes and quality of life.

Keywords:  artificial intelligence; continuous glucose monitoring; diabetes blood glucose; glucose variability; taVNS; time in range

DOI:  https://doi.org/10.3389/fendo.2025.1583227
Stud Health Technol Inform. 2025 Apr 08. 323 317-321

Dialysis Nurses' Experiences with Continuous Glucose Monitoring in Hemodialysis Patients Treated with Insulin.

Sisse H Laursen, Inger V Kristensen, Hanne R Larsen, Peter Vestergaard, Morten H Jensen, Ole K Hejlesen.

  This study examines dialysis nurses' perspectives on the use of continuous glucose monitoring (CGM) in hemodialysis (HD) patients with insulin-treated diabetes. Through eight semi-structured interviews, nurses highlighted how CGM improved the patients' and nurses' ability to monitor patients' glucose levels, enhancing patient engagement and nursing practices. The nurses emphasized the value of real-time glucose data during dialysis sessions, allowing for timely adjustments and better glycemic control. Despite these advantages, they also noted challenges, including a lack of knowledge regarding CGM technology. Overall, the nurses viewed CGM as a beneficial tool, providing a clearer understanding of patients' glucose patterns. Furthermore, the findings reveal that CGM fosters better communication and awareness among healthcare professionals and patients, ultimately improving the care provided to insulin-treated HD patients.

Keywords:  Continuous glucose monitoring; hemodialysis patients; insulin-treated diabetes; nursing perspective

DOI:  https://doi.org/10.3233/SHTI250103
Endocrinol Metab (Seoul). 2025 Apr 08.

Advances in Continuous Glucose Monitoring: Clinical Applications.

So Yoon Kwon, Jun Sung Moon.

  Continuous glucose monitoring (CGM) has revolutionized diabetes management, significantly enhancing glycemic control across diverse patient populations. Recent evidence supports its effectiveness in both type 1 and type 2 diabetes management, with benefits extending beyond traditional glucose monitoring approaches. CGM has demonstrated substantial improvements in glycemic control across multiple metrics. Studies report consistent glycosylated hemoglobin reductions of 0.25%-3.0% and notable time in range improvements of 15%-34%. CGM effectively reduces hypoglycemic events, with studies reporting significant reductions in time spent in hypoglycemia. CGM also serves as an educational tool for lifestyle modification, providing real-time feedback that helps patients understand how diet and physical activity affect glucose levels. While skin-related complications remain a concern, technological advancements have addressed many initial concerns. High satisfaction rates and long-term use suggest that device-related issues are manageable with proper education and support. Despite high initial costs, CGM's prevention of complications and hospitalizations ultimately reduces healthcare expenditures. With appropriate training and support, CGM represents a transformative technology for comprehensive diabetes care.

Keywords:  Blood glucose self-monitoring; Continuous glucose monitoring; Diabetes mellitus

DOI:  https://doi.org/10.3803/EnM.2025.2370
J Diabetes Sci Technol. 2025 Apr 05. 19322968251329952

Accounting for Hypoglycemia Treatments in Continuous Glucose Metrics.

Elliott C Pryor, Anas El Fathi, Marc D Breton.

   BACKGROUND: Continuous glucose monitoring (CGM) is increasingly used in the management of diabetes, providing dense data for patients and clinical providers to review and identify patterns and trends in blood glucose. However, behavioral factors like hypoglycemia treatments (HTs) are not captured in CGM data. Hypoglycemia treatments, by definition, reduce the visibility (frequency and duration) of hypoglycemia exposure recorded by CGM, which can lead to errors in treatment management when relying solely on CGM metrics.
METHODS: We propose a method to incorporate HTs into CGM-based metrics and standardize hypoglycemia exposure quantification for a variety of HT behaviors; specifically (1) treatment proactiveness and (2) potential severity of the avoided hypoglycemia. In addition, we introduce an HT detector to identify instances of HT using in CGM data that otherwise lack HT documentation. We then use the HT-modified hypoglycemia metrics in a previously published run-to-run treatment adaptation system using CGM-based metrics.
RESULTS: Using in-silico data to correct time-below-range with HT, we reconstruct the avoided hypoglycemia exposure with high fidelity (R2 = .94). Our HT detector has an F1 score of 0.72 on clinical data with labeled HT. In the example run-to-run application, we reduce the average number of HT per day from 3.3 in the HT-unaware system to 1.6, while maintaining 84% time in 70 to 180 mg/dL.
CONCLUSION: This new metric integrates HT behaviors into CGM-based analysis, offering a behavior-sensitive measure of hypoglycemia exposure for more robust T1D management. Our results show that HT can be seamlessly incorporated into existing CGM methods, enhancing treatment insights by accounting for HT variability.

Keywords:  automated insulin delivery; continuous glucose monitoring; hypoglycemia fear; hypoglycemia treatment; outcome metrics

DOI:  https://doi.org/10.1177/19322968251329952
Diabetes Technol Ther. 2025 Apr 10.

Significant Hyperglycemia in Hospitalized Total Pancreatectomy Patients: Results of a Pilot Prospective Continuous Glucose Monitoring Study.

Shafaq R Rizvi, Ayan Bannerjee, Ravinder J Kaur, Sandeep Gupta, Corey Kurek, Donna Desjardins, Michael L Kendrick, Mark J Truty, Yogish C Kudva.

  Background: Glycemic control has been studied in hospitalized patients with type 1 diabetes (T1D) or type 2 diabetes (T2D) and more recently type 3C diabetes who had total pancreatectomy (TP) and islet autotransplantation (IAT). To our knowledge, we report the first study using continuous glucose monitoring (CGM) to assess glucose control in TP without IAT during hospitalization. Methods: We completed blinded CGM (Dexcom G6 PRO) studies in 27 subjects in a nonintensive care unit setting who had TP at Mayo Clinic, Rochester, and compared two cohorts (CGM data < 30 days of TP [cohort 1] vs. ≥ 30 days of TP [cohort 2]). CGM glucose metrics were calculated as per American Diabetes Association guidelines for hospitalized patients with diabetes mellitus. CGM values were compared with point-of-care testing glucose within 5 min. Results: The baseline characteristics were not significantly different between the cohorts Table 1. The average TIGHT range (140 mg/dL-180 mg/dL) on day 1 was 13 ± 12.0 % in cohort 1 and 17.4 ± 27.7 % in cohort 2 without any statistically significant difference, while average TIGHT percentage on days 2-10 was 23.4% ± 25.8% and 20.6% ± 25.7% in cohort 1 and cohort 2, respectively (no statistically significant difference). Both cohorts spent the majority of (>50%) time above target range on day 1 and days 2-10. There were no significant differences in CGM metrics between the two cohorts. Overall, mean absolute relative difference was 19.6% ± 10, and number of readings meeting % 20/20, was 68% with least accuracy on day 1 of sensor insertion. There were no device-related adverse events. Conclusion: Hospitalized TP patients spend considerable time above 180 mg/dL demonstrating the unmet need of optimal glucose monitoring in this cohort.

Keywords:  continuous glucose monitoring; hospitalization; hyperglycemia; point-of-care testing glucose; total pancreatectomy

DOI:  https://doi.org/10.1089/dia.2024.0583
PLOS Digit Health. 2025 Apr;4(4): e0000815

Chronobiologically-informed features from CGM data provide unique information for XGBoost prediction of longer-term glycemic dysregulation in 8,000 individuals with type-2 diabetes.

Jamison H Burks, Leslie Joe, Karina Kanjaria, Carlos Monsivais, Kate O'laughlin, Benjamin L Smarr.

Type 2 Diabetes causes dysregulation of blood glucose, which leads to long-term, multi-tissue damage. Continuous glucose monitoring devices are commercially available and used to track glucose at high temporal resolution so that individuals can make informed decisions about their metabolic health. Algorithms processing these continuous data have also been developed that can predict glycemic excursion in the near future. These data might also support prediction of glycemic stability over longer time horizons. In this work, we leverage longitudinal Dexcom continuous glucose monitoring data to test the hypothesis that additional information about glycemic stability comes from chronobiologically-informed features. We develop a computationally efficient multi-timescale complexity index, and find that inclusion of time-of-day complexity features increases the performance of an out-of-the-box XGBoost model in predicting the change in glucose across days. These findings support the use of chronobiologically-inspired and explainable features to improve glucose prediction algorithms with relatively long time-horizons.

DOI: https://doi.org/10.1371/journal.pdig.0000815
Mol Nutr Food Res. 2025 Apr 07. e70056

Circulating Polyunsaturated Fatty Acids and Glycemic Dynamics Assessed by Continuous Glucose Monitoring: A 13-Year Prospective Cohort Study.

Jingnan Chen, Yan Yan, Zelei Miao, Ke Zhang, Congmei Xiao, Xinxiu Liang, Fengzhe Xu, Wanglong Gou, Yuanqing Fu, Haili Zhong, Yu-Ming Chen, Ju-Sheng Zheng.

  Polyunsaturated fatty acids (PUFA) have been proposed to be involved in the diabetes etiology, but the associations between PUFA biomarkers and glycemic dynamics are unclear. We aimed to investigate the prospective associations of PUFA with glycemic dynamics assessed by continuous glucose monitoring (CGM). A total of 1216 middle-aged and elderly participants were included. We performed a multivariable-adjusted linear regression to investigate the association between baseline measurement of erythrocyte fatty acids and follow-up CGM profiles. During a median 13.6-year follow-up, the highest (vs. lowest) quartile of γ-linoleic acid (GLA) was associated with an increment (in standard deviation unit) of 0.27 (95% CI: 0.10, 0.42) for time above range, 0.31 (0.14, 0.47) for high blood glucose index, 0.26 (0.10, 0.41) for mean of daily differences, and 0.31 (0.15, 0.46) for mean blood glucose (MBG). In contrast, higher linoleic acid (LA) was associated with lower nocturnal MBG, with beta (95% CI) -0.23 (-0.38, -0.08); higher arachidonic acid (AA) was associated with lower diurnal MBG (-0.25 [95% CI: -0.41, -0.10]). None of the n-3 PUFA biomarkers was associated with CGM measures. Circulating GLA was positively associated with high-glucose level, glycemic variability, and mean glucose measures, while LA and AA were inversely associated with nocturnal and diurnal MBG, respectively. Our findings may provide new insights into the relationship between PUFA metabolism and glycemic control.

Keywords:  continuous glucose monitoring; glycemic control; glycemic variability; polyunsaturated fatty acids; prospective cohort study

DOI:  https://doi.org/10.1002/mnfr.70056
Small Sci. 2024 Feb;4(2): 2300189

Nanocomposite Hydrogel-Based Optical Fiber Probe for Continuous Glucose Sensing.

Israr Ahmed, Said El Turk, Amal Al Ghaferi, Yarjan Abdul Samad, Haider Butt.

  Diabetes mellitus (DM) presents a substantial global health concern due to elevated blood glucose levels, necessitating an affordable, rapid, and reliable continuous glucose monitoring (CGM) solution. In this pursuit, a pioneering approach is introduced utilizing optical fiber (OF) sensors based on nanocomposite photonic hydrogel functionalized with phenylboronic acid (PBA) for precise CGM. The fabrication of OF sensors involves a streamlined process, involving one-step polymerization of PBA-based hydrogel onto a commercial fiber tip and the integration of gold nanoparticles (AuNPs) via a simple dipping process. These sensors offer robust performance within the physiological glucose range (0-20 mm), exhibiting a remarkable 25% increase in transmission intensity and a 4 nm blue shift in the surface plasmon resonance with increasing glucose concentration. Additionally, there is a noticeable elevation in reflection intensity, affirming the sensor's suitability for remote sensing applications. These results are further validated using a green laser, underlining the method's reliability. The sensors exhibit a swift 30 s response time, followed by a 5 min saturation period, for all measurements. Practicality is demonstrated through smartphone readouts, utilizing the phone's photodiode to measure optical power changes concerning various glucose concentrations. These OF sensors hold great promise for CGM integration, enhancing diabetic management.

Keywords:  glucose sensing; hydrogels; nanocomposite; optical fiber

DOI:  https://doi.org/10.1002/smsc.202300189
J Diabetes Sci Technol. 2025 Apr 10. 19322968251331526

A Randomized Clinical Trial of a Culturally Tailored Diabetes Education Curriculum With and Without Real-Time Continuous Glucose Monitoring in a Latino Population With Type 2 Diabetes: The CUT-DM With Continuous Glucose Monitoring Study.

Nicole Ehrhardt, Laura Montour, Peter Berberian, Ana Gabriela Vasconcelos, Bryan Comstock, Lorena Alarcon-Casas Wright.

   BACKGROUND: Data on culturally tailored diabetes education with and without real-time continuous glucose monitoring (RT-CGM) in Latinos with type 2 diabetes, who are not on intensive insulin management, is lacking.
RESEARCH DESIGN AND METHODS: This is an open-label randomized control trial of Latinos with uncontrolled (HbA1c > 8.0%) type 2 diabetes conducted in a Federally Qualified Health Center (FQHC). All participants received 12 one-hour culturally tailored education sessions. Patients were randomized (1:1) to education sessions only (blinded CGM) or cyclic (50 days wear: 10 days on, 7 days off) RT-CGM. The primary outcome was a change in HbA1c from baseline to 12 weeks in those with or without CGM. Secondary outcomes included 24-week HbA1c, CGM, and metabolic parameters.
RESULTS: Participants (n = 120) were 46 years old on average, 44% female, 98% preferred Spanish language, 30% with income <$25,000, 68% uninsured and 26% using basal insulin only. Mean 1-hour session attendance and RT-CGM wear was 7.0 (±4.4) and 27.9 (±20.5) days, respectively. Mean baseline HbA1c was 10.5% (±1.8). HbA1c reduced by 1.9% (95% confidence interval [CI]: 1.5-2.3) overall (P < .001). Participants in the RT-CGM group reduced HbA1c at 12 weeks by 2.3% (95% CI: 1.5-3.2) compared to 1.5% (95% CI: 0.6-2.3) in the blinded CGM group (P =.04). At 24 weeks, overall HbA1c reduction was maintained but between-group differences attenuated.
CONCLUSIONS: In a Latino type 2 diabetes population that was primarily noninsulin-requiring, virtually delivered, culturally tailored education improved HbA1c, with RT-CGM conferring greater improvement. RT-CGM should be an adjunctive therapy to diabetes education, irrespective of insulin use but continued cyclic CGM use may be needed for sustained effect.

Keywords:  continuous glucose monitoring; diabetes education; health equity; lifestyle changes; type 2 diabetes

DOI:  https://doi.org/10.1177/19322968251331526
JMIR Diabetes. 2025 Apr 10. 10 e67867

Early Detection of Elevated Ketone Bodies in Type 1 Diabetes Using Insulin and Glucose Dynamics Across Age Groups: Model Development Study.

Simon Cichosz, Clara Bender.

   Background: Diabetic ketoacidosis represents a significant and potentially life-threatening complication of diabetes, predominantly observed in individuals with type 1 diabetes (T1D). Studies have documented suboptimal adherence to diabetes management among children and adolescents, as evidenced by deficient ketone monitoring practices.
Objective: The aim of the study was to explore the potential for prediction of elevated ketone bodies from continuous glucose monitoring (CGM) and insulin data in pediatric and adult patients with T1D using a closed-loop system.
Methods: Participants used the Dexcom G6 CGM system and the iLet Bionic Pancreas system for insulin administration for up to 13 weeks. We used supervised binary classification machine learning, incorporating feature engineering to identify elevated ketone bodies (>0.6 mmol/L). Features were derived from CGM, insulin delivery data, and self-monitoring of blood glucose to develop an extreme gradient boosting-based prediction model. A total of 259 participants aged 6-79 years with over 49,000 days of full-time monitoring were included in the study.
Results: Among the participants, 1768 ketone samples were eligible for modeling, including 383 event samples with elevated ketone bodies (≥0.6 mmol/L). Insulin, self-monitoring of blood glucose, and current glucose measurements provided discriminative information on elevated ketone bodies (receiver operating characteristic area under the curve [ROC-AUC] 0.64-0.69). The CGM-derived features exhibited stronger discrimination (ROC-AUC 0.75-0.76). Integration of all feature types resulted in an ROC-AUC estimate of 0.82 (SD 0.01) and a precision recall-AUC of 0.53 (SD 0.03).
Conclusions: CGM and insulin data present a valuable avenue for early prediction of patients at risk of elevated ketone bodies. Furthermore, our findings indicate the potential application of such predictive models in both pediatric and adult populations with T1D.

Keywords:  diabetic complication; diabetic ketoacidosis; ketone level; machine learning; prediction model; type 1 diabetes

DOI:  https://doi.org/10.2196/67867
Am J Obstet Gynecol. 2025 Apr 09. pii: S0002-9378(25)00217-0. [Epub ahead of print]

Evidence for improved glucose metrics and perinatal outcomes with continuous glucose monitoring compared to self-monitoring in diabetes during pregnancy.

Jessica Burk, Glynis P Ross, Teri L Hernandez, Stephen Colagiuri, Arianne Sweeting.

   OBJECTIVE: Continuous glucose monitoring (CGM) is recommended for pregnant women with type 1 diabetes (T1D), due to associations with decreased HbA1c and large-for-gestational age (LGA). However, its benefit in type 2 diabetes (T2D) and gestational diabetes (GDM) is not established. This systematic review and meta-analysis compared usage of CGM to self-monitoring of blood glucose (SMBG) both across and within diabetes in pregnancy (DIP), and determined which glucose metrics are associated with perinatal outcomes, to potentially inform treatment targets in DIP.
DATA SOURCES: We searched Medline, Embase, CENTRAL, CINAHL and Scopus, from January 2003 to August 2024.
STUDY ELIGIBILITY CRITERIA: Randomized controlled trials and quasi-experimental studies comparing CGM with SMBG in DIP were included.
STUDY APPRAISAL AND SYNTHESIS METHODS: RCTs and quasi-experimental studies were analyzed separately. Data were extracted on CGM glucose metrics, HbA1c, rates of cesarean delivery, LGA, small-for-gestational age (SGA), neonatal hypoglycemia and neonatal intensive care unit (NICU) admission, summarized as mean differences (MD) or odds ratios (OR) with 95% Confidence Intervals (95%CI) and 95% Prediction Intervals (95%PI). Prespecified subgroup analyses were undertaken by DIP subtype, including duration of CGM use (continuous vs intermittent) for LGA. Certainty of evidence was assessed using the Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) framework.
RESULTS: Across DIP, CGM (vs SMBG) decreased HbA1c (MD -0.22% [95%CI: -0.37, -0.08]) (7 RCTs, moderate-certainty evidence). Within DIP, CGM use (vs SMBG) showed similar but stronger benefits in both T1D when used throughout pregnancy (HbA1c MD -0.18% [95%CI: -0.36, 0.00], LGA OR 0.51 [0.28, 0.90]) (1 RCT, high-certainty evidence), and GDM when used intermittently (HbA1c MD -0.18 [95%CI: -0.33, -0.02]) (5 RCTs, moderate-certainty evidence) and LGA (OR 0.46 [0.26, 0.81]) (1 quasi-experimental study, low-certainty evidence), with insufficient data for CGM benefit in T2D. Increased pregnancy %time-in-range (T1D) and decreased mean sensor glucose (T1D/GDM) were associated with decreased LGA.
CONCLUSIONS: Usage of CGM (vs SMBG) reduces HbA1c and possibly LGA across DIP. Greatest benefit was evidenced in T1D, followed by GDM, although CGM duration differed. Mean sensor glucose and pregnancy %time-in-range are important CGM metrics for reducing LGA.

Keywords:  Continuous glucose monitoring (CGM); Gestational diabetes (GDM); Large-for-gestational age (LGA); Meta-analysis; Systematic review; Type 1 diabetes (T1D) in pregnancy; Type 2 diabetes (T2D) in pregnancy

DOI:  https://doi.org/10.1016/j.ajog.2025.04.010
Natl Sci Rev. 2025 May;12(5): nwaf039

A pretrained transformer model for decoding individual glucose dynamics from continuous glucose monitoring data.

Yurun Lu, Dan Liu, Zhongming Liang, Rui Liu, Pei Chen, Yitong Liu, Jiachen Li, Zhanying Feng, Lei M Li, Bin Sheng, Weiping Jia, Luonan Chen, Huating Li, Yong Wang.

  Continuous glucose monitoring (CGM) technology has grown rapidly to track real-time blood glucose levels and trends with improved sensor accuracy. The ease of use and wide availability of CGM will facilitate safe and effective decision making for diabetes management. Here, we developed an attention-based deep learning model, CGMformer, pretrained on a well-controlled and diverse corpus of CGM data to represent individual's intrinsic metabolic state and enable clinical applications. During pretraining, CGMformer encodes glucose dynamics including glucose level, fluctuation, hyperglycemia, and hypoglycemia into latent space with self-supervised learning. It shows generalizability in imputing glucose value across five external datasets with different populations and metabolic states (MAE = 3.7 mg/dL). We then fine-tuned CGMformer towards a diverse panel of downstream tasks in the screening of diabetes and its complications using task-specific data, which demonstrated a consistently boosted predictive accuracy over direct fine-tuning on a single task (AUROC = 0.914 for type 2 diabetes (T2D) screening and 0.741 for complication screening). By learning an intrinsic representation of an individual's glucose dynamics, CGMformer classifies non-diabetic individuals into six clusters with elevated T2D risks, and identifies a specific cluster with lean body-shape but high risk of glucose metabolism disorders, which is overlooked by traditional glucose measurements. Furthermore, CGMformer achieves high accuracy in predicting an individual's postprandial glucose response with dietary modelling (Pearson correlation coefficient = 0.763) and helps personalized dietary recommendations. Overall, CGMformer pretrains a transformer neural network architecture to learn an intrinsic representation by borrowing information from a large amount of daily glucose profiles, and demonstrates predictive capabilities fine-tuned towards a broad range of downstream applications, holding promise for the early warning of T2D and recommendations for lifestyle modification in diabetes management.

Keywords:  continuous glucose monitoring; diabetes; glucose dynamics; pretrained model

DOI:  https://doi.org/10.1093/nsr/nwaf039
J Diabetes Sci Technol. 2025 Apr 05. 19322968251331628

Evaluation of a Continuous Glucose Monitoring-Based Insulin Titration Protocol for Inpatients With Type 2 Diabetes in Nonintensive Care Unit Settings.

Mikkel Thor Olsen, Ulrik Pedersen-Bjergaard, Signe Hjejle Jensen, Louise Mathorne Rasmussen, Carina Kirstine Klarskov, Birgitte Lindegaard, Jonas Askø Andersen, Hans Gottlieb, Suzanne Lunding, Katrine Bagge Hansen, Peter Lommer Kristensen.

   BACKGROUND: No widely adopted continuous glucose monitoring (CGM)-based insulin titration protocol exists, which may limit the effects of inpatient CGM on glycemic and clinical outcomes. We evaluate the acceptability and operability of the protocol proposed by Olsen et al for inpatients with type 2 diabetes in non-intensive care unit (non-ICU) settings.
METHOD: 7 inpatient diabetes team members, responsible for daily insulin titration, decided on insulin adjustments for 353 days. The members had the option to follow the CGM-based insulin protocol or override it for basal, prandial, and correctional insulin, separately, in 84 inpatients monitored by CGM. Questionnaires were used to evaluate the protocol's operability by the teams.
RESULTS: Of 456 basal insulin titration decisions, 439 (96.3%) adhered to the protocol. For prandial insulin, adherence rates were 83.9% (125/149) for breakfast, 87.2% (130/149) for lunch, and 92.6% (138/149) for dinner (p=0.163). All correctional insulin titrations adhered to the protocol. All team members expressed a preference for having a protocol for CGM-based insulin titration and rated the protocol's usability on a 1 to 10 scale, with mean scores (SD) of 8.7 (0.9) for basal insulin, 8.3 (1.4) for prandial insulin, and 7.4 (1.9) for correctional insulin.
CONCLUSIONS: The CGM-based insulin titration protocol by Olsen et al has been successfully implemented for titrating basal, prandial, and correctional insulin in inpatients with type 2 diabetes in non-ICU settings. It was highly accepted by inpatient diabetes teams and provides a framework for effective CGM implementation in these settings.

Keywords:  continuous glucose monitoring; hospital; inpatient; insulin titration; type 2 diabetes

DOI:  https://doi.org/10.1177/19322968251331628
Comput Methods Programs Biomed. 2025 Apr 02. pii: S0169-2607(25)00154-3. [Epub ahead of print]265 108737

Personalized glucose forecasting for people with type 1 diabetes using large language models.

Francisco J Lara-Abelenda, David Chushig-Muzo, Pablo Peiro-Corbacho, Ana M Wägner, Conceição Granja, Cristina Soguero-Ruiz.

   BACKGROUND AND OBJECTIVE: Type 1 Diabetes (T1D) is an autoimmune disease that requires exogenous insulin via Multiple Daily Injections (MDIs) or subcutaneous pumps to maintain targeted glucose levels. Despite the advances in Continuous Glucose Monitoring (CGM), controlling glucose levels remains challenging. Large Language Models (LLMs) have produced impressive results in text processing, but their performance with other data modalities remains unexplored. The aim of this study is three-fold. First, to evaluate the effectiveness of LLM-based models for glucose forecasting. Second, to compare the performance of different models for predicting glucose in T1D individuals treated with MDIs and pumps. Lastly, to create a personalized approach based on patient-specific training and adaptive model selection.
METHODS: CGM data from the T1DEXI study were used for forecasting glucose levels. Different predictive models were evaluated using the mean absolute error (MAE) and the root mean squared error and considering the Prediction Horizons (PHs) of 60, 90, and 120 min.
RESULTS: For short-term PHs (60 and 90 min), the personalized approach achieved the best results, with an average MAE of 15.7 and 20.2 for MDIs, and a MAE of 15.2 and 17.2 for pumps. For long-term PH (120 min), TIDE obtained an MAE of 19.8 for MDIs, whereas Patch-TST obtained a MAE of 18.5.
CONCLUSION: LLM-based models provided similar MAE values to state-of-the-art models but presented a reduced variability. The proposed personalized approach obtained the best results for short-term periods. Our work contributes to developing personalized glucose prediction models for enhancing glycemic control, reducing diabetes-related complications.

Keywords:  Continuous glucose monitor; GPT; Glucose forecasting; Large language models; Time series forecasting; Transformers; Type 1 diabetes

DOI:  https://doi.org/10.1016/j.cmpb.2025.108737
J Am Pharm Assoc (2003). 2025 Apr 04. pii: S1544-3191(25)00076-7. [Epub ahead of print] 102397

Evaluation of pharmacist-led management of type 2 diabetes using personal continuous glucose monitors across a large tertiary academic health system.

Christina Mnatzaganian, Mark Bounthavong, Wendelle Abalos, Truc Chau, Oluchi Nwosu, Andrew Yi, Ila Saunders, Panteha Kelly.

   BACKGROUND: There is limited but positive evidence of the impact of pharmacists in managing patients with type 2 diabetes (T2D) using a personal continuous glucose monitor (CGM). Previous studies have been limited to single clinic pilots or community pharmacies with small sample sizes.
OBJECTIVES: To evaluate the impact on glycemic outcomes of an innovative pharmacist-led Diabetes Management and Education Clinic (DMEC) on patients with T2D using a personal CGM.
PRACTICE DESCRIPTION: The DMEC operates in primary care settings in a large, tertiary academic health care system serving a large patient population. Pharmacists manage care for patients with T2D who are referred by primary care and specialty medical providers under a collaborative practice agreement.
PRACTICE INNOVATION: To use CGM data to guide decision making for clinical pharmacists seeing patients with T2D in the DMEC.
EVALUATION METHODS: This was a retrospective study conducted at the DMEC over two years. Demographics and glycemic outcomes were collected from the electronic medical record for patients who had a personal CGM prior to the initial clinic visit, supplied during the visit, or ordered as a prescription. A descriptive analysis was completed for this study.
RESULTS: DMEC pharmacists used CGMs to guide treatment decisions for 165 patients. The average hemoglobin A1c decreased by 1.48% at three months (p<0.001) and 1.74% at six months (p<0.001) after initial visit. Time in range improved by 8.2% at three months (p<0.001) and 12.1% at six months (p <0.001). The glucose management indicator decreased by 0.27% at three months (p<0.001) and 0.53% at six months (p<0.001). The average glucose decreased by 13.5 mg/dL at three months (p<0.001) and 18.8 mg/dL at six months (p<0.001).
CONCLUSION: Pharmacist-led management of T2D using personal CGMs can improve diabetes outcomes in a large academic health care system.

Keywords:  continuous glucose monitor; diabetes; health care system; pharmacist

DOI:  https://doi.org/10.1016/j.japh.2025.102397
Mayo Clin Proc Digit Health. 2023 Jun;1(2): 189-200

Screening for Impaired Glucose Homeostasis: A Novel Metric of Glycemic Control.

Jaycee M Kaufman, Lennaert van Veen, Yan Fossat.

Objective: To investigate the use of a mathematical model of glucose homeostasis, fit to continuous glucose monitor data, as a metric of dysfunctional glycemic control.
Patients and Methods: Three hundred eighty four participants recruited from 2 studies between October 2020 and June 2022 were equipped with a continuous glucose monitor, and interstitial glucose data were automatically collected for 2 weeks. The participants were assessed by a physician and diagnosed as being diabetic, prediabetic, or healthy according to the American Diabetes Association guidelines. A mathematical model of glucose homeostasis was fitted to the glucose data, and model parameter values were obtained. The participants were classified into the following 2 groups on the basis of their glucose homeostasis parameters: effective and impaired. Finally, glycemic variability metrics were compared with glucose homeostasis classification.
Results: The homeostasis classification resulted in a specificity, sensitivity of individuals with prediabetes, and sensitivity of individuals with type 2 diabetes (T2D) of 0.78, 0.86, and 1.00, respectively, for women and 0.71, 0.86, and 1.00, respectively, for men. This sensitivity was similar to that of glycated hemoglobin A1c measurement (a sensitivity of 0.89 for women and 0.90 for men for prediabetes and a sensitivity of 1.00 for T2D) and superior to that of the oral glucose tolerance test (a sensitivity of 0.18 for women and 0.24 for men for prediabetes and a sensitivity of 0.75 for women and 0.86 for men for T2D). Overall, the individuals classified as impaired had increased glucose variability metrics than the individuals classified as effective (P<.05).
Conclusion: The classification of glucose homeostasis on the basis of mathematical modeling of continuous measurements has promising applications as a new metric of dysfunctional glycemic control.
Trial Registration: clinicaltrials.gov Identifier: NCT04529239; clinical trial registry identifier: CTRI/2021/08/035957.

DOI: https://doi.org/10.1016/j.mcpdig.2023.02.008
Digit Health. 2025 Jan-Dec;11:11 20552076251328980

Exploring the potential of deep learning models integrating transformer and LSTM in predicting blood glucose levels for T1D patients.

Xin Xiong, XinLiang Yang, Yunying Cai, Yuxin Xue, JianFeng He, Heng Su.

   Objective: Diabetes mellitus is a chronic condition that requires constant blood glucose monitoring to prevent serious health risks. Accurate blood glucose prediction is essential for managing glucose fluctuations and reducing the risk of hypo- and hyperglycemic events. However, existing models often face limitations in prediction horizon and accuracy. This study aims to develop a hybrid deep learning model combining Transformer and Long Short-Term Memory (LSTM) networks to improve prediction accuracy and extend the prediction horizon, using personalized patient information and continuous glucose monitoring data to support better real-time diabetes management.
Methods: In this study, we propose a hybrid deep learning model combining Transformer and LSTM networks to predict blood glucose levels for up to 120 min. The Transformer Encoder captures long-range dependencies, while the LSTM models short-term patterns. To improve feature extraction, we integrate Bidirectional LSTM and Transformer Encoder layers at multiple stages. We also use positional encoding, dropout layers, and a sliding window technique to reduce noise and manage temporal dependencies. Richer features, including meal composition and insulin dosage, are incorporated to enhance prediction accuracy. The model's performance is validated using real-world clinical data and error grid analysis.
Results: On clinical data, the model achieved root mean square error/mean absolute error of 10.157/6.377 (30-min), 10.645/6.417 (60-min), 13.537/7.283 (90-min), and 13.986/6.986 (120-min). On simulated data, the results were 1.793/1.376 (15-min), 2.049/1.311 (30-min), and 3.477/1.668 (60-min). Clark Grid Analysis showed that over 96% of predictions fell within the clinical safety zone up to 120 min, confirming its clinical feasibility.
Conclusion: This study demonstrates that the combined Transformer and LSTM model can effectively predict blood glucose concentration in type 1 diabetes patients with high accuracy and clinical applicability. The model provides a promising solution for personalized blood glucose management, contributing to the advancement of artificial intelligence technology in diabetes care.

Keywords:  Clarke Error Grid Analysis; deep learning model; predictive analytics; short-term blood glucose prediction; type 1 diabetes

DOI:  https://doi.org/10.1177/20552076251328980