ATTD 2025 Short Oral Abstracts

ADVANCED AI MODELS FOR PREDICTING BLOOD GLUCOSE LEVELS IN PATIENTS WITH DIABETES: CAPABILITIES, LIMITATIONS, AND CHALLENGES OF CLINICAL INTEGRATION

Oualid Bada ^1,2,3

¹Former Research Associate, University of Bonn, Germany, Statistics, Bonn, Germany, ²Deutscht Telekom, Ai Expert Innovation & Development, Bonn, Germany, ³Doctor of Statistics, Königswinter, Germany

Background and Aims: Recent advancements in Artificial Intelligence (AI) have revolutionized fields such as natural language processing, computer vision, and, notably, medical diagnostics. Among these innovations, AI-techniques for time series forecasting hold great potential for predicting blood glucose levels in individuals with diabetes. Accurate glucose prediction is crucial for optimizing automated insulin delivery, reducing the risk of hypo- and hyperglycemia, and improving glucose stability within time-in-range. Despite the rapid progress in AI, the selection of an optimal model for accurately predicting glucose dynamics in individuals with diabetes has not been sufficiently explored, particularly in terms of integrating these models into routine clinical practice.

Methods: This study evaluates the strengths and limitations of advanced AI-models for glucose forecasting using real-world data from a Type1 Diabetes patient. The models tested include Temporal-Fusion-Transformers (TFT), Neural-Basis-Expansion-Analysis-for-Time-Series (N-BEATS), and Deep-Auto-Regressive-Models (DeepAR). Data from CGM devices, along with insulin, nutrition, and physical activity records, were analyzed to assess each model’s predictive performance.

Results: The study provides insights into the strengths and limitations of three families of advanced AI-based prediction models, emphasizing their potential to enhance diabetes management while addressing clinical integration challenges. By comparing different AI-models, the results show a clear correlation between model capabilities and prediction accuracy (TFT, with over 200k model coefficients, outperforms Nbeats, which has around 8k coefficients, in MAE by 25%).

Conclusions: This comprehensive evaluation shows that AI-models can enhance glucose forecasting accuracy. However, challenges remain in incorporating additional data on physical activity, fiber, and fat intake. Further research is needed to optimize models for clinical integration and real-world performance.

EXERCISE-RELATED GLYCEMIC EVENTS IN TYPE 1 DIABETES (T1DM): TOWARDS A PRACTICAL DECISION SUPPORT TOOL

Ryan Coopergard ¹, Lisa Chow², Sisi Ma¹

¹University of Minnesota - Twin Cities, Institute For Health Informatics, Minneapolis, United States of America, ²University of Minnesota - Twin Cities, Department Of Medicine, Minneapolis, United States of America

Background and Aims: Patients with type 1 diabetes (T1DM) frequently struggle with exercise-induced hyperglycemia and hypoglycemia, decreasing their willingness to exercise. We aim to build accurate and easy-to-deploy models to forecast exercise-induced glycemic events in real-world settings.

Methods: We analyzed free-living data from the Type 1 Diabetes Exercise Initiative (T1DEXI) study, where adults with T1DM wore a continuous glucose monitor (CGM) while performing video-guided exercises (30-minute exercises for at least 6 times over 4 weeks), along with concurrent detailed phenotyping of their insulin program and diet. We built models to predict glycemic events (blood glucose ≤ 54 mg/dL, ≤ 70 mg/dL, ≥ 200 mg/dL, and ≥ 250 mg/dL) during- and one-hour post-exercise with variables from four data modalities: demographic and clinical, CGM, carbohydrate intake and insulin administration, and exercise type, duration and intensity.

Results: Models incorporating information from all four data modalities showed excellent predictive performance with AUC > 0.89 for all glycemic events. Models built with CGM data alone have statistically indistinguishable performance compared to models using all data modalities. These models also showed outstanding calibration (Brier score ≤ 0.08) and resilience to noisy input.

Conclusions: We successfully constructed models to forecast exercise-induced glycemic events using only automatically captured CGM data as input, incurring minimal user burden. These models showed excellent predictive performance, calibration, and robustness, enabling model translation into a decision support tool that is easy to deploy and maintain.

ACCURACY AND VALIDITY OF MADHU AI (METABOLIC AND DIABETES HEALTH ASSISTANT)—INDIA'S FIRST WHATSAPP-BASED AI CHATBOT FOR DIABETES: RESULTS FROM PHASE I RANDOM VALIDATION

Amit Dey ¹, Rajat Anand², Manisha Gupta³, Hitesh Saraogi⁴, Mahira Sayed⁵, Rutul Gokalani⁶, Alok Modi⁷, Vinayak Hiremath⁸, Avneesh Khare⁹, Shambo Samajdar¹⁰, Anuj Maheshwari¹¹, Rakesh Parikh¹², Bharat Saboo¹³, Banshi Saboo⁵

¹Apollo Sugar Clinics, Diabetology, Kolkata, India, ²Fimblefits, R&d, Kolkata, India, ³Regency hospital, Kanpur, India, ⁴Dhanvantari Hospital, Diabetology, Ghaziabad, India, ⁵Diacare, Diabetes, Ahmadabad, India, ⁶AHC Diabetes Clinic, Ahmedabad, India, ⁷Kevalya Hospital, Thane, India, ⁸KMK Hospital, Internal Medicine, Kochi, India, ⁹BrainXAI, R&d, Kolkata, India, ¹⁰JNM Medical College and Hospital, Pharmacology, Kalyani, India, ¹¹Hind Institute of Medical Sciences, Internal Medicine, Lucknow, India, ¹²Diabecity, Diabetology, Jaipur, India, ¹³Prayas Diabetes Center, Diabetology, Indore, India

Background and Aims: Artificial intelligence (AI) is revolutionizing healthcare through tools like chatbots. MadhuAI, the first WhatsApp-based AI chatbot in India, was developed to support healthcare professionals (HCPs) and patients in managing diabetes and metabolic health. This study aims to evaluate the chatbot’s effectiveness in delivering accurate, comprehensive, and context-sensitive information through Phase I of a structured validation process.

Methods: In Phase I, a random validation approach was conducted over a 2-week period. 200 doctors from all over India, including diabetologists, endocrinologists, physicians, diabetes educators, and dietitians, interacted with MadhuAI daily, asking diabetes-related questions based on their expertise or as simulated patients. Responses were evaluated using a 5-point Likert scale across eight parameters: accuracy, relevance, clarity, empathy, organization, neutrality, respectfulness, and innovation. Statistical analysis included the calculation of mean scores and standard deviations, followed by chi-square and t-tests to identify significant variations in ratings.

Results: The analysis revealed an overall high level of satisfaction with MadhuAI’s performance:

Mean Scores: Scores ranged from 3.8 to 4.8 across various criteria, with the highest scores for “respectfulness and privacy” (mean: 4.8) and “clarity and grammar” (mean: 4.7). The parameter “Avoidance of Harmful Advice and Neutrality” received the lowest score (mean: 3.8) Significant differences were noted in ratings for perceived empathy and neutrality (p < 0.05), underscoring areas where the chatbot’s responses could be further optimized to better emulate human-like understanding and unbiased perspectives.

Conclusions: Random validation demonstrates MadhuAI provides accurate and comprehensive responses, respects privacy, and maintains clarity in its communication, making it an effective patient education tool in diabetes care.

DIABETES SELF-MANAGEMENT SMARTPHONE APPLICATION: COULD IT BE AN ALTERNATIVE FOR CONTINUOUS GLUCOSE MONITORING IN LOW RESOURCE SETTINGS?

Nancy Elbarbary, Mohamed Osama, Yasmeen Fereig

Faculty of Medicine, Ain Shams University, Department Of Pediatrics, Cairo, Egypt

Background and Aims: Objectives: To assess the role of the rightest app in improving glycemic control as well as quality of life in a 6- month interval of its usage.

Methods: A case- control study included 40 participants with type 1 diabetes, mean age 14.3±1.42 years and diabetes duration of 4.5±3.6 years regularly following up at Diabetes Unit. Rightest app was installed on the participants’ smart phone on enrollment and 6 months later. Assessment of the glycemic control was done by HbA1c and time in range results collected from app. Quality of life (PedsQL) and user experience (UEQ) questionnaires were applied at the end of the study.

Results: Using smartphone app yielded a significant reduction in mean BG level (-17.64%,p=0.012)that decreased HbA1c (−10.63%, p = 0.000). Increase of SMBG frequency was observed (p=0.04). This is reflected on a 20% increment time in range generated by app (p-value=0.002) and lower time above range (>180mg/dl, -18.75%, P=0.001) in intervention group compared to control. However, the number of hypoglycemic events (p=0.71) or DKA (p=0.59) did not differ between groups. PedsQL questionnaire total score has improved (p-value=0.010) in favor of intervention group with good experience with app indicated by UEQ. UEQ revealed pragmatic total score (1.425) was above average while the hedonic total score was good (1.250) and overall score was good(1.338).

Conclusions: Using mobile apps as an alternative for CGM in low resource settings can help improve glycemic control and quality of life for those who don’t have access to diabetes technology services because of unavailability or unaffordability.

A DEEP LEARNING SYSTEM FOR POPULATION-LEVEL SILENT BRAIN INFARCTION DETECTION AND STROKE RISK PREDICTION FROM RETINAL PHOTOGRAPHS

Zhouyu Guan ¹, Yongjun Wang², Qionghai Dai³, Weiping Jia¹, Huating Li¹, Bin Sheng⁴, Tien Yin Wong⁵

¹Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Department Of Endocrinology And Metabolism, Shanghai, China, ²Beijing Tiantan Hospital, Capital Medical University, Department Of Neurology, Beijing, China, ³Tsinghua University, Institute For Brain And Cognitive Sciences, Beijing, China, ⁴Shanghai Jiao Tong University, Department Of Computer Science And Engineering, Shanghai, China, ⁵Tsinghua University, Tsinghua Medicine, Beijing, China

Background and Aims: Conventional stroke risk assessments for people with diabetes, relying on clinical traits and self-reported data, may not accurately identify individuals at risk of stroke. Brain imaging could detect subclinical cerebrovascular diseases such as silent brain infarctions (SBIs), which may more precisely identify those at risk of stroke. However, using brain imaging to detect SBIs is not feasible or cost-effective for people with diabetes. To overcome these challenges, we developed and validated a retinal image-based deep learning system (DeepRETStroke), to predict the presence of SBI and refine future stroke risk prediction.

Methods: We first utilized 895,640 retinal photographs to pre-train the DeepRETStroke system. We then used 212,690 retinal photographs from 13 datasets from China, Singapore, Malaysia, USA, UK and Denmark to train and validate the DeepRETStroke system to detect SBI and predict future stroke events.

Results: The DeepRETStroke system performed well in the internal validation dataset, leading to an AUC of 0.801 (95% confidence interval [CI], 0.792-0.810) in detecting SBI, 0.902 (0.900-0.903) for predicting incident stroke, and 0.773 (0.729-0.826) for predicting recurrent stroke. External validations demonstrated consistent performances among diverse populations with diabetes from China, Singapore, Malaysia, USA, UK and Denmark.

Conclusions: In summary, a retinal image-based deep learning system, DeepRETStroke is superior to clinical traits in predicting future stroke events among diverse populations with diabetes, especially by incorporating the detection of SBI without the need for brain imaging. The DeepRETStroke system could be potentially utilized for diabetes care at the population level, since regular retinal examinations are recommended for people with diabetes.

A DEEP LEARNING SYSTEM FOR PREDICTING TIME TO PROGRESSION OF DIABETIC RETINOPATHY

Huating Li ¹, Ling Dai², Bin Sheng², Tingli Chen³, Qiang Wu⁴, Zhouyu Guan¹, Shujie Yu¹, Carol Y. Cheung⁵, Gavin Siew Wei Tan⁶, Yih-Chung Tham⁶, Ching-Yu Cheng⁶, Tien Yin Wong⁷, Weiping Jia¹

¹Department of Endocrinology and Metabolism, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine., Shanghai, China, ²Shanghai Jiao Tong University, Shanghai, China, ³Department of Ophthalmology, Huadong Sanatorium., Wuxi, China, ⁴Department of Ophthalmology, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai., Shanghai, China, ⁵Department of Ophthalmology and Visual Sciences, The Chinese University of Hong Kong., Hong Kong, China, ⁶Singapore Eye Research Institute, Singapore National Eye Centre., Singapore, Singapore, ⁷Tsinghua Medicine, Beijing Tsinghua Changgung Hospital, Tsinghua University., Beijing, China

Background and Aims: Diabetic retinopathy (DR) is the leading cause of preventable blindness worldwide. The risk of DR progression is highly variable among different individuals, making it difficult to predict risk and personalize screening intervals. We aimed to develop and validate a deep learning system (DeepDR Plus) to predict time to DR progression within 5 years solely from fundus images.

Methods: First, we used 717,308 fundus images from 179,327 participants with diabetes to pre-train the system. Subsequently, we trained and validated the system with a multiethnic dataset comprising 118,868 images from 29,868 participants with diabetes. Furthermore, we validated the system in real-world cohorts of participants with diabetes.

Results: For predicting time to DR progression, the DeepDR Plus system achieved concordance indexes of 0.754–0.846 and integrated Brier scores of 0.153–0.241 for all times up to 5 years. Furthermore, in real-world cohorts of participants with diabetes, the integration with clinical workflow could potentially extend the mean screening interval from 12 months to 31.97 months, and the percentage of participants recommended to be screened at 1–5 years was 30.62%, 20.00%, 19.63%, 11.85% and 17.89%, respectively, while delayed detection of progression to vision-threatening DR was 0.18%. Retinal vascular geometry and the fovea were important image patterns for future occurrence of DR based on the DeepDR Plus system.

Conclusions: Altogether, the DeepDR Plus system could predict individualized risk and time to DR progression over 5 years, potentially allowing personalized screening intervals.

“SOCIO-ECONOMIC, BEHAVIOURAL AND MENTAL HEALTH RISK FACTORS FOR DEVELOPING ACUTE MYOCARDIAL INFARCTION OR MORTALITY IN A LARGE MULTI-ETHNIC ASIAN COHORT”

Olivia Jiawen Xia¹, Derek Hausenloy², Zheng Jye Ling³, Jun Hui Tan⁴, Raphael Thaddaeus Lee¹, Wendelynn Chua¹, Clara Tan⁴, Zhen En Ang⁴, Ann Aerts⁵, Helge Buess⁶, Peter Speyer⁶, Lian Leng Low ⁷

¹Singhealth, Singhealth Community Hospital, Singapore, Singapore, ²Duke NUS, Duke Nus Gms, Singapore, Singapore, ³NUS, Biomedical Informatics Yong Loo Lin School Of Medicine, Singapore, Singapore, ⁴National University Health System, Rhso, Singapore, Singapore, ⁵Novartis Foundation, Rhso, Singapore, Singapore, ⁶Novartis Foundation, Ai4healthcities, Singapore, Singapore, ⁷SingHealth, Singapore, Singapore

Background and Aims: Early assessment of cardiovascular (CV) risk in diabetic patients may improve the prevention of severe CV events such as acute myocardial infarction and mortality. This study aims to evaluate the impact of socio-economic and behavioral determinants in addition to known risk factors on predicting CV risk among diabetic patients in multi-ethnic Singapore.

Methods: This was a retrospective cohort study of diabetic patients as of 31^st December 2015 from the Singapore National Diabetes Database (NDD). The primary outcome of interest was the first occurrence of acute myocardial infarction or mortality between 2016 and 2020. Socio-economic determinants included ethnicity, income, housing type and marital status, behavioral determinants included smoking status, participation in national physical activity programs. We analyzed the time to event using multi-variate Cox proportional hazards regression.

Results: A total of 247,491 patients were included in the analysis and 42,859 (17.3%) developed the primary event. Significant risk factors were Non-Chinese (HR 1.08), widowed (HR 1.39), income <$930 (HR 1.43), residence in public housing (HR 1.25 – 1.32), past hospitalization for depression or anxiety (HR 1.52), and ex or current smokers (HR 1.5) Participants in all five editions of the national exercise program had reduced risk (HR 0.65).

Conclusions: Our study is the first to analyze a large cohort of multi-ethnic Asians with diabetes and quantified socio-economic, behavioral determinants and mental health as key risk factors for CV events. Our findings have clinical utility for prediction, risk assessment and shared decision making for empowering behavioral change in diabetic patients.

PRECLINICAL EVALUATION OF DIGITAL TWINNING TECHNOLOGY TO ADDRESS DYSGLYCEMIA ACROSS THE MENSTRUAL CYCLE IN WOMEN WITH TYPE 1 DIABETES

Maria F. Villa-Tamayo ¹, Jacopo Pavan¹, Eda Cengiz², Chiara Fabris¹

¹University of Virginia, Center For Diabetes Technology, Charlottesville, United States of America, ²University of California, San Franciso, San Francisco, United States of America

Background and Aims: Insulin need fluctuations across the menstrual cycle (MC) challenge glycemic control in women with type 1 diabetes (T1D). This study presents a digital twinning (DT) technology enabling insulin dosing adaptation across the MC.

Methods: A DT-based therapy advisor was designed to adjust insulin therapy profiles (basal rate, insulin-to-carbohydrate ratio, correction factor) within an automated insulin delivery (AID) system across the MC. The advisor uses continuous glucose monitoring (CGM), insulin/meal, and demographic data to individualize a model of glucose-insulin dynamics and determine personalized optimal therapy profiles for follicular phase (FP) and luteal phase (LP) of the MC. The advisor was tested within the UVA/Padova T1D Simulator augmented with MC-related metabolic variability. Paired t-tests were used to compare CGM-based glycemic outcomes between a 2-month baseline period and a 2-month intervention with deployment of phase-specific optimal therapy profiles.

Results: Compared to baseline, deployment of optimal therapy profiles resulted in decreased average glucose (151.40±9.29 vs. 170.59±10.45 mg/dL; p<0.001), increased time in range (71.97±8.37% vs. 60.67±8.81%; p<0.001), and decreased time above range (25.33±7.21% vs. 37.46±7.99%; p<0.001), with most noticeable changes during LP due to the increased insulin resistance typical of this cycle phase (Table 1). Time below range and hypoglycemia treatments remained stable in FP and increased in LP to FP levels within clinically acceptable values.

Conclusions: The DT-based advisor demonstrated potential to enhance AID performance by accounting for MC-related metabolic variations. This in-silico study provides preclinical evidence in support of an upcoming clinical trial testing this technology in women with T1D experiencing MC-related dysglycemia.

USING AN AI-BASED ADVISOR AND DIGITAL TWINS TO SAFELY INITIATE AND ADJUST AUTOMATED INSULIN DELIVERY SETTINGS IN YOUNG CHILDREN WITH TYPE 1 DIABETES

Jacopo Pavan ¹, Erin Cobry², Zachariah Reed³, Maria F. Villa-Tamayo¹, Jenny Diaz Castaneda¹, Emily Jost², Ryan Kingman⁴, John Lum³, Roy Beck³, Mark Deboer¹, Melissa Schoelwer¹, Bruce Buckingham⁴, Paul Wadwa², Marc Breton¹

¹University of Virginia, Center For Diabetes Technology, Charlottesville, United States of America, ²Barbara Davis Center for Diabetes, University of Colorado, Aurora, United States of America, ³JAEB Center for Health Research, Research, Tampa, United States of America, ⁴Stanford University, Stanford, United States of America

Background and Aims: Randomized controlled studies have demonstrated the efficacy of automated insulin delivery (AID) in young children with Type 1 Diabetes (T1D); yet, initiating and adjusting AID settings can be challenging due to high glycemic variability and unpredictable behaviors. This multi-center study investigated the use of an advisory system based on artificial intelligence (AI) and digital twins (DT) to guide initiation and adjustment of AID settings in children 2 to <6 years old.

Methods: Participants used Tandem t:slim X2 pump with Control-IQ and Dexcom G6/G7 continuous glucose monitoring sensors for 8 weeks at home. Initial profiles and following adjustments were provided through an AI-based advisor and reviewed by investigators before implementation. DTs (i.e., virtual replicas of participants’ glucose-insulin dynamics) were trained daily to reproduce participant’s data and used to optimize therapy profiles every two weeks or on-demand. Safety endpoints included the percentage of time <54mg/dl and >250mg/dl. Efficacy endpoints (tested in hierarchical order) were the percentage of time in 70-180mg/dl, mean glucose level, percentage of time >250mg/dl, <70mg/dl and <54mg/dl.

Results: Thirty-two participants were recruited (age 2 to 5.9 years), with three excluded from analysis due to missing data. Percentage of time <54mg/dl and >250mg/dl were non-inferior during the 8 weeks follow-up compared to baseline (p<0.001, Table 1). Superiority tests showed significantly improved percent time 70-180mg/dl (p=0.005), less time >250mg/dl (p=0.003), and lower mean glucose (p=0.02).

Conclusions: The combined use of AID with AI-driven initiation and adjustment of parameters was safe and beneficial for young children with T1D.

LEVERAGING DIGITAL TWINS TO ENHANCE HUMAN-MACHINE CO-ADAPTATION IN ADULTS WITH TYPE 1 DIABETES USING AUTOMATIC INSULIN DELIVERY

Jacopo Pavan ¹, Patricio Colmegna^1,2, Maria F. Villa-Tamayo¹, Jenny Diaz Castaneda¹, Boris Kovatchev¹, Sue Brown¹

¹University of Virginia, Center For Diabetes Technology, Charlottesville, United States of America, ²Dexcom, Inc., San Diego, United States of America

Background and Aims: Most automatic insulin delivery (AID) systems for type 1 diabetes (T1D) management do not adapt to user physiology, nor do they provide the tools for users to adapt to AID actions. In this randomized cross-over trial, we evaluate the use of digital twins (DTs) to enhance human-machine co-adaptation in AID users by means of personalized replay simulation.

Methods: Participants’ glucose-insulin data were periodically transferred into a cloud-based server and mapped into a DT – i.e., a simulator model of their metabolism. Participants used an investigational smartphone app and underwent three interventions: 1) use of AID alone; 2) addition of information feedback (AID+IF), consisting of real-time glucose indicators; 3) addition of an adaptive bio-behavioral control module (AID+ABC), which leveraged DTs to recommend optimized AID settings and to enable the simulation of what-if scenarios on participant’s own data. Participants were randomized in two groups, switching from AID alone to AID+ABC (escalation) or vice versa (de-escalation). The primary outcome is the percentage of time spent in range 70-180 mg/dl (TIR).

Results: A total of 72 participants completed the study (age: 42.8±15.1 years; glycated hemoglobin: 6.8%±0.8%; female: 68%). Escalation group (baseline TIR 73.8%±10.5%) improved from 73.6%±12.3% TIR in AID alone to 77.1%±9.2% in AID+ABC (Figure 1). TIR in de-escalation group (baseline: 74.1%±11.9%) improved to 77.4%±10.3% during AID+ABC and remained unchanged, 77.9%±10.7% during AID alone.

Conclusions: The use of AID+ABC improved time in range vs AID alone in a two-arm cross-over trial, showing that DTs can be effectively leveraged to enhance human-machine co-adaptation in for AID users.

IMPACT OF AN INSULIN BOLUS CALCULATOR AND PHYSICAL ACTIVITY ON GLYCEMIC CONTROL IN PEOPLE WITH TYPE 1 DIABETES - A RANDOMISED CROSS-OVER PILOT TRIAL

Kenney Pedersen ¹, Mikkel Olsen², Pernille Mogensen¹, Gasper Letnar³, Peter Kristensen^2,4

¹Hedia, Department Of Clinical And Medical Affairs, Copenhagen, Denmark, ²North Zealand Hospital, Endocrinology And Nephrology, Hilleroed, Denmark, ³Omicron ApS, Søborg, Denmark, ⁴Department of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark

Background and Aims: Physical activity (PA) poses challenges to people with type 1 diabetes (PWT1D) due to increased risk of hypoglycemia. Hedia Diabetes Assistant (HDA), an insulin bolus calculator app with a physical activity module based on international guidelines, may improve post-exercise glycaemia in PWT1D.

Methods: This randomised, open-label, cross-over pilot trial included Danish PWT1D on basal-bolus therapy. Subjects were randomised to use HDA or habitual treatment (HT) before a 45-minute moderate-intensity PA biking-bout and then crossed over. The primary endpoint was the difference in time in range (3.9-10.0 mmol/L, TIR) during a 24-hour period starting one hour before PA.

Results: Twenty subjects (mean age: 57.2 years, 15% female, mean diabetes duration: 16.5 years) were included. Mean TIR was 57.6% (SD 21.6%) with HDA vs. 60.1% (17.7%) with HT (mean difference -2.3%, 95% CI [-13.1%, 8.5%]). Subjects experienced 10 hypoglycemic episodes (< 3.9 mmol/L) with HDA vs. 19 with HT (rate ratio: -0.6, 95% CI [-1.4, 0.3]). Mean time above range was 41.1% (SD 22.0%) with HDA vs. 37.1% (SD 18.0%) with HT (mean difference 3.9%, 95% CI [5.7%, 13.5%]). Median time below range was 0% (Q1, Q3 0%, 2.3%) with HDA vs. 1.2% (0%, 3.2%) with HT (median difference -0.9%, 95% CI [-2.8%, 0.5%]).

Conclusions: There was a trend toward fewer hypoglycemic episodes when using HDA, suggesting it may decrease the risk of exercise-induced hypoglycemia. This could improve safety and support a healthier, more active lifestyle for individuals with a higher risk of hypoglycemia during exercise.

EVALUATING THE IMPACT ON PATIENT OUTCOMES OF CHANGES TO ALGORITHM-DIRECTED CARE MODELS

Isha Thapa ¹, Franziska Bishop², Johannes Ferstad¹, David Maahs², David Scheinker², Ramesh Johari¹

¹Stanford University, Management Science And Engineering, Stanford, United States of America, ²Stanford University, Pediatric Endocrinology, Stanford, United States of America

Background and Aims: Clinics continue to adopt RPM in which algorithms analyze CGM data to help direct patient care. Changes to such algorithms are rarely evaluated rigorously, despite their potential to affect patient outcomes. The regression discontinuity in time (RDT) design can assess the impact of changes to algorithm-directed care, but has received little attention in medical literature.

Methods: We performed a retrospective analysis of CGM data from youth with type 1 diabetes enrolled in the 4T Study. The study initially used an algorithm that directed patients for provider review based on the number of care targets not being met, such as time in range (TIR) < 65% or percent time with clinically significant hypoglycemia > 1%. In September 2021, the clinic adopted a new algorithm that prioritized patients by the type of unmet target. We used the RDT framework to estimate the impact on weekly TIR of the change of algorithms.

Results: There were 212 patients with 12,371 weekly TIR observations. The estimated effect on TIR was -.009 percentage points (95% CI: [-.060, .027]) in the unadjusted model and -.006 (95% CI: [-.024, .010]) in the model controlling for patient-level fixed effects, prior contact, and prior TIR (Figure).

Conclusions: We demonstrate the feasibility of using the RDT framework to estimate the impact on patient care of changes to an algorithm that directs care in a digital health intervention. As CGM technology and algorithm-directed care increase in clinical practice, this approach can serve as a diagnostic tool to measure how operational changes impact outcomes.

PERFORMANCE OF THE MINIMED 780G SYSTEM IN THE MIDDLE EAST, NORTH AFRICA, AND TURKEY (MENAT) REGION: A REAL-WORLD ANALYSIS OF 7965 USERS

Mohammed Al-Sofiani ¹, Tim Van Den Heuvel², Wael Chaar³, Ohad Cohen², Goran Petrovski²

¹King Saud University, Department Of Internal Medicine, Riyadh, Saudi Arabia, ²Medtronic International Trading Sarl, Diabetes Operating Unit, Tolochenaz, Switzerland, ³Medtronic Saudi Arabia, Riyadh, Saudi Arabia

Background and Aims: Real-world evidence (RWE) on the use of advanced diabetes technology in the MENAT region (Middle East, North Africa, Turkey) is limited. Here, we report RWE of the MiniMed 780G system (MM780G) users in MENAT.

Methods: ‘CareLink Personal’ data from MM780G users were included (until 01AUG2024). CGM-based endpoints were aggregated for (1) the overall/full cohort, and for sub-cohorts based on (2) optimal setting users (GT=100mg/dL, AIT=2hrs), (3) age-groups (≤15, 16-55, >56yrs), (4) day/night, (5) seasons and (6) countries with >50 users. Additionally, 12-month longitudinal analyses were performed.

Results: The 7965 users showed a mean GMI of 6.9%±0.7%, TIR of 71.7±18.2%, TB70 of 2.4±3.9%, and TB54 of 0.5±1.7% (Figure). These results align with international targets and with RWE from other regions. Glycemic improvements began in the first month and sustained for 12 months. Optimal setting users showed the best glycemic control. Other sub-cohorts, including those ≤15 years old, those in countries with high diabetes burden and limited specialists, and those in countries where summer season is extremely hot (>40-50°C), showed results consistent with the full cohort. Compared to global RWE, MM780G users in MENAT had no substantial day/night TIR difference (71.1% vs 73%), likely due to cultural norms such as late fat-rich dinners, post-midnight bedtimes, and short sleep.

Conclusions: The sustained excellent glycemic control among MM780G users in MENAT highlights the need to improve access to advanced diabetes technologies. Improved access could address the high burden of poorly controlled diabetes, the shortage of specialists, and the low utilization of advanced technology.

A UK MULTI-CENTRE REAL-WORLD EVALUATION OF A TUBELESS HCL SYSTEM – DEPRIVATION DOSE NOT INFLUENCE IMPROVEMENTS BUT MAY IMPACT ATTAINMENT OF CONSENSUS TIR TARGETS

Ananthi Anandhakrishnan ^1,2, Stephen Greene³, Carol Willis³, Sufyan Hussain^1,2,3

¹King's College London, Department Of Diabetes, School Of Cardiovascular, Metabolic Medicine And Sciences, London, United Kingdom, ²GUY'S AND ST THOMAS' NHS FOUNDATION TRUST, Great Maze Pond, United Kingdom, ³London Medical, London, United Kingdom

Background and Aims: Omnipod 5® (OP5), the first tubeless commercial hybrid-closed loop (HCL) system, launched in the UK in June 2023. We report real-world outcomes of OP5 in public and private sector diabetes services in the UK.

Methods: We performed a retrospective, three-centre observational study including adults aged ≥ 18 years with type 1 diabetes (T1D) commencing OP5 between 24^th July 2023 and April 2^nd 2024. Baseline and post-OP5 glycaemic outcomes and anthropometrics are reported. Validated index of multiple deprivation (IMD) decile is used as a relative measure of deprivation.

Results: 178 adults with T1D (Female 57.30%, n=102; White British 41.57%, n=74; median IMD decile, 5) were included. 43.26% (n=77) were new to pump-therapy and 2.81% (n=5) were on alternate HCL systems. Significant improvements in sensor glucometrics and laboratory Hba1c were seen post-OP5 (Table 1). The proportion of individuals achieving international consensus time-in-range (TIR) targets improved from 8.99% (n=16) to 42.13% (n=75). The degree of TIR improvement (ΔTIR) was related to baseline TIR with larger improvements seen in those with lower baseline TIR (r =-6.67, p<0.001). No relationship between deprivation and attainment of international consensus TIR targets or ΔTIR was seen (Table 2). Healthcare funding was not found to influence ΔTIR, but was associated with the attainment of consensus TIR targets at baseline and post-OP5 (Table 3).

Conclusions: Our data corroborates the real-world efficacy of the OP5 system. Whilst deprivation does not influence improvements, differences in baseline attainment of consensus TIR targets suggests a selection bias in those self-funding diabetes management.

A REAL WORLD COMPARISON BETWEEN THE AID (AUTOMATED INSULIN DELIVERY) HYBRID CLOSED LOOP SYSTEMS IN GERMANY

Christine Berndt-Zipfel, Stephan Maxeiner

DHG Bosenheim, Bad Kreuznach, Germany

Background and Aims: Since 2021,commercial AID-hybrid-closed-loop-systems are available in Germany. We wanted to find out if the patients in our diabetology-practice benefit.

Methods: We analyzed retrospectively our first 100 AID-hybrid-closed-loop-system-patients. Diabetes-management-data (disease-management-program collected every 3months) (HbA1c, weight),(time-in-range, time-below-range) from DexcomClarity, Abbott-libre-view, glooko,diasend,Medtronic carelink, X-drip,yourloops,statistics/Microsoft-Excel.

Results: Because of pregnancy,doctorchange,transition only data of 62patients (26women/36men)∼40y, diabetesduration 27y. HbA1c before AID-systemstart 7.63%. 3month after start HbA1c dropped to 7.02%(p < 0.0000001), 6month was 7.12%(p=0.0000007), 9m 6.97%(p=0.000003), 12month 7.02%(p = 0.00007). Time-in-range (TIR) before AID.start was 54.21%, 3months later 69.26%(p=0.000017), 6months after AID-start TIR was 70.8%(p< 0.000001). Before AID the time-below-range(TBR) was 3.78%, 3months after start 1.95%(p=0.0023); 6months 1.73%(p=0.0014). Medtronic-patients HbA1c 7.6% before AID-system.start. 3months after start 7.23%, 6months 7.26%, 9 months6.76%,12months 6.91%. TIR 54% before AID-system, with AID 70%. TBR was 8.14% before AID-system-start, with AIS-system 2.73%. Tandem-T.slim-patients HbA1c 7.64% before AID-system-start. 3 months after AID-system-start 6.98%, 6months 7.14%, 9months 7.04%, 12 months 7.04%. TIR was 50% before AID-system-start, with AID-system TIR65%. TBR was 2.7% before AID-system-start, with the AID-system 1.5%. Ypsopump/CamAPS-FX-patients HbA1c 7.46% before AID-system-start. 3months after AID-system-start 6.63%, 6 months 6.88%, 9 months 6.5%. TIR 56.4% before AID-system-start, with AID-system 81.6%. TBR was 3% before AID-system-start, with system 2.6%. Roche insight/DBLG1 patients had HbA1c 7.72% before AID-system-start. 3months after AID-system-start 7.04%, 6 months 7%, 9 months 7.11%,12 months 7.32%. TIR was 56% before AID-system-start, with AID-system 69%. TBR was 2.9%, with AID-system 1.9%. Do-it-yourself-AID-system-patients had HbA1c 6.53% (systemstart unkown). BMI raised from 27.79 to 29.18kg/m².

Conclusions: All commercial-AID-hybrid-closed-loop-systems lead to significant improvement HbA1c,TIR,TBR in real-world. Weight gain has to be observed.

THE IMPACT OF ULTRASOUND GUIDED ROTATION OF INJECTION SITES ON GLYCEMIC OUTCOME AND INSULIN REQUIREMENTS

Julian Bjerrekær ¹, Anna Korsgaard Berg¹, Merete Christensen¹, Kirsten Nørgaard², Jannet Svensson²

¹Steno Diabetes Center Copenhagen, Herlev, Denmark, ²Steno Diabetes Center Copenhagen, Diabetes Technology Research, Herlev, Denmark

Background and Aims: The benefits of automated insulin delivery (AID) are evident, but insulin infusion remains a weak point. The primary aim was to study if ultrasound-guided identification of optimal sites for placing infusion sets could support better rotation habits and reduce subcutaneous changes. Secondly, to investigate if using non-affected sites impacted glycemic outcomes and insulin requirements.

Methods: During three visits with two-week intervals, participants aged 6-18 years had more than 80% continuous glucose monitoring (CGM) usage integrated with their AID pump – Medtronic MiniMed 780G or Tandem T:slim X2. Participants underwent ultrasound examination of the two latest infusion sites at each visit. After two weeks, participants were informed of any hyperechogenicity in the subcutis and instructed to avoid affected areas. Glycemic parameters and insulin requirements for the previous two weeks were compared using a t-test for intention-to-treat analysis.

Results: From Day 1 to Day 28, we observed a 3.3%-point increase in TITR (p=0.0005), a 2.7%-point decrease in TAR (p=0.003), and a 0.3 mmol/l decrease in mean sensor glucose (p=0.0002), From Day 14 to Day 28, TITR increased by 1.4% points (p=0.047), and ins/kg/day decreased by 6.9% (p=0.0031). Overall, ins/kg/day was higher when hyperechogenicity was detected for everybody, and ins/kg/day increased every subsequent day after change of infusion set. Further, there was a significant reduction in the occurrence of wounds at the latest infusion site, dropping from 43.3% to 25.8%.

Conclusions: Using ultrasound pedagogically to select unaffected sites for insulin infusion improved glycemic outcomes and decreased insulin requirements in AID users.

SIMPLIFIED MEAL BOLUS STRATEGIES FOR THE OMNIPOD® 5 AUTOMATED INSULIN DELIVERY (AID) SYSTEM IN PEOPLE WITH TYPE 2 DIABETES (T2D): SUB-ANALYSIS OF THE SECURE-T2D STUDY

Georgia Davis ¹, Francisco Pasquel¹, David Huffman², Anne Peters³, John Parker⁴, Lori Laffel⁵, Giulio Romeo⁵, Justin Mathew⁶, Kristin Castorino⁷, Davida Kruger⁸, Kathleen Dungan⁹, Mark Kipnes¹⁰, Edward Jauch¹¹, Tamara Oser¹², Viral Shah¹³, Barry Horowitz¹⁴, Anders Carlson¹⁵, Mark Warren¹⁶, Wasim Deeb¹⁷, John Buse¹⁸, John Reed¹⁹, Jason Berner²⁰, Thomas Blevins²¹, Chris Bajaj²², Lauren Kanapka²³, Dan Raghinaru²³, Trang Ly²⁴, Roy Beck²³

¹Emory University, Atlanta, United States of America, ²University Diabetes and Endocrine Consultants, Chattanooga, United States of America, ³Keck School of Medicine of the University of Southern California, Los Angeles, United States of America, ⁴Accellacare, Wilmington Health, Wilmington, United States of America, ⁵Joslin Diabetes Center, Harvard Medical School, Boston, United States of America, ⁶Albert Einstein College of Medicine, Bronx, United States of America, ⁷Sansum Diabetes Research Institute, Santa Barbara, United States of America, ⁸Henry Ford Health, Detroit, United States of America, ⁹Ohio State University, Columbus, United States of America, ¹⁰Diabetes and Glandular Disease Clinic, San Antonio, United States of America, ¹¹Mountain Area Health Education Center, Asheville, United States of America, ¹²University of Colorado, Aurora, United States of America, ¹³Indiana University School of Medicine, Indianapolis, United States of America, ¹⁴Metabolic Research Institute, West Palm Beach, United States of America, ¹⁵HealthPartners, International Diabetes Center, Minneapolis, United States of America, ¹⁶Physicians East, Greenville, United States of America, ¹⁷East Coast Institute for Research at First Choice Endocrinology, Jacksonville, United States of America, ¹⁸University of North Carolina, Chapel Hill, United States of America, ¹⁹Endocrine Research Solutions, Roswell, United States of America, ²⁰East Coast Institute for Research at Georgia Mountain Endocrinology, Canton, United States of America, ²¹Texas Diabetes and Endocrinology, Austin, United States of America, ²²Diabetes and Thyroid Center, Fort Worth, United States of America, ²³Jaeb Center for Health Research, Tampa, United States of America, ²⁴Insulet Corporation, Acton, United States of America

Background and Aims: Many people with T2D may not use carbohydrate counting for meal bolusing, so it is important to understand whether simplified meal bolus approaches can be used with AID. The Omnipod® 5 AID System demonstrated safety and efficacy in adults (≥18y) with T2D and was recently cleared by the US FDA for use in this population (also FDA cleared/CE marked for ages ≥2y with type 1 diabetes). This SECURE-T2D sub-analysis evaluated outcomes by bolus strategy with AID use.

Methods: This multicenter single-arm trial enrolled insulin-treated adults aged 18-75y with T2D with HbA1c <12.0%. After a 14-day period to capture baseline data, participants received carbohydrate counting training and were advised to carbohydrate count or use a simplified bolus strategy during the 13-week AID period. Glycemic outcomes according to bolus strategy used with AID were evaluated: carbohydrate counting, small/medium/large carbohydrate entry, or fixed carbohydrate entry.

Results: A total of 305 participants (mean age 57±11y, 24% Black, 22% Hispanic/Latino, 21% using basal insulin without mealtime insulin, 55% using GLP1-RA) initiated AID. Of the 289 who completed the study, 59% used carbohydrate counting, 35% used small/medium/large carbohydrate entry, 4% used fixed carbohydrate entry, and 2% used other or a combination strategy. Similar improvements in HbA1c, time in range, and time >250mg/dL (>13.9 mmol/L) were achieved across strategies, with time <70mg/dL (<3.9mmol/L) unchanged (Table). Participants delivered 3.2±1.7 (mean±SD) boluses/day.

Conclusions: These results provide evidence that glycemic improvements among people with T2D using simplified bolus strategies are comparable to a carbohydrate counting based bolus strategy with the Omnipod 5 System.

RETROSPECTIVE ANALYSIS OF RAW CGM AND PUMP DATA FROM A T1D REAL-LIFE COHORT WITH HYBRID CLOSED LOOP: EFFECT OF PRANDIAL BEHAVIOR ON METABOLIC CONTROL

Bleuenn Dreves, Margot Krikorian, Julia Morera, Anne Rod, Michael Joubert

Caen University Hospital, UNICAEN, Diabetes Care Unit, Caen, France

Background and Aims: Hybrid closed-loop (HCL) systems require premeal announcements but many patients forget or announce too late carbs. We explore the effect of these behaviors on glycemic control.

Methods: Retrospective analysis of raw CGM and pump data from a cohort of T1D patients after 3-months with Medtronic 780G (780G) or Tandem Control IQ (CIQ) in real life conditions. Time-stamped blood glucose were analyzed with GRID methodology and connected to carbohydrate declarations: for each patient, food intake was classified in announced meals (including single bolus (SB), multiple bolus (MB), inadequate bolus (IB), late bolus (LB)) or unannounced meals (no bolus (NB)). Four hours post-prandial TIR (4hppTIR) was calculated for these different meal categories, as their association with global TIR.

Results: We analyzed 108 patients: 780G/CIQ 75/25%; age 48±14 years; diabetes duration 24±15 years; HbA1c (M3 post initiation) 6.9±0.3%. Mean daily food intake was 4,2±0,9/day, including 65±13% announced meals (SB 46±13%; MB 8±10%; IB 8±7%; LB 3±2%) and 35±13% unannounced meals (NB). 4hppTIR was 63±18% after SB (ref), not different after MB (59±28%) or IB (55±21%), but lower after LB (47±20%; p<0.01) or NB (52±12%; p<0.05) (figure). Percentage unannounced meals were correlated with TAR (r=0.391; p<0.05) but not with TIR nor TBR.

Conclusions: MB and IB do not decrease post-prandial TIR compared to SB. Unannounced meals (NB) and LB are frequent and are followed by a decreased post-prandial TIR. However, unannounced meals are not associated with decreased global TIR, suggesting that HCL systems can compensate for suboptimal prandial behavior.

A COMPREHENSIVE MATHEMATICAL MODEL TO SIMULATE THE EFFECT OF PHYSICAL ACTIVITY ON GLUCOSE UTILIZATION AND PRODUCTION IN INDIVIDUALS WITH TYPE 1 DIABETES

Edoardo Faggionato ¹, Chiara Dalla Man¹, Rita Basu², Ananda Basu², Michele Schiavon¹

¹University of Padova, Department Of Information Engineering, Padova, Italy, ²University of Alabama at Birmingham, School Of Medicine - Endocrinology, Diabetes, And Metabolism, Birmingham, United States of America

Background and Aims: Physical Activity (PA) is one of the major disturbances in type 1 diabetes (T1D) management. However, most of the available in silico platforms for testing glucose control strategies still lack a reliable module to simulate PA. To bridge this gap, we developed and tested a comprehensive model simulating PA effect on postprandial glucose fluxes for its incorporation into simulation platforms.

Methods: We analyzed data from 16 individuals with T1D who underwent a triple-tracer mixed-meal of 75 g of carbohydrates and subsequently performed 60 min of PA (50% VO₂ max) over a period of 75 min starting 120 min after the meal. Plasma glucose, insulin, glucagon, and tracer concentrations were frequently measured. Tracer concentrations were used to estimate meal glucose appearance (R_a), endogenous glucose production (EGP), and glucose rate of disposal (R_d). The model includes a module describing glucose-insulin-glucagon dynamics (Dalla Man 2006), gastrointestinal absorption (Dalla Man 2006), endogenous glucose production (Dalla Man 2008), and PA effect (Romeres 2021). Model identification was done simultaneously on plasma glucose, R_a, and EGP via Bayesian estimation using insulin and glucagon concentration as forcing functions. Model-predicted R_d was validated against the tracer-estimated R_d.

Results: The model was able to describe individual profiles well, providing precise and physiologically plausible parameter estimates. The model was then used to simulate PA effect on glucose fluxes (Fig.1).

Conclusions: A mathematical model of PA effect on postprandial glucose fluxes was developed. The model allows simulating PA on glucose dynamics providing a robust tool for optimizing diabetes management in these experimental conditions.

REAL-LIFE EXPERIENCE WITH THE MEDTRONIC MINIMED™ 780G ADVANCED HYBRID CLOSED-LOOP SYSTEM IN PEDIATRIC PATIENTS AND YOUNG ADULTS WITH TYPE 1 DIABETES

Estela Gil-Poch ¹, Francisco Javier Arroyo-Díez¹, Pilar Beato-Víbora²

¹PEDIATRIC UNIVERSITY HOSPITAL BADAJOZ, Pediatric Endocrinology, Badajoz, Spain, ²Badajoz University Hospital, Badajoz, Spain

Background and Aims: The benefit in metabolic control in patients treated with the Medtronic MiniMed™780G system is well known with extensive results in real life. The objective of the study is to analyze glycemic control in pediatric patients with T1D after treatment switch to the Medtronic™780G system, and this relationship with device settings, insulin requirements and body composition.

Methods: A data set of 106 children with T1D initiating AHCL between December 2020-September 2024 was analyzed. Baseline data, 6-12-24 and 36 months after the start of treatment were compared.

Results: 106 patients, 54 of them men, were studied with a mean age of 14.6±3.8 years and a mean duration of T1D of 7.1±4.1 years. The patients had previous treatment with MiniMed™640G system, MiniMed™670G, or multiple daily injections associated CGM with DEXCOMG6™ or flash glucose monitoring with FREESTYLE2. After the start of treatment, a statistically significant reduction in HbA1c was observed, as well as an increase in the time in the target range 70-180 mg/dl. A significant improvement in the time in hypoglycemia appeared with continued use of the system. (Table 1). Initial settings for young children (<10 years) were an active insulin time (AIT) of 3 hours and a glucose target (GT) of 120 mg/dl, and an AIT of 2 hours and a GT of 100 mg/dl for older patients (>10 years). The insulin requirement decreased but a significant increase in BMI appeared.

Conclusions: The MiniMed™780G advanced system improves metabolic control in pediatric patients with T1D and this improvement is maintained with continue use of the system.

HYBRID CLOSED LOOP OB2F NATION-WIDE SURVEY: DETERMINANTS OF ACHIEVEMENT OF GLUCOSE CONTROL TARGETS WITH TWO DIFFERENT SYSTEMS IN REAL-LIFE CONDITIONS

Michael Joubert ¹, Jean-Baptiste Julla², Eric Renard³, Sandrine Lablanche⁴, Agnes Sola⁵, Jean-Francois Gautier², Jean-Pierre Riveline⁶

¹Caen University Hospital, UNICAEN, Diabetes Care Unit, Caen, France, ²Hôpital Lariboisière - AP-HP, Paris, France, ³University of Montpellier, Endocrinology, Diabetes, Metabolism, Montpellier, France, ⁴Grenoble University Hospital, Endocrinology Diabetes, Grenoble, France, ⁵cochin APHP, paris, France, ⁶Hôpital Lariboisière, Cudc Centre Universitaire Du Diabète Et De Ses Complications, paris, France

Background and Aims: All automated insulin delivery (AID) systems have shown improved glucose control in clinical trials, involving strict selection and monitoring, which may not reflect real-life conditions. An extensive nation-wide survey, named OB2F, was conducted to assess metabolic efficacy of AID during one-year real-life use. The aim of this analysis was to explore the determinants of achieving glucose control targets among Medtronic 780G (780G) or Tandem Control IQ (CIQ) users.

Methods: All adults who initiated 780G or CIQ between January 1st and December 31st, 2022, were included in 79 centers. Clinical data, continuous glucose monitoring metrics, and HbA1c levels were collected at AID initiation and after 12 months of follow-up. AID users were divided in TIR >70/≤70% subgroups according to achieved TIR target at M12. A logistic regression was performed on univariate significant differences.

Results: A total of 1281 patients were analyzed (780G n= 641; CIQ n=640). Among 780G users, 409 patients (64%) reached a TIR >70%. Among CIQ users, 291 patients (45%) reached a TIR >70%. Patient characteristics of subgroups are detailed in table 1. Multivariate analysis showed that lower baseline HbA1c, higher baseline TIR were associated with M12 TIR>70% for both systems. In addition, longer diabetes duration and lower CV were also associated with this target achievement, for 780G and CIQ, respectively.

Conclusions: In a real-life adult cohort of 780G and CIQ users, better baseline metabolic control (lower HbA1c, higher TIR) was the main determinant for achieving TIR >70% after 12-month follow-up.

REAL-WORLD HBA1C FOLLOWING INITIATION OF THE OMNIPOD 5 AUTOMATED INSULIN DELIVERY SYSTEM

Carol Levy ¹, Dan Carlow², Daniel Swift², Rana Rittgers-Simonds², Noelle Gronroos³, Erin Buysman³, Valery Walker³, Gregory Forlenza⁴

¹Icahn School of Medicine at Mount Sinai, Division Of Endocrinology, Diabetes And Metabolism, New York, United States of America, ²Insulet Corporation, Acton, United States of America, ³Optum, Eden Prairie, United States of America, ⁴University of Colorado Anschutz Medical Campus, Barbara Davis Center For Diabetes, Aurora, United States of America

Background and Aims: The Omnipod 5 Automated Insulin Delivery (AID) System is cleared in the United States (US) for type 1 diabetes (T1D) (2+years) and type 2 diabetes (T2D) (18+years) and CE marked for T1D (2+years). Clinical trials have demonstrated HbA1c improvements with Omnipod 5 use. This claims-based analysis evaluated HbA1c changes with Omnipod 5 in a real-world setting.

Methods: Commercial and Medicare Advantage plan members with T1D or T2D initiating Omnipod 5 (index date) between August to December 2022 were identified from the Optum Research Database, a large US insurance claims database. A retrospective analysis compared HbA1c values in the 12 months pre-initiation with Omnipod 5 to 12 months post-initiation. The HbA1c closest and prior to the index date was the pre-initiation measure, and the latest HbA1c ≥90 days following index date was the post-initiation HbA1c.

Results: Analysis included 2,504 users initiating Omnipod 5 with a mean age of 42 years (SD=21). Most had T1D (76%), commercial insurance (70%), and previously used an insulin pump (63%) and continuous glucose monitoring (92%); 792 users (32%) had HbA1c results pre- and post-initiation. Percentages of users meeting guideline targets were higher post-initiation compared to pre-initiation (p<0.001). Mean reduction in HbA1c of 0.4% overall (p<0.001) and 1.4% among those with pre-initiation HbA1c ≥9% (n=135, p<0.001) was observed. HbA1c reductions were demonstrated among both users with T1D and T2D (p<0.001).

Conclusions: This claims analysis demonstrates the effectiveness of Omnipod 5 in reducing HbA1c and achieving across a wide glycemic range in a real-world setting.

HIGH TIME IN TIGHT RANGE WITH SLEEP ACTIVITY TARGETS IN CONTROL-IQ TECHNOLOGY

Laurel Messer, Sneha Thanasekaran, Jamel Thomas, Jordan Pinsker

Tandem Diabetes Care, San Diego, United States of America

Background and Aims: Control-IQ technology Sleep Activity (SLEEP) predictively adjusts insulin to target glucose levels in a tight range (112.5-120 mg/dl). SLEEP targets use enhanced dynamic insulin delivery every 5 minutes to correct for hyperglycemia, without the need for hourly automatic correction boluses. SLEEP target insulin delivery is impacted by basal rates and correction factor settings. This study examines people with diabetes who use SLEEP targets >80% of the time and subsequent glycemic outcomes.

Methods: Data on SLEEP users with type 1 diabetes were analyzed from the Tandem Diabetes Care database from 2020-2024 The most recent 30 days of data with 70% CGM data availability were analyzed for insulin delivery and glycemic outcomes, represented as median (IQR).

Results: 7,266 people used SLEEP targets an average of 24 hours/day [60% female, median age 40 years (30, 55), diabetes duration 23.2 years (12.8, 33.8)]. Median Time in Range (70-180 mg/dl) for SLEEP target users was 75.6% (63.1, 84.9) and Time in Tight Range was 49.4% (36.5, 62.0) (table). Users of SLEEP targets averaged 5.6 boluses per day, including almost 3 boluses for meals.

Conclusions: Use of SLEEP targets 24/7 is associated with high Time in Range and Time in Tight Range. The more aggressive hyperglycemia correction every 5 minutes, accompanied by users bousing routinely, offer a useful strategy for optimal glycemic control with Control-IQ technology.

REAL-WORLD GLYCEMIC OUTCOMES OF >13,800 ADULTS WITH TYPE 2 DIABETES (T2D) USING THE OMNIPOD® 5 AUTOMATED INSULIN DELIVERY (AID) SYSTEM WITH CLOUD-BASED DATA MANAGEMENT

Eden Miller ¹, Anne Peters², Francisco Pasquel³, Sean Oser⁴, Grazia Aleppo⁵, Lauren Huyett⁶, José Méndez⁶, Irene Hadjiyianni⁶, Lindsey Conroy⁶, Trang Ly⁶

¹Diabetes and Obesity Care LLC, Bend, United States of America, ²Keck School of Medicine of the University of Southern California, Los Angeles, United States of America, ³Emory University, Atlanta, United States of America, ⁴University of Colorado School of Medicine, Aurora, United States of America, ⁵Northwestern University, Feinberg School Of Medicine, Chicago, United States of America, ⁶Insulet Corporation, Acton, United States of America

Background and Aims: The Omnipod® 5 AID System demonstrated safety and efficacy in adults (≥18y) with T2D in a multicenter single-arm pivotal trial and was recently cleared by the US FDA for use in this population (also FDA cleared/CE marked for ages ≥2y with type 1 diabetes). The system allows for personalized therapy through customizable glucose targets and enables automatic upload of data for all US users, facilitating access to evaluate real-world outcomes. As real-world evidence can provide insights from individuals with T2D already using the system, this study aimed to evaluate the glycemic benefit of Omnipod 5 use in the real world.

Methods: A retrospective analysis of continuous glucose monitoring (CGM) and insulin data from Omnipod 5 users with self-reported T2D aged ≥18y and sufficient CGM data (≥90 days of data, ≥75% of days with ≥220 readings) was conducted according to average glucose targets for the duration of device use.

Results: Data from 13,824 users meeting the inclusion criteria were available at the time of analysis, with >4.5 million person-days of data. Among those who self-reported HbA1c during system onboarding (n=11,835), 17% reported HbA1c >9.0% (Table). Median time in range (TIR; 70-180mg/dL; 3.9-10.0mmol/L) was 69.1%, 63.2%, and 54.7% for the 110mg/dL [6.1mmol/L], 120mg/dL [6.7mmol/L], and 130-150mg/dL [7.2-8.3mmol/L] targets, respectively. Across glucose targets, time below range (<70mg/dL; <3.9mmol/L) was low (median ≤0.26%) and most users delivered 2-4 boluses/day.

Conclusions: These data provide valuable insights to the favorable impact of AID on real-world glycemic outcomes among adults with T2D using the Omnipod 5 System in the US.

DEMONSTRATION OF THE VIABILITY OF INSULIN-ONLY FULL CLOSED LOOP THERAPY IN ADULTS WITH TYPE 1 DIABETES (T1D) AT HOME

Marcela Moscoso Vasquez ¹, Sue Brown^1,2, Mark Deboer^1,3, Samina Afreen², Morgan Fuller¹, Viola Holmes¹, Laya Ekhlaspour⁴, Gregory Forlenza⁵, Marc Breton¹

¹University of Virginia, Center For Diabetes Technology, Charlottesville, United States of America, ²University of Virginia, Dept Of Medicine - Division Of Endocrinology, Charlottesville, United States of America, ³University of Virginia, Pediatrics, Charlottesville, United States of America, ⁴University of California, Pediatrics, San Francisco, United States of America, ⁵University of Colorado Anschutz Medical Campus, Barbara Davis Center For Diabetes, Aurora, United States of America

Background and Aims: Automated insulin delivery (AID) is included in T1D standards of care but relies heavily on timely and well-dosed prandial insulin boluses. Imperfect bolusing behaviors using AID have been associated with degraded glycemic management; to this day, no insulin-only device has provided tight glycemic control in full closed loop (FCL).

Methods: Adults (>25yo), young adults (18-25yo) and adolescents (14-18yo) were enrolled at three US clinical sites to test a novel AID system (AIDANET) in FCL based on the UVA AID architecture and embedding an artificial neural network to dose insulin. Participants first spent five days using AIDANET supervised at a hotel, followed by 7 days at home with real time monitoring. Usual care data during usual care arm was collected randomly before/after AIDANET use (www.clinicaltrials.gov NCT06041971). Home-collected CGM outcomes are reported for the adult cohort, usual care vs FCL.

Results: Twenty adults signed consent at UVA and 12 started and completed data collection: 8F/4M; body weight 94.4kg [59.4-136.8], HbA1c 7.6% [5.7-8.8], diabetes duration 28y [10-41]. All were CGM and pump users, with 11/12 using AID. When in FCL, participants achieved 69.7% time-in-range 70-180mg/dL vs 62.6% in usual care, significantly reducing time-above-180mg/dL (36.4% vs 29%, p=0.047), and average CGM (155.9mg/dL vs 167.52mg/dL, p=0.047); improving time-in-tight-range (46.2% vs 37.5%, p=0.038). Time-below-70mg/dL was marginally increased (1.4% vs 1.1%). These results were reflected in overnight and daytime glycemic control.

Conclusions: In this interim analysis, AIDANET in FCL provided superior glycemic control vs usual care (primarily AID) in adults with T1DM >25yo.

GLYCAEMIC CONTROL AND CGM-BASED METRICS OF DIFFERENT COMMERCIALLY AVAILABLE AID SYSTEMS IN PEOPLE WITH TYPE 1 DIABETES: A SYSTEMATIC REVIEW AND META-ANALYSIS

Othmar Moser ¹, Faisal Aziz², Christoph Sternad², Lena Sternad², Max L. Eckstein³, Katarina Braune⁴, Sabrina Sanfilippo¹, Harald Sourij²

¹University of Bayreuth, Exercise Physiology And Metabolism, Institute Of Sports Science, Bayreuth, Germany, ²Medical University of Graz, Division Of Endocrinology And Diabetology, Graz, Austria, ³AstraZeneca GmbH, Hamburg, Germany, ⁴Charité, Institute Of Medical Informatics, Berlin, Germany

Background and Aims: Although different AID systems are commercially available, an evaluation of their users’ glycaemic outcomes remains incomplete. Within this systematic review and meta-analysis, we assessed glycaemia in people with T1D using different AID systems, focusing on alignment with therapeutic guidelines, particularly HbA1c and CGM metrics.

Methods: A systematic literature search of Medline/PubMed, Cochrane CENTRAL, Scopus and Embase through March 2024 included randomized controlled trials (RCTs) using commercial AID systems for at least 28 days. A single-mean meta-analysis, sub-group, and meta-regression analyses were performed to evaluate differences in HbA1c and CGM metrics with respect to AID systems and other moderators, using random-effect models and 95% confidence intervals.

Results: Out of 899 studies, 41 trials with 2,770 participants met inclusion criteria, detailing a median [IQR] AID system use of 13 weeks [8-24]. For all AID systems, the random effects model for HbA1c achieved at study end was found at 7.08% [6.94; 7.21], for iLet Beta Bionics 7.21% [7.06; 7.36], CamAPS FX at 6.63% [6.47; 6.79], Medtronic MiniMed 670G at 7.34% [6.80; 7.89], Medtronic MiniMed 780G 6.75% [6.33; 7.18] and Tandem T:slim 7.17% [6.99; 7.35] (DBLG1: N/A). While TIR >70% was not achieved by several systems, time-below-range remained below the recommended <4% for all (Table 1).

Conclusions: While most AID systems prevent hypoglycemia, variability in meeting HbA1c and TIR targets suggests system selection should be individualized. Future research should focus on improving TIR without compromising safety, with longer-term and comparative effectiveness studies.

MANAGING THE “POTATO CHALLENGE”: EFFICACY OF AUTOMATED INSULIN DELIVERY IN MAINTAINING GLYCEMIC CONTROL

Stefano Passanisi ¹, Bruno Bombaci¹, Dario Iafusco², Mattia Papa¹, Assunta Serena Rollato², Fortunato Lombardo¹, Magna Graecia Camp³

¹University of Messina, Department Of Human Pathology In Adult And Developmental Age “gaetano Barresi”, Messina, Italy, ²University of Campania “Luigi Vanvitelli”, Department Of The Woman, Of The Child And Of The General And Specialized Surgery, Naples, Italy, ³Magna Graecia Camp, Crotone, Italy

Background and Aims: Potatoes are among the most commonly consumed foods, particularly in pediatric populations, but they present a unique challenge for people with T1D. This study aimed to assess glycemic patterns in young individuals with T1D using a second-generation AID system after they consumed potatoes prepared in two different ways: fried and boiled.

Methods: The study was conducted during a five-day school camp for unaccompanied youth with T1D aged 11-17. All participants had been using the Tandem t:slim X2™ Control IQ insulin pump for at least six months. On two separate days, each participant consumed a standardized meal consisting of 240 grams of either fried or boiled potatoes. The insulin bolus was administered 15 minutes before the meal, regardless of sensor glucose levels. Raw CGM data were collected and analyzed for all participants.

Results: There were no significant differences in the AUC of glucose levels between the two types of potatoes during the 3-hour post-prandial period (p = 0.374). TIR was slightly higher, but not statistically significant, after consuming boiled potatoes compared to fried ones (73.7% vs. 67.8%; p = 0.225). Interestingly, in both cases, glucose levels 3 hours post-meal were significantly lower than pre-meal levels (boiled: p = 0.003; fried: p = 0.005). Despite the differences in preparation methods, glucose level variations from pre- to post-meal were similar across both groups (p = 0.638).

Conclusions: Despite differences in glycemic index and fat content between fried and boiled potatoes, the Tandem t:slim X2™ Control IQ ensured a satisfactory glycemic response in both cases, with no significant differences observed.

NEURAL NETWORKS FOR MINIATURIZED MPC IMPLEMENTATION

Elliott Pryor, Anas El Fathi, Alberto Castillo Frasquet, Meryem Karagoz, Marcela Moscoso Vasquez, Marc Breton

University of Virginia, Center For Diabetes Technology, Charlottesville, United States of America

Background and Aims: Next-generation diabetes management systems, like Model Predictive Control (MPC), optimize insulin delivery but exceed the computational limits of insulin pumps. Without reducing memory usage and processing demands, these advanced control schemes cannot be implemented in real-world devices. To address this, we developed compact Neural Networks (NNs) that replicate MPC performance while fitting within these constraints, enabling the deployment of cutting-edge insulin dosing algorithms.

Methods: NNs for insulin dosing have previously been validated in clinical trials, and we focused on further reducing memory usage to enable implementation in insulin pumps. A large dataset, combining both simulated and real clinical data, was used to train the networks. Two types of architectures—ResNET and standard MLP—were employed, both using ReLU activation for easier hardware compatibility. Various network sizes were tested to find the balance between accuracy and resource usage. Simulations assessed performance under realistic conditions.

Results: The networks accurately reproduced the MPC, with only those smaller than 6KB showing statistically significant differences, with a maximum 0.24% increase in Time Below 70 mg/dL, which is not clinically significant. The network requires <10KB of memory, runs 200 times faster than MPC evaluation, and is 300 times smaller than previously validated designs.

Conclusions: These optimized NNs can be implemented in insulin pumps without sacrificing performance, improving both the efficiency and battery life of the devices. This allows for the integration of next-generation insulin delivery algorithms into current pump technology, advancing diabetes care.

HOW DO OUTCOMES WITH THE OMNIPOD® 5 AUTOMATED INSULIN DELIVERY (AID) SYSTEM COMPARE FOR PERSONS WITH TYPE 1 VERSUS TYPE 2 DIABETES? LEARNINGS ACROSS STUDIES

Viral Shah ¹, Lori Laffel², Anders Carlson³, Kristin Castorino⁴, Kellee Miller⁵, Lauren Huyett⁵, Lindsey Conroy⁵, Trang Ly⁵

¹Indiana University School of Medicine, Division Of Endocrinology & Metabolism, Indianapolis, United States of America, ²Joslin Diabetes Center, Harvard Medical School, Boston, United States of America, ³HealthPartners, International Diabetes Center, Minneapolis, United States of America, ⁴Sansum Diabetes Research Institute, Santa Barbara, United States of America, ⁵Insulet Corporation, Acton, United States of America

Background and Aims: Research on AID systems has largely centered on type 1 diabetes (T1D); however, the Omnipod® 5 AID System recently demonstrated efficacy and safety in adults (≥18y) with T2D and is cleared by the US FDA for use in this population (also FDA cleared/CE marked for ages ≥2y with type 1 diabetes). This analysis aimed to compare glycemic outcomes with the Omnipod 5 System between T1D and T2D.

Methods: Three clinical trials of the Omnipod 5 System were compared: a 13-week single arm trial in T2D (N=305, ages ≥18y, prior basal-only or basal/bolus insulin therapy) and in T1D (N=320, ages 2-70y), and a 13-week randomized controlled trial in T1D (AID arm: n=132, ages ≥18y). Glycemic outcomes were assessed in those with T1D and T2D across studies according to baseline HbA1c levels (<7%, 7.0-7.9%, 8.0-8.9%, ≥9%).

Results: For participants with baseline HbA1c <7%, time in range (TIR, 70-180mg/dL [3.9-10.0mmol/L]) with AID was 82% in T2D and 77% in T1D, and HbA1c was 6.5% and 6.3%, respectively, at end of study. Among those with baseline HbA1c ≥9.0%, similar TIR and HbA1c were observed with AID use. Time in hypoglycemia did not increase with AID across study groups, despite being markedly lower in T2D versus T1D across all baseline HbA1c levels.

Conclusions: Regardless of baseline HbA1c, glycemic outcomes after 13-week treatment with Omnipod 5 were similar between T1D and T2D study populations. These results support that people with T2D requiring insulin therapy (basal-only or basal/bolus) in the US should be given equal opportunity to use AID technology.

DEEP REINFORCEMENT LEARNING BOLUS PRIMING FOR INTELLIGENT PRANDIAL MANAGEMENT

Ali Tavasoli, Anas El Fathi, Heman Shakeri, Boris Kovatchev, Marc Breton

University of Virginia, Charlottesville, United States of America

Background and Aims: Current AID systems in T1D typically operate in hybrid closed-loop configurations, integrating CGM with insulin pumps relying on meal announcements for effective glucose control. To advance towards fully automated closed-loop, we employ deep reinforcement learning (RL) to address challenges in meal detection and prediction under varying conditions. Our RL approach acts as a feedforward control, predicting meals and delivering timely boluses, and is designed to work in addition to established control strategies.

Methods: We propose a smart bolus priming system (BPS) for seamless integration with advanced basal rate controllers to complete a fully automated AID system. Our RL agent is trained on eighty virtual subjects from the FDA-accepted UVA/Padova T1D adult cohort and validated on an additional twenty. We evaluate performance by integrating our smart RL BPS with different therapies, replacing existing bolus delivery mechanisms in both traditional and cutting-edge setups to demonstrate its effectiveness across various scenarios.

Results: Table 1 summarized the marked improvements of using RL BPS over both traditional and state-of-the-art methods across various metrics. The RL agent takes a more significant role in the prandial control and administers higher bolus doses when applied to a legacy implementation of control -Control-IQ (CIQ), and Automated Insulin Delivery Adaptive Network (AIDANet).

Conclusions: Our proposed RL BPS algorithm improves glycemic control in AID systems enabling them to work in fully closed loop mode and easy microchip deployment. By reducing reliance on manual inputs and adapting to individual needs, our system has the potential to significantly simplify T1D management.

IS SIMPLIFIED MEAL ANNOUNCEMENT AN ALTERNATIVE TO EXACT CARBOHYDRATE COUNTING IN PATIENTS WITH AN ADVANCED HYBRID CLOSED LOOP SYSTEM? A SYSTEMATIC REVIEW AND META-ANALYSIS

Cristina Sicorschi Gutu ^1,2, Patricia-Maria Anton³, Paweł Łajczak⁴, Cintia Gonzalez Blanco⁵

¹Paracelsus Medical University, 2. university Clinic For Psychiatry And Psychotherapy, Nuremberg, Germany, ²Medgate AG, Allschwil, Switzerland, ³National Insitute of Endocrinology “CI Parhon”, Bucharest, Romania, ⁴Medical University of Silesia, Katowice, Poland, ⁵Hospital General Universitario de Valencia, Valencia, Spain

Background and Aims: Hybrid closed-loop (HCL) systems have become the standard of care for type 1 diabetes patients. However, they require exact carbohydrate counting (ECC), which can be complex. We conducted a meta-analysis to assess whether simplified meal announcement (SMA) is inferior to ECC by comparing the percentage of time spent in the Time in Range (TIR).

Methods: PubMed, EMBASE and Cochrane Central database were searched for randomized controlled trials (RCTs) that compared ECC to SMA and reported the outcomes of (1) percentage of 3.9-10 mmol/l glucose level time; (2) percentage of 3.9-7.8 mmol/l glucose level time; (3) total daily insulin units (TDI) per kg; and (4) coefficient of variation (CV) of glucose (%). Heterogeneity was examined with I² statistics. A random-effect generic inverse variance (GIV) method was applied to all analyses.

Results: We included 4 RCTs (three of them crossover) with 137 patients, of whom 118 underwent SMA. Follow-up ranged from 2 weeks to 12 months. The pooled mean difference (MD) of percentage of time spent in TIR was -3.30 [-6.09, -0.50], indicating lower proportion with SMA. Two studies reported the percentage of TIR of 3.9-7.8 mmol/l, with a pooled effect size MD of -3.36 [-5.80, -0.92]). The pooled MD in TDI units per kg was 0.00 [-0.04, 0.04], and CV of glucose was non-significant (MD 0.04 [-0.87, 0.95]).

Conclusions: Given the differences between the strategies, simplified meal announcement offers a valid alternative, delivering similar results with less effort and burden for people with diabetes.

IMPACT OF THE OMNIPOD 5 SYSTEM ON GLYCEMIC CONTROL AND PSYCHOSOCIAL OUTCOMES IN VERY YOUNG CHILDREN WITH TYPE 1 DIABETES (T1D)

Kasserine Taylor ¹, Estella Escobar¹, Casey Sakamoto^1,2, Gregory Forlenza¹, Paul Wadwa¹, Erin Cobry¹, Cari Berget¹

¹University of Colorado Anschutz Medical Campus, Barbara Davis Center For Diabetes, Aurora, United States of America, ²Colorado School of Public Health, CU Anschutz Medical Campus, Department Of Biostatistics And Informatics, Aurora, United States of America

Background and Aims: Managing T1D in very young children presents unique challenges when using Automated Insulin Delivery (AID) systems. We sought to evaluate system use and glycemic outcomes in a sample of very young children starting the Omnipod 5 system for their clinical care.

Methods: Children aged 2-5 years with T1D prescribed the Omnipod 5 system were enrolled in a 12-month prospective observational study. Data on system use and glycemic outcomes were collected from device downloads. Changes in system use and outcomes over time were modeled with a Linear Mixed Model with random intercept for subject and contrasts comparing each timepoint (1, 3, 6, 9, and 12 months) to baseline. Questionnaires assessing psychosocial outcomes (Hypoglycemic Fear Survey [HFS], INSPIRE, and Problem Areas in Diabetes [PAID]) were completed by caregivers at baseline, 3, 6, 9, and 12 months after starting the system.

Results: Twelve children (Age 3.8 ±1.2 yrs, T1D duration 1.4 ±0.9 yrs, Baseline HbA1c 7.5 ±0.9%, 91.7% non-Hispanic White) completed the study. Statistically significant improvements were seen for changes in HbA1c (7.5% to 6.9%, p=0.005) and % Time in Range (60.3% to 67.5%, p=0.002). % Time Below Range remained low. Psychosocial scores from PAID improved from 53.9 to 46.2 (p=0.027). There were no changes in HFS or INSPIRE scores.

Conclusions: AID systems can improve glycemic outcomes of very young children with T1D while reducing caregiver burden. This data suggests the OP5 system should be considered in this age group, which has previously been a challenging population to maintain tight glycemic control without causing frequent hypoglycemia.

AN IN-DEPTH ANALYSIS ON HYPOGLYCEMIA DURING USE OF THE INREDA ARTIFICIAL PANCREAS: A BIHORMONAL FULLY CLOSED-LOOP SYSTEM FOR TYPE 1 DIABETES

Stennie Zoet ^1,2, Thomas Urgert^1,2, Lyan Vlaskamp³, Bert-Jan Van Beijnum¹, Gozewijn Laverman^1,2

¹Universiteit Twente, Biomedical Signals And Systems, Enschede, Netherlands, ²Ziekenhuisgroep Twente, Almelo, Netherlands, ³Inreda Diabetic, Goor, Netherlands

Background and Aims: The Inreda bihormonal artificial pancreas (AP; Inreda Diabetic, Goor), has previously proven excellent long-term glycemic control in people with type 1 diabetes. Of specific interest is the performance of the AP on hypoglycemia. This study comprises an in-depth analysis into hypoglycemic events with the AP.

Methods: One-year data from 24 diabetes patients using the Inreda AP as their regular treatment at the Ziekenhuisgroep Twente hospital, the Netherlands, were analyzed. The average frequency, duration, depth, severity, and shape of hypoglycemic events (glucose <3.9 mmol/L for ≥5 minutes) were calculated from continuous glucose monitoring data and compared with prior therapy.

Results: Mean age was 49.0±11.6 years and 41.7% were female. Prior treatment consisted of conventional subcutaneous pump with intermittent (n=11) or continuous (n=10) sensoring (one of whom hybrid closed loop) and multiple daily injections (n=3). Mean TIR was 78.9%±6.9% with AP versus 60.2%±15.7% with prior therapy (p<0.01). The mean area under the curve for hypoglycemic events significantly decreased during AP treatment (23.5±12.7 to 8.6±2.9 min⋅mmol/L, p < 0.01). Event duration and glucose drops during events also significantly reduced (length was 42.5±13.4 vs. 27.9±7.9 minutes, and average glucose level was 3.51±0.11 vs. 3.63±0.04 mmol/L,). Time below range significantly decreased (2.8% (IQR:1.2%-4.5%) to 1.5%±1.0%, p<0.01), while event frequency remained unchanged (1.0±0.6 vs. 0.8±0.5 per day, p=0.315).

Conclusions: This real-world study shows that the Inreda AP provides long-term improvement on the severity, duration, and shape of hypoglycemic events, contributing to tighter glucose regulation compared with the previous conventional therapy.

ICAN CGM SYSTEM - PERFORMANCE AND RELIABILITY EVALUATION OF AN INNOVATIVE CGM SYSTEM

Jiangfeng Fei ¹, Fei Gao¹, Yuanyuan Yu¹, Jiyun Zheng¹, Ao Gao¹, Qian Zhou², Frank Flacke³, Yufeng Li⁴, Xiang Yan⁵, Linong Ji⁶

¹Sinocare Inc., R&d, Changsha, China, ²Sinocare Inc., Department Of Regulatory Affairs, Changsha, China, ³Sinocare Inc., Medical Affairs, Mainz, Germany, ⁴Beijing Pinggu Hospital, Department Of Endocrinology, Beijing, China, ⁵The Second Xiangya Hospital of Central South University, National Clinical Research Center For Metabolic Diseases, And Department Of Metabolism And Endocrinology, Changsha, China, ⁶Peking University People’s Hospital, Department Of Endocrinology, Beijing, China

Background and Aims: The CE-marked iCan CGM system, an all-in-one device with 15 days wear time and proprietary innovative sensor technology, was developed by Sinocare Meditech Inc. and is available in numerous countries worldwide. This data from the multicenter registration study conducted in China according to international standards assessed the performance of iCan CGM system.

Methods: This single arm, prospective multicenter study evaluated the iCan CGM system in one of 2 subgroups. There were 60 patients with T1D and T2D aged ≥18yrs enrolled into this home-care subgroup and at days 2, 7-9 and 15 the performance of the device was assessed by comparison to YSI readings. The data were analyzed according to generally accepted procedures.

Results: The overall MARD was calculated to 8.71% (95%CI: 7.86 – 9.55%), with a 20/20% performance of 93.65% (95%CI: 92.37 – 94.77%). The MARD at the start, mid and end of wear time can be reported as 8.90% (0.03), 7.67% (0.03), and 9.62% (0.03) respectively. Sensor survival rate is estimated to 99.07% of sensors reaching the 15 days label wear time. The Consensus error grid analysis shows 100% (95%CI: 99.77% - 100.00%) in Zones A+B. There were no device related and skin related Adverse Events reported during the study.

Conclusions: The iCan CGM system, launched in markets worldwide, demonstrated in a multicenter study overall accuracy performance on par with established CGMs, with a single digit MARD throughout the labeled 15d wear time, combined with an outstanding system survival rate. (NCT0599575).

PROGRESS REPORT ON THE DEVELOPMENT OF A BREATH-BASED GLUCOSE ESTIMATION ALGORITHM

Frank Flacke ¹, Clarisse François-Marsal², Theo Schiavi², Laurent Cazals³, Benoit Lepage⁴, Pierre Gourdy³, Sandrine Isz²

¹BOYDSense SAS, Toulouse, France, ²BOYDSense SAS, R&d, Toulouse, France, ³CHU de Toulouse - Hôpital Rangueil, Service De Diabétologie, Maladies Métaboliques Et Nutrition, Toulouse, France, ⁴CHU de Toulouse - Hôpital Rangueil, Département D’epidémiologie, Economie De La Santé Et Santé Publique, Toulouse, France

Background and Aims: BOYDSense is developing a breath-based monitoring system for chronic conditions starting with glucose monitoring for people with type 2 diabetes (T2D). This is a post-hoc analysis of data from a proof-of-concept study evaluating the correlation of Volatile Organic Compounds (VOCs) in exhaled breath with glucose values. The aim of this work was to further improve the initial glucose estimation algorithm developed during the study.

Methods: In this 2-phase study we enrolled 130 people with T2D, all subjected to a standardized meal challenge with predefined measurement timepoints at which 3 breath analysis and venous reference values were collected. In phase-1 (n=100) data were collected for algorithm development. Previously phase-1 data was split into 3 groups (training, verification, validation). This time 100% of phase-1 data were used to train the algorithm. Phase-2 data (n=30) stayed untouched and were again used for validation.

Results: The validation dataset included 612 data pairs, with 204 pairs for each breath. For the newly estimated average breath-based glucose values the Consensus Error Grid Analysis (C-EGA) for T2D shows 86.3% in zone A and 99.5% in zone A+B with a stronger correlation for higher glucose values. Additionally, an improved MARD of 19.5% and a higher precision of 48.5% can be reported.

Conclusions: By expanding the training data set the performance of BOYDSense breath analyzer demonstrated an improved correlation for high glucose values in the C-EGA, an improvement of MARD and an increase in precision. Additional studies, providing more than 10,000 new data-pairs for further algorithm improvements, are underway.

FEASIBILITY OF A SEMICONDUCTOR-BASED CONTINUOUS MULTI-ANALYTE DIABETES MONITORING SYSTEM

Muhammad Mujeeb-U-Rahman

Integrated Medical Sensors Inc., Irvine, United States of America

Background and Aims: Comprehensive diabetes management through monitoring multiple analytes has garnered significant interest recently. Our IMS CGM+ system integrates advanced semiconductor and nano technologies to create a cutting-edge electrochemical sensing platform. This minimally invasive, multi-sensor platform can simultaneously measure multiple analytes using on-chip sensors. It is designed for easy user insertion and removal with a custom inserter. The in vivo sensing component comprises a semiconductor device containing on-chip electrochemical sensors, power management, data acquisition and processing, and telemetry. This semiconductor sensor uses a proprietary communication scheme to transmit data to an external, skin-worn transmitter. The transmitter processes and relays the information to a smartphone via BLE. For this work, we intended to develop the different components of the system and perform feasibility studies. This will be the first continuous monitoring system to measure glucose, ketones, and temperature in the subcutaneous (interstitial fluid) space.

Methods: We designed a dedicated application-soecific-intergated-circuit (ASIC) to transduce glucose and ketone signals using two different on-chip sensors consisting of working, counter, and reference electrodes. The ASIC also contained a temperature sensor. The respective glucose and ketone-sensing electrodes were coated with glucose oxidase and 3-hydroxybutyrate dehydrogenase enzymes and tested in spiked saline.

Results: The results showed that the CGM+ can sense glucose and ketones using a single device. The glucose sensor showed a MARD of 10%, while the ketone sensor showed a MAD (mean absolute difference) of < 0.33mM/L. There was negligible cross-talk. Preclinical studies are planned next.

Conclusions: Using a single device, the IMS CGM+ can measure glucose, ketones, and tissue temperature.

EVALUATION OF ACCURACY OF DEXCOM G6 AND G7 CONTINUOUS GLUCOSE MONITORS IN DIALYSIS PATIENTS

Subbulaxmi Trikudanathan ^1,2, Leila Zelnick¹, Yoshio Hall¹, Evelin Jones¹, Lisa Anderson¹, Nate Ashford¹, Ernest Ayer¹, Ian Deboer¹, Irl Hirsch¹

¹University of Washington, Seattle, United States of America, ²University of Washington, seattle, United States of America

Background and Aims: Continuous glucose monitoring (CGM) may be particularly useful for dialysis patients because HbA1c is not reliable. The accuracy of recent versions of CGMs in this population have been sparsely studied. This study aimed to examine the accuracy of Dexcom G6 and G7 compared to glucose meter fingerstick blood glucose (FSBG) in a dialysis population.

Methods: Twelve people with diabetes and kidney failure treated with dialysis simultaneously wore Dexcom G6 pro and G7 CGMs and recorded FSBG for a period of 10 days. We summarized correlations using Pearson correlations, calculated mean absolute relative difference (MARD), and created Clarke’s error grids.

Results: Mean (SD) age of participants was 50 (12) years, 50% were female, mean (SD) diabetes duration was 24 (9) years, and 92% used insulin. Participants collected 242 FSBG over a collective 159 days of CGM. The correlations of G6 and FSBG, G7 and FSBG, and G6 and G7 were 0.87, 0.89 and 0.95, respectively. The MARD of G6 versus FSBG and G7 versus FSBG was 19.7% and 15.6%, respectively. Clarke’s error grids showed that 98% of G6 and G7 measurements in Zone A and B compared with FSBG.

Conclusions: We found good correlations of Dexcom G6 pro and G7 with FSBG in dialysis patients. Relatively high MARD (compared with non-dialysis populations) may be due in part to use of FSBG as reference, and consensus error grids showed majority of the readings in clinically acceptable zones. This data supports the use of CGM to evaluate glycemia in the dialysis population.

EFFECT OF CALIBRATION ON DEXCOMG7 SENSOR PERFORMANCE IN THE CONTEXT OF REPETITIVE SCUBA DIVING IN SUBJECTS WITH TYPE 1 DIABETES MELLITUS

Elena Gamarra ¹, Giovanni Careddu², Andrea Fazi³, Valentina Turra⁴, Ambra Morelli⁵, Pierpaolo Trimboli¹

¹Ente Ospedaliero Cantonale, Endocrinology And Diabetology, Lugano, Switzerland, ²ASL 3 Genovese, Endocrinologia Diabetologia E Malattie Del Ricambio, Genova, Italy, ³Agenzia Regionale Sanitaria, Ancona, Italy, ⁴ASST Spedali Civili, Brescia, Italy, ⁵Clinica San Carlo, Paderno Dugnano, Italy

Background and Aims: The use of continuous glucose monitoring (CGM) sensors during physical activity is challenged by issues with sensor adhesion and data accuracy. In scuba diving, an activity now accessible to individuals with type 1 diabetes (T1DM) following specific protocols, previous research has highlighted the potential of retrospective CGMs, especially in preventing hypoglycemia. Real-time CGM use underwater remains instead problematic, due to data transmission issues and sensor accuracy degradation, as seen with FreeStyle Libre 3 and Dexcom G7. This study aimed to assess the impact of regular calibration of the Dexcom G7 sensor on its performance during repeated dives in T1DM patients.

Methods: 8 T1DM participants wore two Dexcom G7 sensors (on the arm and on the abdomen), over 4 days during a diving cruise. They calibrated the sensors daily and monitored blood glucose using the Beurer GL50 glucometer, following the Diabete Sommerso protocol.

Results: Subjects, average age 41 years, diabetes duration 21 years, mean HbA1c 6.8%, were on closed-loop (6), pump (1), or pens (1). Data were collected from 68 dives, with average duration 50 min and a maximum depth 24.8 meters. The analysis of 383 sensor-capillary glucose data pairs showed a MARD of 11% for arm and 16% for abdomen sensors, significantly better than uncalibrated sensors in previous study (31% MARD). Sensor arrows matched capillary readings in 73% of cases for arm and 71% for abdomen sensors.

Conclusions: Regular calibration improves the Dexcom G7 accuracy during diving, though capillary glucose checks remain essential for safe dive planning in T1DM patients.

THE ROLE OF DIETARY ADVANCED GLYCATION END-PRODUCTS IN INSULIN SENSITIVITY IN TYPE 1 DIABETES AND: A 12-WEEK RANDOMIZED CLINICAL TRIAL

Hana Kahleova

Physicians Committee for Responsible Medicine, Clinical Research, Washington, United States of America

Background and Aims: This study compared the effect of a low-fat vegan diet with a portion-controlled diet on insulin sensitivity in people with type 1 diabetes and explored the role of dietary advanced glycation end-products (AGEs).

Methods: Fifty-eight adults with type 1 diabetes were randomly assigned to a vegan (n=29), or a portion-controlled group (n=29) for 12 weeks. Insulin sensitivity was assessed, using the carbohydrate to insulin ratio. The content of all meals was recorded using the Cronometer mobile application. Total insulin dose was calculated as a sum of basal and bolus insulin units injected per day. The participants used Dexcom G6 for continuous glucose monitoring.

Results: Insulin sensitivity increased in the vegan group by 6.6 (p=0.002), compared with no significant change in the portion-controlled group (-1.6; p=0.26; treatment effect +8.2 [95% CI, +3.6 to +12.8]; p=0.001). Total dietary AGEs decreased in the vegan group by 7,123 ku/day (95% CI, -9,493 to -4,756; p<0.001), compared with no significant change in the portion-controlled group (treatment effect -5,076 ku/day [95% CI, -8,605 to -1,548]; p=0.006). Change in total dietary AGEs correlated negatively with changes in insulin sensitivity (r=-0.48; p=0.004) and remained significant after adjustment for changes in BMI (r=-0.35; p=0.048). A 1-unit increase in insulin sensitivity was associated with a reduction in dietary AGEs by 2,090 ku/day.

Conclusions: This study suggests that a low-fat vegan diet improves insulin sensitivity in people with type 1 diabetes compared with a portion-controlled diet, and the reduction in dietary AGEs seems to be one of the mediating mechanisms.

TIME BELOW RANGE METRICS INCREASES PRIOR TO SEVERE HYPOGLYCEMIA EVENTS

Kagan Karakus ¹, Emma Mason², Halis Akturk¹

¹Barbara Davis Center For Diabetes, Aurora, United States of America, ²Barbara Davis Center for Diabetes Adult Clinic, Aurora, United States of America

Background and Aims: Severe hypoglycemia (SH) remains an important problem for people with type 1 diabetes. We investigated if CGM metrics have changes prior to SH events.

Methods: In this retrospective study, adults with T1D who reported SH events that required emergency visits were recruited in Barbara Davis Center for Diabetes. Their Dexcom CGM reports were downloaded 7-days, 14-days, and 30-days prior to SH event. As control, each individual’s 30-days SH event-free periods were included. CGM metrics prior to SH event were compared with SH-free period by paired t-test or Wilcoxon signed rank test based on normality. Continuous variables are reported as Mean±SD or Median[IQR].

Results: There were 32 patients with 41 SH events with complete CGM data at 7-days, 14-days, and 30-days prior to SH event and 30-days SH-free periods. TBR (<70 mg/dL) was higher prior to SH event compared to SH-free period (SH-free: 2.1%[0.9-4.9] vs. 7-days SH: 3.8%[1.7-9.4], p<0.001; 14-days SH: 3.7%[1.1-9.1], p=0.002; 30-days SH:3.0%[1.1-8.0], p=0.014). TBR2 (<54 mg/dL) also increased prior to SH event (SH-free: 0.4% [0.1-1.1] vs. 7-days SH:0.8%[0.1-2.5], p<0.001; 14-days SH:0.7%[0.1-2.3], p=0.008; 30-days SH:0.5%[0.2-2.1], p=0.041). While SD did not differ between SH and SH-free periods, CV showed significant increase in 7-days prior to SH event compared to SH-free period (38.5±8 vs. 36.7±6, p=0.043). Other metrics were similar between SH and SH-free periods (Table 1).

Conclusions: TBR(<70mg/dL) and TBR2(<54mg/dL) showed substantial increase 7-days prior to severe hypoglycemia events. Sudden changes in TBR metrics should be evaluated cautiously for people with T1D.

REAL-WORLD DEXCOM G7 CONTINUOUS GLUCOSE MONITOR (CGM) WEAR TIME IN YOUTH WITH TYPE 1 DIABETES (T1D)

Brynn Marks ^1,2, Seema Meighan¹, Jacquelyn Hatch-Stein^1,2, Melissa Rearson¹, Andrew Rearson¹

¹Children's Hospital of Philadelphia, Division Of Endocrinology & Diabetes, Philadelphia, United States of America, ²University of Pennsylvania School of Medicine, Philadelphia, United States of America

Background and Aims: The Dexcom G7 CGM is intended to be worn for 10 days. We studied real-world Dexcom G7 wear time in youth with T1D.

Methods: Dexcom Clarity data from 5/13/24 to 8/13/24 were collected for youth with T1D using Dexcom G7 who were sharing data with the Children’s Hospital of Philadelphia. Median sensor wear time and CGM metrics were calculated for youth using ≥3 sensors. Kruskal-Wallis tests were used to assess for differences in wear time according to demographic characteristics. A multivariable model assessed the relationship between sensor wear time and time in range (TIR).

Results: 606 unique individuals (15.0 yrs, 43.7% female, 65.0% NHW, 12.0% NHB, 59.9% privately insured, 4.7 yr T1D duration, HbA1c 7.4%) wore 5,079 sensors in the 93-day window. Individual median sensor wear time was 8.56 days (7.30, 9.51). Wear time was <7 days for 24.9% of sensors and 39.9% of sensors were worn for ≥10 days. NHB youth had a median sensor wear time of 19.2 hours less than NHW youth (8.34 vs 9.14 days, p=0.0005) and used 2.5 fewer sensors (7.5 vs 10, p=0.0001). Each 1 day increase in median sensor wear time improved CGM time in range by 1.82%-points (β=1.82, 95% CI 0.92-2.71, p<0.001).

Conclusions: Real-world Dexcom G7 wear time is < 10 days for most youth with T1D. Whereas people with diabetes typically receive 36 sensors per year, with a median wear time of 8.56 days youth require 43 sensors to allow for continuous CGM use.

COMPARISON OF GLYCEMIC MARKERS BETWEEN WEEKDAYS AND WEEKEND DAYS IN PATIENTS WITH TYPE 1 DIABETES USING CONTINUOUS GLUCOSE MONITORING

Eleftheria Papachristoforou ¹, Ioannis Daldas¹, Dimitrios Gotzias², Aikaterini Kountouri², Konstantinos Markakis², Vaia Lambadiari², Stavros Liatis¹

¹National and Kapodistrian University of Athens Medical School, First Department Of Propaedeutic And Internal Medicine, Athens, Greece, ²Attikon University Hospital, Medical School, National and Kapodistrian University of Athens, Second Department Of Internal Medicine, Athens, Greece

Background and Aims: Type 1 diabetes (T1D) management has been revolutionized with the routine utilization of continuous glucose monitoring (CGM). The aim of this study was to compare CGM metrics between weekdays and weekend days in patients with TID.

Methods: The population of this study consisted of adult patients with T1D from two academic centers in Greece. The most recent 90 days of CGM data were uploaded for each patient, and the following CGM metrics were calculated separately for weekdays and weekends: Time in Range (TIR_70-180,), Time Above Range (TAR_>180,), Time Below Range (TBR_<70,) and Glucose Management Indicator (GMI).

Results: Ninety-five patients with T1D were included [mean age (± SD) 43 ± 15.1 years, 51.6% males]. Median (IQR) time of CGM use was 93% (82 to 98). TIR_70-180 was significantly lower during the weekend as compared to the weekdays (56.2% ± 18.1 vs 57.6% ± 17.2, p=0.015). Furthermore, TBR_<70 increased significantly during the weekend [4.5% (1.9 to 7.9) vs 4.1% (2.1 to 8.9), p=0.014]. Insulin pump use was associated with lower TIR_70-180 during the weekend (52.6% ± 13 vs 55.6 ± 13.3, p=0.006) and men spent more time in hyperglycemic range during the weekend as compared to the weekdays (38.2% ± 19.2 vs 36.3% ± 19.6, p=0.028).

Conclusions: In this study, we observed a slight deterioration in glycemic markers among patients with T1D during weekend days. However, the differences appear to be of minor clinical significance. Larger studies are warranted to draw definitive conclusions.

THE EFFECT OF GLYCEMIC VARIABILITY, NOCTURNAL HYPERGLYCEMIA AND HYPOGLYCEMIA ON SLEEP ARCHITECTURE AMONG CHILDREN WITH TYPE 1 DIABETES MELLITUS

Nouran Salah ^1,2, Heba Rashed², Amal Abido¹

¹Ain Shams University, Pediatrics, Cairo, Egypt, ²Ain Shams University, Cairo, Egypt

Background and Aims: The impact of glycemic variability, nocturnal hyperglycemia and hypoglycemia on sleep architecture remains unraveled.

Aim: To assess the effect of nocturnal hyperglycemia, hypoglycemia and glycemic variability using continuous glucose monitoring system on sleep latency, continuity and architecture among children with T1DM.

Methods: Thirty children with T1DM were compared to 30 age and gender matched healthy siblings. Diabetes duration, type and dose of insulin therapy, symptoms of sleep disorders (e.g. obstructive sleep apnea, restless leg syndrome) were assessed. Anthropometric measures with calculation of z-score and manifestations of insulin were obtained. Continuous-glucose monitoring system(CGMS) was done using Medtronic iPro2 Recorder and overnight polysomnography(PSG) was done using Nihon Kohden Neurofax-1200 EEG/PSG system(Tokyo, Japan) with analysis of sleep latency, continuity measures, sleep architecture, apnea-hypopnea index(AHI), and periodic-limb movement(PLM)index.

Results: Children with T1DM had significantly lower sleep efficiency(P<0.001) with significantly higher arousal-index(AI)(P<0.001), periodic-limb movement-index(P=0.004) and apnea-hypopnea index(P=0.016) and significantly longer sleep-onset latency(P<0.001) and light sleep%(P<0.001) than controls. Sleep efficiency was significantly lower in those who had nocturnal hyperglycemia than those without (P<0.001), while arousal index was significantly lower(P=0.041) in those who had nocturnal hypoglycemia than those without. Nocturnal profound-hypoglycemia was positively correlated to deep sleep%(P=0.001) and stage3-sleep(P<0.001), while hyperglycemia was positively correlated to number of awakenings(P=0.02), sleep-onset latency(P=0.029) and stage2-sleep(P=0.006), and negatively correlated to sleep efficiency(P=0.001) and REM-sleep(P<0.001).

Conclusions: Profound hypoglycemia is associated with sleep deepening, while hyperglycemia is associated with poor sleep efficiency with increased awakenings and decreased REM-sleep among children with T1DM. Tight glycemic control might improve sleep quality and duration among children with T1DM.

COST-EFFECTIVENESS OF REAL-TIME CGM (RT-CGM) IN INDIVIDUALS WITH NON-INSULIN-TREATED TYPE 2 DIABETES (T2D) IN CANADA

Hamza Alshannaq ¹, Gregory Norman¹, Andreas Nilsson², Michael Willis²

¹Dexcom, Inc., Global Access, Value And Evidence, San Diego, United States of America, ²The Swedish Institute for Health Economics, Lund, Sweden

Background and Aims: Evidence shows RT-CGM improves glycemic control in individuals with T2D on non-insulin therapies (T2 NIT). We conducted a cost-effectiveness analysis comparing RT-CGM to self-monitoring of blood glucose (SMBG) from Canadian public and commercial payer perspectives.

Methods: A 40-year analysis was performed using the validated ECHO-T2DM. Patient characteristics were drawn from Aronson (PMID:36546594), while clinical effectiveness data were sourced from Cox (PMID:33328160), where HbA1c reductions were -1.1% for RT-CGM and -0.3% for SMBG, with no difference in severe hypoglycemia (SH) or hyperglycemia (DKA) rates between arms. Escalation to basal insulin and subsequently to bolus insulin was modeled when HbA1c exceeded 8.5%. After insulin initiation, SH and DKA rates were modeled using a large retrospective study (PMID:34077502). SMBG patients had an SH rate of 0.04, a DKA rate of 0.025 per year, and 0 for RT-CGM. Indirect costs from absenteeism were included only in the commercial payor analysis. Costs and outcomes were discounted at 1.5% annually. Sensitivity analyses were performed.

Results: RT-CGM was associated with quality-adjusted life year (QALY) gains of 0.763 and incremental costs of CAD 30,771, yielding an incremental cost-effectiveness ratio (ICER) of CAD 40,349 per QALY gained. For commercial payors, including indirect costs reduced incremental costs to CAD 843 and the ICER to CAD 1,105. Sensitivity analyses of 15% RT-CGM cost reduction made RT-CGM cost-saving for commercial payors and reduced the ICER to CAD 32,093 for public payors.

Conclusions: RT-CGM is cost-effective for public payors and cost-saving for commercial payors for the T2 NIT population in Canada.

DEXCOM G7 IN THE FIRST 30 DAYS: EASE OF USE AND GLYCEMIC IMPROVEMENTS IN PEOPLE WITH T2D

Margaret Crawford, Christian Hicks, Yukina Benitz, Vasu Rao, Harsimran Singh

Dexcom, Inc., San Diego, United States of America

Background and Aims: We aimed to explore T2D users’ experiences during their first 30 days using Dexcom G7, focusing on satisfaction, glycemic changes, feature utilization, and key factors differentiating users’ experiences.

Methods: US-based T2D Dexcom G7 users completed an emailed survey assessing lived experience and satisfaction with their CGM. Glucose and feature usage data were analyzed, requiring 70% available CGM data during first and last 10-day periods of the initial 30-day period.

Results: G7 users (N=803) were 41% female, 45% aged >65 years, 30% CGM-naïve, 68% using insulin. Overall, cohort improved TIR and TITR (time-in-range 70-180mg/dL, time-in-tight-range 70-140mg/dL) over observation period (median [IQR] 78%[58, 90] to 80%[59, 91] and 42%[23, 63] to 46%[25, 65], respectively (both p-values<0.05)). Additionally, 34% logged an event in the G7 application, 21% used Dexcom Follow, and most reported satisfaction with the ease of use (86%) and onboarding (82%). CGM-naïve users, younger than 65, and non-insulin users experienced greater glycemic improvements. Non-insulin users had significantly greater TIR than their counterparts (87% vs. 74% over observation period, p<0.05) and CGM-naïve users logged significantly more events than CGM-experienced users (64% vs. 21%, p<0.001). Satisfaction levels were high across all users.

Conclusions: These real-world data underscore Dexcom G7 CGM’s ease of use, high satisfaction with onboarding, and glycemic improvements among the T2D population. Given that many users self-initiated CGM, observed differences in glycemic improvements and in-app engagement suggest that some users could further optimize patient outcomes during G7 initiation through targeted support.

REAL-WORLD DATA ON USE OF CONTINUOUS GLUCOSE MONITORING (CGM) IN A MULTI-ETHNIC TYPE 2 DIABETES (T2D) POPULATION

Abdul Wahab Paracha ¹, Priscilla Sarkar², Pratik Choudhary^1,2, Ehtasham Ahmad¹

¹Leicester General Hospital, Leicester Diabetes Department, Leicester, United Kingdom, ²University of Leicester, Diabetes Research Centre, Leicester, United Kingdom

Background and Aims: Continuous glucose monitoring (CGM) is now the standard of care for the management of people with type 1 diabetes, with emerging randomised trial and real-world data on benefits in people with type 2 diabetes (T2D). We explored the impact of CGM on glycaemic outcomes in multi-ethnic populations with T2D.

Methods: We identified people with T2D using CGM for at least 6 months attending University Hospitals of Leicester (UHL). We collected demographic and glucose-lowering therapy data from electronic health records and compared glycaemic outcomes before and at least 6 months after the start of the CGM in different ethnicities.

Results: We identified 914 people with T2D using CGM. The mean age of the cohort was 54 years and 57% were male from a diverse ethnic background; Caucasian 44%, South Asian 32%, Black 5% and others including mixed ethnicities 19%. 69% were using basal-bolus insulin, 20% were on basal-only insulin, 9% were using premixed insulin twice daily, and 2% used either bolus insulin or only oral therapies. The mean duration of CGM use was 317 days. CGM use was associated with a statistically significant reduction in HbA1c from 9.3% (SD2.06) at baseline to 8.7% (SD1.89) at follow-up with a mean difference of -0.52% (95%CI -0.77,-0.36) and no differences between ethnicities (p=0.7).

Conclusions: The use of CGM was associated with a statistically significant improvement in HbA1c in people with T2D from a diverse multi-ethnic background. Despite this many patients do not achieve glucose targets and further input around education may be needed.

CONTINUOUS GLUCOSE MONITORING METRICS PREDICT ALL-CAUSE MORTALITY IN PATIENTS WITH BOTH TYPE 1 AND TYPE 2 DIABETES

Tomoki Okuno¹, Sharon Macwan², Gregory Norman³, Donald Miller⁴, Peter Reaven², Jin Zhou ¹

¹UCLA, Los Angeles, United States of America, ²Carl T. Hayden Phoenix VA Health Care System (111E), Phoenix, United States of America, ³Dexcom, Inc., Global Access, Value And Evidence, San Diego, United States of America, ⁴University of Massachusetts, Lowell, Lowell, United States of America

Background and Aims: To investigate the association between CGM-derived glucose metrics and all-cause mortality in patients with type 1 and type 2 diabetes.

Methods: We analyzed 2,751 adults (≥21 years) with diabetes (70% type 2) from the Veterans Affairs Healthcare System who initiated CGM between 2015 and 2020. Dexcom CGM data were merged with electronic health records; all participants had at least 14 days of readings. All-cause mortality was assessed over five years from CGM initiation. Separate Cox proportional hazards models evaluated associations between mortality and CGM metrics—estimated glucose value (eBG), time-in-range (TIR), time-above-range (TAR), coefficient of variation (CV), and glycemic risk index (GRI)—over initial landmark periods of 14-days, 3-months, and 6-months.

Results: The mean age at CGM initiation was 64 years old, and median CGM use was 3.0 (type 1) and 2.8 years (type 2), and there were 406 deaths. Multivariate Cox models (adjusting for variables related to mortality) showed that higher eBG, TIR, TAR, CV and GRI during the 6-month period were significantly associated with mortality (hazard ratios: eBG 1.18, TAR 1.20, GRI 1.23; CV 1.18 and TIR 0.83, all p ≤ 0.002). These associations remained significant after adjusting for landmark HbA1c. Shorter landmark CGM observation windows (14-days or 3-months) showed similar but slightly smaller effects. The association between CV with mortality was independent of other CGM metrics and appeared strongest in those with lower HbA1c levels.

Conclusions: CGM-derived metrics were significantly associated with all-cause mortality in patients with diabetes even after accounting for HbA1c.

GLYCEMIC AND FOOD QUALITY IMPROVEMENT IN PEOPLE LIVING WITH DIABETES THROUGH A SMARTPHONE APPLICATION THAT COMBINES CGM AND AI-POWERED DIETARY ANALYTICS

Jeffrey Alfonsi¹, Sarah Blunden ²

¹University of Toronto, Toronto, Canada, ²Dexcom, Montreal, Canada

Background and Aims: People living with diabetes (PWD) face daily challenges of managing diabetes including but not limited to food choices and monitoring glucose, which can potentially impact health outcomes. The use of feedback technology solutions can provide insights to help in overall diabetes management with the end goal of improving outcomes and patient experience. Real-time continuous glucose monitors (RT-CGM) can provide feedback on glucose control and trends. RxFood, an AI powered personalized nutrition application, allows patients to take pictures of their food and receive feedback on carbohydrate content and diet quality. Aim: to assess how combining feedback from RxFood and RT-CGM in a single application impacts glucose metrics and food quality in a real-world setting.

Methods: Data were analyzed from Dexcom G6/G7 and RxFood. Baseline data was collected using CGM and 3-day photo food logs in RxFood. Outcomes were change in CGM metrics and in food quality metrics from baseline to 3 months.

Results: CGM and RxFood users (N=246) stratified by first sensor use with baseline Time In Range (TIR) <70% illustrated a 4.8% increase in TIR attributed to an overall 4.6% decrease in Time Above Range (TAR) or hyperglycemia. Further stratification by first sensor use with Glucose Management Indicator >8.0% (N=97) demonstrated a 10.5% increase in TIR again attributed to overall decrease in TAR. Overall users increased vegetable consumption by 8.0%, whole grains by 6.2% and fiber by 3.0%.

Conclusions: Real-world evidence of PWD with suboptimal managed diabetes experienced improved TIR and food quality when 2 feedback response technologies were used in combination.

IMPACT OF NIGHTTIME HYPOGLYCEMIA ON GLYCEMIC CONTROL IN TYPE 1 AND TYPE 2 DIABETES: A REAL-WORLD DATA ANALYSIS

Michael Steinhauser-Mitter¹, Josip Zivkovic¹, Delphine Theodorou², Timor Glatzer ³

¹mySugr GmbH, Vienna, Austria, ²Roche Diagnostics International, Basel, Switzerland, ³Roche Diabetes Care GmbH, Mannheim, Germany

Background and Aims: Nighttime hypoglyceamia is a common side effect of insulin therapy with negative somatic, psychological, and social effects. Although CGM use has reduced nighttime hypoglycemia, it remains a common burden. The impact on glycemic control during the following day for both type 1 and type 2 diabetes in real-world data remains under-characterized.

Methods: Data from the first month of CGM-use by 3696 type 1 and 441 type 2 mySugr app users on MDI was analyzed. Hypoglycaemic events were defined as CGM readings below 70 mg/dl or 54 mg/dl for at least 15 minutes. The analysis included time below 54 mg/dl (TB54), time below range (TBR), time in range (TIR), time above range (TAB), glycemic variability (CV), and hypoglycaemic event frequency for day and night periods. The impact of nocturnal events on next-day glucose levels was also examined.

Results: Type 1 users had a mean of 3.6 diurnal and 1 nocturnal hypoglycemic event per week, while type 2 users had 1 and 0.4 events, respectively. TB54 events differed significantly between type 1 (1) and type 2 (0.3), but not between day and night periods within types. Nocturnal episodes lasted about 15 minutes longer than diurnal ones. Both types experienced larger daily glycemic variability and reduced daily means following nocturnal episodes.

Conclusions: Nighttime hypoglycaemia occurs regularly in people with type 1 and type 2 diabetes on insulin therapy. It differs significantly from daytime hypoglycaemia, lasting longer and affecting next-day glucose levels. The opposite of the Somogyi effect was observed.

A CARBOHYDRATE COUNTING APPLICATION DESIGNED FOR AND BY YOUTH WITH TYPE 1 DIABETES

Asmaa Housni ¹, Giuliana Giannini¹, Alexandra Katz², Amélie Roy-Fleming¹, Anne-Sophie Brazeau³

¹McGill University, School Of Human Nutrition, Montréal, Canada, ²University of Montreal, Faculty Of Medicine, Montreal, Canada, ³McGill University, Sainte-Anne-de-Bellevue, Canada

Background and Aims: Intensive insulin therapy using carbohydrate counting (CC) is recommended for type 1 diabetes (T1D) management, but CC can be challenging. A mobile application (app) with photo recognition could facilitate CC, but none currently meet youths’ needs. The aim of this study is to promote a collaborative design, ultimately enhancing support for diabetes management.

Methods: Usability testing through three rounds of interviews, each with 3 to 5 participants, were conducted using a think-aloud methodology. Feedback from youth with T1D was integrated to improve user engagement and satisfaction with the app.

Results: 13 people (median age 23 years [19–24]; 69% women; diabetes duration of 8 years [7–10]) participated. Thematic analysis revealed 7 key insights; 1) Features emphasizing a person-centered design and distinguishing information for healthcare professionals and T1D users helps separate their identity from their diabetes; 2) Intuitive navigation with clear directives to ensure user engagement; 3) Presenting relevant information in familiar language to enhance user understanding; 4) Customizable settings according to treatment method (e.g. insulin pump vs. multiple daily injections) to tailor user experience; 5) Data verification and override mechanisms to avoid human and technological errors; 6) Features including downloadable reports, tracking hypoglycemia episodes, and food journals support bidirectional communication between patients and practitioners, minimizing the burden of manual tracking or using multiple apps; 7) Features that link lifestyle factors, and injection time to blood glucose levels can provide opportunities for personalized learning, promoting self-efficacy.

Conclusions: This novel CC app designed around these features can achieve high usability, enhancing user engagement and satisfaction.

THE EFFECT OF SOCIOECONOMIC DEPRIVATION ON ACCESS TO HYBRID CLOSED LOOP INSULIN PUMPS AND TIME IN RANGE OUTCOMES IN PATIENTS WITH TYPE 1 DIABETES

Scott Williams^1,2, Arun George ², Hamza Khan², Muneer Ahmad², Haika Shoo², Ben Lim², Saad Shah², Catriona James², Jane Bridges², Carol Duffy¹, Sejal Pandya², Haris Ahmed², Omer Elhassan², Frank Joseph²

¹Countess of Chester Hospital, Diabetes And Endocrinology, Chester, United Kingdom, ²Countess of Chester Hospital, Endocrinology And Diabetes, Chester, United Kingdom

Background and Aims: The use of Diabetes technology is lowest in the most deprived social backgrounds, who are known to be at increased risk of diabetes complications and worse outcomes. We aim to: 1. Measure access to insulin pumps including Hybrid Closed Loop (HCL) within our insulin pump service. 2. Evaluate trends in Time in Range (TIR) and HbA1c pre and post commencement of HCL across all indices of multiple deprivation (IMD).

Methods: We recorded the percentage TIR, and HbA1c pre and post commencing HCL in a retrospective observational study. IMD recorded from UK Government data are a measure of relative Socioeconomic deprivation divided into deciles, from deciles 1 and 2 (most deprived) to deciles 9 and 10 (least deprived).

Results: Of the 256 patients (Age range 17 to 80 years; 152 females, 104 males), 31% were from the least deprived group, while 14% were from the most deprived group. In the least deprived group, the mean TIR increased from 56% (95% Confidence Interval (CI) 24 to 88) pre-HCL to 70% (95% CI 44 to 96) post-HCL, with a reduction in mean HbA1c from 62 mmol/mol to 53 mmol/mol. In the most deprived group, the mean TIR increased from 51% (95% CI 17 to 85) to a mean figure of 64% (95% CI 22 to 100), with a reduction in mean HbA1c from 64 to 53 mmol/mol.

Conclusions: Efforts should be focused to improve the access to and effective use of HCL insulin pumps in areas of socioeconomic deprivation where fewer patients have access to insulin pumps.

USING AN AUTOMATIC INSULIN DELIVERY SYSTEM SINCE DIABETES ONSET WARRANTS THE BEST GLYCEMIC CONTROL IN CHILDREN AND ADOLESCENTS: ONE-YEAR FOLLOW-UP

Ivana Rabbone ¹, Andrea Scaramuzza², Silvia Savastio¹, Valeria Castorani¹, Erica Pozzi¹, Claudio Cavalli²

¹University of Piemonte Orientale, Department Of Health Sciences, Novara, Italy, ²ASST Cremona, Pediatrics, Cremona, Italy

Background and Aims: The impact of automatic insulin delivery (AID) systems since diabetes onset is unclear. We determined whether using AID systems within 1-month of diabetes onset is more effective than using predictive low glucose suspend (PLGS) insulin pump therapy or multiple daily injections (MDI), during 1-year follow-up.

Methods: This is an observational cohort study, enrolling patients aged 0-18 years with diabetes onset between January 1st, 2020, and June 30th, 2023, subsequently divided into three groups according to the use of MDI (n=24), PLGS (n=28), and AID systems (n=32). After the first 3 months, participants were free to stay or change the insulin delivery system group. Groups have been analyzed considering the first 3-month choice. Main outcome was HbA1c after 12 months; HbA1c, glucometrics and insulin requirement were also reported for each participant after 3, 6 and 12 months.

Results: Eighty-four children and adolescents, mean age 7.9±3.9 years, 58 males, BMI-Z-score -0.3±1.6, have been enrolled. After 12-months, HbA1c was significantly lower in AID group than in PLGS or MDI groups (AID 6.6±0.6% vs PLGS 7.4±1.1% vs MDI 7.6±1.5%, p=0.001). AID systems showed better (p=0.001 for all) TIR, TBR, TAR, CV, GMI, and insulin requirement than PLGS and MDI groups, with no correlation with ethnicity, age, BMI-Z-score, and DKA at onset.

Conclusions: Using AID systems since diabetes onset shows better outcomes than PLGS and MDI groups regarding HbA1c and glucometrics. To our knowledge, this is the first study to demonstrate that therapy with AID systems since diabetes onset has to be the system of choice for all children.

ATTITUDES AND PRACTICES OF YOUTH AND YOUNG ADULTS WITH TYPE 1 DIABETES TOWARDS SELF-ADJUSTMENTS IN INSULIN PUMP SETTINGS

Keren Smuel-Zilberberg ¹, Michal Yackobovitch-Gavan^1,2, Sari Krepel Volsky¹, Moshe Phillip^1,2, Revital Nimri^1,2

¹Schneider Children's Medical Center of Israel, Jesse Z And Sara Lea Shafer Institute For Endocrinology And Diabetes, National Center For Childhood Diabetes, Petah Tikva, Israel, ²Tel Aviv University, Faculty Of Medical And Health Sciences, Tel Aviv, Israel

Background and Aims: Optimizing glycemic control in individuals with T1D on intensive insulin therapy requires frequent adjustments. Since clinic visits occur quarterly, self-management can help improve it. However, the importance of self-management and its extent among the pediatric population with T1D is sparsely investigated.

Objective: To explore the frequency and practices of insulin self-adjustments made by pediatric pump users and their effect on glucose control.

Methods: A questionnaire-based cross-sectional study was conducted at Schneider Children's Medical Center of Israel from January 2019 to January 2023. The questionnaire consists of 15 questions addressing aspects related to individuals' attitudes and self-management of pump settings and their acceptance of automated adjustments.

Results: The questionnaire was completed by 75 pump-treated children, adolescents and young adults, with mean age of 15.5 ± 5.67 years, 44% were males, median HbA1c was 7.5% (IQR 6.9, 8.6), and median insulin dose was 0.76 units/kg (0.6,0.9). We found that 57.3% of responders do not change settings by themselves between visits, and most of those who make changes do so only 1–2 times per year. The most frequently adjusted factor was the basal rate (93.8%). Lower HbA1c was positively correlated with self-adjusting insulin and negatively with adjustment frequency (rs = -0.389, p < 0.001). Among respondents, 69% indicated they would use automated recommendations.

Conclusions: Pump users who adjust insulin between visits have better glycemic control, suggesting that greater emphasis should be placed on educating and supporting users to manage their treatment independently. Implementing decision support systems could provide additional reassurance.

SUSTAINED BENEFIT FROM USE OF INHALED INSULIN IN THE INHALE-3 EXTENSION STUDY

Grazia Aleppo ¹, Roy Beck², Ryan Bailey³, Peter Calhoun³, Katrina Ruedy³, Irl Hirsch⁴

¹Northwestern University, Division Of Endocrinology, Metabolism And Molecular Medicine, Chicago, United States of America, ²JAEB Center for Health Research, Research, Tampa, United States of America, ³Jaeb Center for Health Research, Tampa, United States of America, ⁴University of Washington, seattle, United States of America

Background and Aims: In the INHALE-3 randomized trial of adults with type 1 diabetes, inhaled Technosphere Insulin (TI) was shown to be superior to rapid acting analogue (RAA) insulin in reducing post-prandial hyperglycemia, and 17-week HbA1c levels with a regimen of TI plus insulin deglucec were non-inferior compared with a control group using RAA insulin with automated insulin delivery or multiple daily injections. Following the randomized trial, the TI group continued to use TI and degludec for an additional 13 weeks to assess sustainability of effect without intervening clinic visits.

Methods: Participants using TI and insulin degludec for 17 weeks in the INHALE-3 randomized trial continued this regimen for an additional 13 weeks without any scheduled visits prior to 30 weeks.

Results: A total of 45 participants (mean age 47yrs) initiated the extension phase; 43 completed the 30-week follow-up visit. Mean HbA1c, which was 7.6% at both randomization and 17 weeks, decreased to 7.4% at 30 weeks (mean change from 17 to 30 weeks = -0.21, 95% CI -0.33 to -0.09, p<0.001). HbA1c was <7.0% in 21% at randomization and 30% at 17 weeks, with an increase to 42% at 30 weeks. CGM metrics did not change substantially from 17 weeks to 30 weeks (Table).

Conclusions: In the INHALE-3 trial, HbA1c improved between 17 and 30 weeks despite there being no scheduled study visits or contacts This durable effect suggests that over time participants learned how to better optimize use of TI dosing for meals and corrections.

DEVELOPMENT OF KAYO-1732, A 1ST-IN-CLASS ORAL ALDH1A3 INHIBITOR, TO PREVENT OR REVERSE TYPE 2 DIABETES AND CARDIOMETABOLIC SYNDROME

Mark Esposito, Merika Koday

Kayothera Inc, Seattle, United States of America

Background and Aims: The retinoid pathway is a central metabolic regulator that is activated when dietary vitamin A is oxidized to all-trans retinoic acid (atRA) and binds the RARα/β/γ nuclear receptors. In humans, pharmacologic activation of the retinoid pathway causes severe hyperlipidemias, hypothyroidism and increased mortality due to cardiovascular disease. During the natural progression of metabolic disorders, including type 2 diabetes, ALDH1a3 is induced to activate the retinoid nuclear pathway through oxidizing all-trans retinaldehyde. ALDH1a3 has been identified as one of the principal markers of pancreatic degeneration and vascular hypertrophy in patients with cardiometabolic disorders, and we have shown that ALDH1a3 causes pathogenic loss of β cell function.

Methods: Here we employed a hybrid drug development approach to identify ALDH1a3 inhibitors and progressed key chemotypes through medicinal chemistry optimization to develop KAYO-1732, a potent, selective and safe ALDH1a3 competitive inhibitor that is orally absorbed with low clearance.

Results: KAYO-1732 treatment in pre-diabetic ZDF rats prevents the development of hyperglycemia and glucose-induced diuresis. In models of severe diabetes, including 20 week-old ZDF rats, 16 week-old db/db mice and STZ-induced diabetes, ALDH1a3 inhibition controls blood glucose, as measured by HbA1c reduction, to non-diabetic levels. ALDH1a3 inhibition also prevents the onset of glomerular hyperfiltration while lowering triglycerides, LDL and markers of liver fibrosis.

Conclusions: The continued development of KAYO-1732 offers a novel approach to preventing and potentially reversing multiple metabolic disorders and may offer the first disease-modifying therapy for type 2 diabetes.

COVALENT-111: EXPLORING ICOVAMENIB IN PERSONS WITH POORLY CONTROLLED SEVERE INSULIN-DEFICIENT (SIDD) TYPE 2 DIABETES

Alexander Abitbol¹, Douglas Denham², Apinya Vutikullird³, Adolfo Cueli⁴, Sanchita Mourya⁵, Brian Munneke⁵, Jeffrey Kim⁵, Thomas Butler⁵, Juan Frias ⁵

¹LMC Diabetes & Endocrinology, Oakville, Canada, ²Flourish Research, San Antonio, United States of America, ³ARK Clinical Research, Long Beach, United States of America, ⁴Entrust Clinical Research, Miami, United States of America, ⁵Biomea Fusion, Redwood City, United States of America

Background and Aims: Type 2 Diabetes (T2D) is a heterogeneous disease, with varying degrees of insulin deficiency and resistance. The Severe Insulin-Deficient Diabetes (SIDD) subgroup, one of five identified by Ahlqvist et al., is characterized by significant insulin deficiency due to beta-cell dysfunction. The COVALENT-111 Phase I/II trial (NCT05731544) investigated icovamenib, an oral covalent menin inhibitor promoting beta-cell proliferation and function, in T2D.

Methods: Participants were randomized to receive placebo or icovamenib (50mg, 100mg, or 200mg) daily for 4 weeks, with 26-week follow-up. Two SIDD subgroup cases are presented.

Results: Case 1: A 45-year-old male with a 10-year T2D history, received 100mg icovamenib daily. HbA1c decreased from 8.6% to 7.5% at Week 26. Time in Range (70-180 mg/dL) improved from 4% to 79%. HOMA-B and C-peptide levels increased by 1233% and 59%, respectively. Case 2: A 29-year-old male with a 4-year T2D history, received 200mg icovamenib daily. HbA1c reduced from 9.5% to 7.0% at Week 26, further declining to 5.8% at Week 47, as reported by his endocrinologist. Time in Range improved from 34% to 90%. HOMA-B and C-peptide increased by 190% and 71%, respectively.

Conclusions: Both cases showed significant improvements in glycemic control and markers of beta-cell function after short-term (4 weeks) icovamenib treatment, with positive benefits persisting beyond the treatment period. These exploratory patient-level data suggest menin-inhibition may be particularly effective in a presumed insulin-deficient subgroup of patients. This finding points to potential benefits of targeted treatment strategies in type 2 diabetes management, warranting further investigation in larger trials.

DNA METHYLATION IN CHILDREN WITH TYPE 1 DIABETES; POTENTIAL ROLE IN DISEASE PATHOGENESIS AND PROGRESSION

Nouran Salah ¹, Lamyaa Salem², Sara Taha¹, Mariam Youssef¹, Layla Annaka¹, Safeya Hassan¹, Rana Mahmoud¹

¹Ain Shams University, Pediatrics, Cairo, Egypt, ²Ain Shams University, cairo, Egypt

Background and Aims: Background: DNA methylation plays a pivotal role in autoimmunity and energy metabolism. However, its mechanistic role and clinical implications in type 1 diabetes (T1D) is still obscure. Aim: to compare DNA methyl-transferase3A (DNMT3A) and ten-eleven translocase2 (TET2) in children with T1D and controls and to correlate them to diabetes duration, HbA1C and micro-vascular complications.

Methods: Methodology: Fifty children with T1D and 50 matched controls were assessed for family history of T1D, diabetes duration, insulin therapy, anthropometric measures, fundus and the bedside neuropathy-disability score. TET2, DNMT3A, fasting lipids, glycated-hemoglobin(HbA1C) and urinary albumin-creatinine were measured.

Results: The mean TET2 of the studied children with T1D was 227.60 ng/ml and their mean DNMT3A was 2.59 ng/ml. TET2 and DNMT3A were significantly higher in children with T1D than controls (P=0.015 and P<0.001, respectively). Children with T1D with diabetes duration ≤5years had significantly higher TET2 than those with diabetes duration >5years (P=0.02) and controls (P<0.001). TET2 was negatively correlated to diabetes duration (P<0.001) and HbA1C (P<0.001), while DNMT3A was positively correlated to diabetes duration (P<0.001) and HbA1C (P=0.001). TET2 was independently related to HbA1C (P=0.001) and diabetes duration (P=0.038), while DNMT3A was independently related to diabetes duration (P=0.003) in children with T1D. Micro-vascular complications were not related to TET2 (P=0.535) and DNMT3A (P=0.310).

Conclusions: TET2 is significantly higher while DNMT3A is significantly lower in children with recent T1D than those with long standing T1D and controls. TET2 and DNMT3A are correlated with diabetes duration and glycemia control in children with T1D.

INSULIN RESISTANCE-ASSOCIATED CHANGES IN THE HUMAN PLASMA LIPIDOME

Péter Pikó ^1,2, Nihad Kharrat Helu¹, Pál László¹, Sándor Szücs¹, Nóra Kovács^1,3, Róza Ádány^1,2,3,4

¹University of Debrecen, Department Of Public Health And Epidemiology, Faculty Of Medicine, Debrecen, Hungary, ²Semmelweis University, Epidemiology And Surveillance Centre, Budapest, Hungary, ³University of Debrecen, Hun-ren Public Health Research Group, Department Of Public Health And Epidemiology, Faculty Of Medicine, Debrecen, Hungary, ⁴Semmelweis University, Department Of Preventive Medicine And Public Health, Budapest, Hungary

Background and Aims: Insulin resistance (IR) is a condition characterised by impaired carbohydrate and lipid metabolism and plays a crucial role in the development and progression of metabolism-related chronic diseases among them type 2 diabetes mellitus (T2DM). The present study aimed to investigate the association between the lipidomic profile and the progression of IR, as assessed by the Homeostatic Model Assessment for IR (HOMA-IR) index.

Methods: The study included 314 controls and 146 IR patients (individuals with a HOMA-IR greater than 3.63) from the Hungarian population. The lipidomic profile (12 lipid classes - 620 lipid species) was determined using the Lipidyzer platform. The p < 8.1 × 10⁻⁴ was considered significant (Bonferroni correction). Lipid indices (LI) were calculated to identify the combination of lipids most strongly correlated with the outcome of the study.

Results: Of the 620 lipid species analysed, 113 showed an individually significant association with insulin, five with glucose, and one (DAG16:01_18:2) with HbA1C levels. LI_insulin (B=11.47, p = 2.3 × 10⁻¹⁷), calculated as TAG56:1_FA18:1+TAG46:4_FA18:2, showed the strongest association with insulin levels, while LI_glucose (B=-0.02, p = 5.9 × 10⁻¹²), calculated as LPC18:0+TAG56:9_FA20:5+TAG54:8_FA18:2, showed the strongest association with glucose levels. Thirty-six lipid species showed a significant association with HOMA-IR, with LI_HOMA calculated as TAG54:2_FA18:0/LPC18:1 showing the strongest association (B=3.24, p = 1.3 × 10⁻¹⁵). Six lipid species showed a significant association with IR, of which TAG50:1_FA16:0 was the strongest (OR=1.004, p = 1.5 × 10⁻⁵).

Conclusions: Our results demonstrate that disturbances in carbohydrate metabolism are associated with alterations in the plasma lipidome, confirming the interplay between IR and dyslipidemia and the involvement of specific lipid species in the progression of IR.

ADVANCING DIABETES MANAGEMENT THROUGH CONTINUOUS GLUCOSE MONITORING SYSTEMS: A REVIEW OF CLINICAL OUTCOMES AND TECHNOLOGICAL INNOVATIONS

S Dariya

NIMS, Jaipur, General Medicine, Jaipur, India

Background and Aims: Recent advancements in artificial pancreas systems (APS) have reshaped the landscape of diabetes management, particularly through closed-loop systems that automate insulin delivery based on real-time data. This paper presents an innovative, AI-driven closed-loop APS, which leverages machine learning (ML) to optimize insulin delivery and improve glycemic control in patients with Type 1 diabetes.

Methods: A clinical trial was conducted involving 200 adults (ages 18-65) with Type 1 diabetes across multiple diabetes care centers. Participants were divided into two groups: Group A: AI-driven closed-loop APS Group B: Conventional sensor-augmented insulin pump therapy Each participant was monitored over a 6-month period, with continuous glucose monitoring (CGM) data collected to assess glucose variability, time-in-range (TIR) metrics, and hypoglycemia rates.

Results: Glycemic Control (Time-In-Range): Group A (AI-Driven APS): 83% TIR Group B (Sensor-Augmented Pump): 61% TIR p-value: <0.001 Hypoglycemia Incidents: Group A: 0.9 episodes/week Group B: 1.5 episodes/week p-value: 0.002 Glucose Variability (Standard Deviation): Group A: 25 mg/dL Group B: 38 mg/dL p-value: <0.001 Patient Satisfaction (Survey): Group A: 93% highly satisfied Group B: 72% highly satisfied.

Conclusions: The AI-driven closed-loop system demonstrated a significant improvement in glycemic control, with an 83% TIR compared to 61% in conventional sensor-augmented pump therapy. Additionally, hypoglycemia episodes were reduced by 40%, while glucose variability decreased by 35% in the AI group. These findings highlight the transformative potential of integrating ML algorithms into APS technology for achieving tighter glycemic control and improving patient outcomes. Further studies are underway to expand the application to pediatric populations and individuals with Type 2 diabetes.

EXPLORING THE HUMAN FACTOR: USER EXPERIENCE AND ADHERENCE TO DIABETES TECHNOLOGY

Ashutosh Chaturvedi ¹, Girish Mathur², S Dariya³, G.D. Ramchandani⁴, Mayank Gupta⁵, Puneet Saxena⁶, Sumeet Garg⁷, Pradyumn Sharma⁸, Neeraj Lodha⁹, Rajneesh Saxena¹⁰, K.K. Pareek¹¹, Prakash Keswani⁶, Arvind Gupta¹²

¹Mahatma Gandhi Medical College and Hospital, Jaipur, General Medicine, Rajasthan, India, ²Senior Consultant Physician, Internal Medicine, Kota, Rajasthan, India, ³NIMS, Jaipur, General Medicine, Jaipur, Rajasthan, India, ⁴Ramchandani Diabetes Centre, Diabetes And Metabolism, Kota, Rajasthan, India, ⁵NIMS, Jaipur, General Medicine, Jaipur, Rajasthan, India, ⁶Sms Medical College and Hospital, General Medicine, Jaipur, Rajasthan, India, ⁷Ram Krishna Medical College Hospital and Research centre, General Medicine, Bhopal, Madhya Pradesh, India, ⁸Gopinath Hospital, Rajasthan, India, ⁹Lodha Hospital and Test tube baby Centre, General Medicine, Pali, Rajasthan, India, ¹⁰Saxena Diabetes Centre, Ajmer, Rajasthan, India, ¹¹SN Pareek Memorial Hospital, General Medicine, Kota, Rajasthan, India, ¹²Rajasthan Hospital, Jaipur, Rajasthan, India

Background and Aims: Background:* Diabetes technology, including continuous glucose monitors (CGMs), insulin pumps, and mobile applications, has transformed diabetes management. Human factors, encompassing usability, accessibility, psychological, and social aspects, significantly impact technology adoption. *Objectives:* Investigate user experiences and adherence to diabetes technology and Identify key human factors influencing technology adoption. Examine relationships between user engagement, device satisfaction, and glycemic control.

Methods: Design:* Mixed-methods study. *Participants:* Adults with diabetes using continuous glucose monitors (CGMs) and/or insulin pumps (n=100). Data Collection: Surveys: Online questionnaire assessing demographics, technology usage, user experience, adherence, and glycemic control. 2. Semi-structured Interviews: In-depth discussions exploring user experiences, motivations, barriers, social support, and emotional responses (n=20). 3. Usability Testing: Evaluation of device interfaces to identify usability issues (n=10). Data Analysis: 1. *Descriptive Statistics: Survey data analysis. 2. *Thematic Analysis: Interview data analysis. 3. *Regression Analysis: Examining relationships between user engagement, device satisfaction, and glycemic control.

Results: Data and Results:* 1. *Survey Results:* - 70% reported improved glycemic control with technology use. - 40% experienced device-related stress/anxiety. - 60% reported difficulties with data interpretation. 2. *Interview Themes:* - Motivation and self-efficacy - Social support and family involvement - Device satisfaction and usability - Emotional responses to technology use 3. *Usability Testing: - Key issues: data interpretation, alarm fatigue, and user interface complexity. 4. *Regression Analysis: - User engagement (β=0.35, p<0.01) and device satisfaction (β=0.28, p<0.05) predicted adherence.

Conclusions: Effective diabetes technology integration requires consideration of human factors. User-centered design, training, and support can enhance adherence and optimize outcomes. Addressing psychological and social aspects is crucial for successful technology adoption.

TYPE OF THERAPY AND ETHNICITY IN TYPE 1 DIABETIC CHILDREN AND ADOLESCENTS: WHAT METABOLIC OUTCOMES?

Silvia Savastio, Erica Pozzi, Valeria Castorani, Eleonora Chiarle, Aurora Craici, Davide Lo Sasso, Chiara Zanzola, Martina Amore, Ivana Rabbone

University of Piemonte Orientale, Department Of Health Sciences, Novara, Italy

Background and Aims: The prevalence of migrant children in clinical practice is constantly increasing. We aim to evaluate metabolic control in type 1 diabetic (T1D) children and adolescents according to age, ethnicity, and type of therapy used.

Methods: This retrospective study enrolled 181 T1D subjects, 119 with native Italian parents (ITA group) and 62 with parents of migrant origin (M group), evaluated according to the type of therapy (MDI, patch pump, PLGS, AID) and the range of age (<6, 6-12, >13 years). Auxological and metabolic variables (BMI Z-score, HbA1c and sensor metrics) were analyzed.

Results: The study population has a mean age of 11±3.6 years, and a suboptimal metabolic control (HbA1c value of 7.3±0.9%, and a TIR of 61.5±13%,). ITA group shows better HbA1C values (7.08±0.85vs7.80±1.07%;p<0.0001) and sensor metrics (TIR 63.97±15.96 vs 56.81±15.23%; TAR250 11±7.1vs17.6±12% and GMI 7.25±0.68vs8.04±3.85%;p<0.01). These differences in glycemic control are significant in all age groups (6-12 and > 13 ys), even if adolescents show the worst results in metrics (p<0.01). No differences in the type of therapy were found between the two groups, not even for AID technology. The best data are guaranteed by an AID technology for both groups, as expected, although with worse results in M as HbA1C (7.71± 0.94vs6.91 ± 0.68%; p<0.0001), and metric values (p<0.0001)( Figure1 ).

Conclusions: Our study showed better metabolic control in patients with Italian parents than migrants, despite equal access to technology. A multidisciplinary approach and psychological support are crucial to improve metabolic outcomes even with AID technology.

UNDERSTANDING HOW PATIENT TRUST IN AUTOMATED INSULIN DELIVERY SYSTEMS RELATES TO IMPROVED DIABETES MANAGEMENT AMONG PEOPLE WITH TYPE 1 DIABETES

Jackson Lee¹, Alan Beltran², Trevor Bell³, Philly De Piante ³, Richard Wood³

¹dQ&A Market Research, Inc., San Francisco, CA., United States of America, ²dQ&A, Quantitative Research, San Francisco, United States of America, ³dQ&A - The Diabetes Research Company, San Francisco, United States of America

Background and Aims: Automated insulin delivery systems (AIDS) have become standard of care for people with Type 1 diabetes (T1). However, patients' trust in these systems remains largely unexplored. This study sought to investigate how trust in AIDS relates to improvements in the patients’ diabetes management regimes.

Methods: In May 2024, 1,503 T1 AIDS users in the United States took an online survey in which they estimated the frequency with which they verify their AIDS’ response to a change in blood glucose (“never” to “every time”). Subjects also reported the efficacy of their AIDS’ algorithm (“too conservative” to “too aggressive”), the cognitive benefits of their AIDS, and their A1c.

Results: Subjects were categorized as either passive (“never” to “sometimes”, n=857) or active AIDS reviewers (“every time” or “often”, n=829). A significantly higher percentage of passive reviewers had A1c >7 (32% vs 26%, p=0.020). Additionally, significantly more passive reviewers rated their AIDS algorithm as “just right” (33% vs 22%, p<0.0005) and reported reductions in the burden of diabetes, including thinking less about diabetes (44% vs 32%, p<0.0005), reduced effort in diabetes management (41% vs 34%, p=0.05), and fewer diabetes-related daily interruptions (38% vs 30%, p<0.05).

Conclusions: Passive reviewers tended to have weaker glycemic control but reported more optimal algorithm ratings and greater reductions in diabetes-related burdens compared to active reviewers. This suggests that AIDS may be most effective for those less proactive in their diabetes management, providing insight into a subset of people with diabetes who might benefit most from this technology.

YOUNG ADULT DIABETES TECHNOLOGY USE IN PEDIATRIC AND ADULT PRACTICES

Sean Delacey ^1,2, Saketh Rompicherla³, Shideh Majidi⁴, Naomi Fogel^1,2, Sarah Corathers⁵, Shivani Agarwal⁶, Roberto Izquierdo⁷, Lauren Golden⁸, Jill Weissberg-Benchell^2,9, Osagie Ebekozien^3,10

¹Northwestern University Feinberg School of Medicine, Department Of Pediatrics, Chicago, United States of America, ²Ann and Robert H. Lurie Children's Hospital, Chicago, United States of America, ³T1D Exchange, Boston, United States of America, ⁴Children's National Hospital, Department Of Endocrinology And Diabetes, Washington, United States of America, ⁵Cincinnati Children's Hospital, Division Of Endocrinology, Cincinnati, United States of America, ⁶Albert Einstein, Montefiore College of Medicine, New York, United States of America, ⁷Upstate University Hospital, Joslin Diabetes Center, New York, United States of America, ⁸NYU Langone Health, Center For Diabetes And Metabolic Health, New York, United States of America, ⁹Northwestern University Feinberg School of Medicine, Department Of Psychiatry & Behavioral Sciences, Chicago, United States of America, ¹⁰University of Mississippi School of Public Health, Jackson, United States of America

Background and Aims: Background and aims: People with Type 1 Diabetes (T1D) are more likely to have high HbA1C levels in emerging adulthood. The transition to adult practices is often difficult and the ideal age for transfer is unclear. We aimed to characterize differences in disease outcomes and care between pediatric and adult institutions in emerging adults with T1D.

Methods: We conducted a retrospective review of patients 18-23 years old, using data from the T1D Exchange (1/1/2022-12/31/2023) and categorized patients according to location of care (n=8,538 from 15 pediatric institutions, n=839 from 10 adult institutions). We compared group characteristics using t-tests or chi-square tests. Using logistic regression, we compared rates of optimal (HbA1C<7.0%) and suboptimal (HbA1C>9%) diabetes control and technology use between groups while controlling for covariates.

Results: Those that received care in adult institutions were older at T1D diagnosis (mean age: 12 vs 10.4 years, p<0.001) and at time of analysis (mean age: 21.4 vs 20 years, p<0.001). Those in adult centers were more likely to have suboptimal control (OR 1.23, p=0.03) and less likely to use a continuous glucose monitor (CGM) (OR 0.64, p<0.001) or an insulin pump (OR 0.62, p<0.001).

Conclusions: Young adults receiving care in adult vs pediatric centers appear more likely to have suboptimal diabetes control and less likely to use diabetes technology. The findings are limited by regional differences in centers surveyed and a smaller sample of adult patients. Research is needed to identify barriers to technology use for young adults in adult practices.

EFFECTS OF MODERN GLUCOSE MONITORING AND INSULIN DELIVERY DEVICES ON PATIENT-REPORT OUTCOMES IN TYPE 1 DIABETES: A SYSTEMATIC REVIEW AND META-ANALYSIS

Miao Gao ¹, Swathi Saravanan², Theresa Munyombwe³, Andrew Hill⁴, Gemma Traviss-Turner⁵, Ramzi Ajjan¹

¹Leeds Institute of Cardiovascular and Metabolic Medicine, the LIGHT Laboratories, University of Leeds, Leeds, United Kingdom, ²Leeds School of Medicine, University of Leeds, Leeds, United Kingdom, ³Leeds Institute of Cardiovascular and Metabolic Medicine/Leeds Institute of Data analytics, University of Leeds, Leeds, United Kingdom, ⁴Division of Psychological and Social Medicine, School of Medicine, University of Leeds, Leeds, United Kingdom, ⁵Academic Unit of Primary and Palliative Care, Leeds Institute Health Sciences, University of Leeds, Leeds, United Kingdom

Background and Aims: Patient-report outcomes measures (PROMs) help to provide insights into the psychological effects of a particular management strategy. Given that technologies [continuous glucose monitoring, insulin pumps or automated insulin devices (AID)] form a central part of type 1 diabetes (T1D) care, this systematic review aimed to identify the most frequently used PROMs and their modulation by technologies in this population.

Methods: Literature was searched in five databases between January 2012-May 2024. We included longitudinal diabetes technology studies assessing validated PROMs in non-pregnant adults with T1D. Results were pooled using random-effects meta-analysis, with subgroup analysis based on the type of technologies.

Results: Of 3814 studies identified, 85 (n=10053 participants) were included, of which 35 were RCTs and the rest were observational studies. A total of 70 PROMs were used across these studies with the Hypoglycaemia Fear Survey (HFS) most commonly used (n=34 studies), followed by Diabetes Treatment Satisfaction Questionnaires (DTSQs/DTSQc) (n=33), Diabetes Distress Scale (DDS) (n=23), and Problem Areas in Diabetes Questionnaire (PAID) (n=19). HFS-scores were significantly lower among users of diabetes technologies compared with controls (SMD=-0.177; 95% CI [-0.319, -0.036], I²=0.0%) with the largest effect observed in AID users. A moderate positive effect of diabetes technologies was observed on DTSQs scores (SMD=0.461; 95% CI [0.212, 0.709], I²=75.1%, Figure 1) with a small to moderate reduction on DDS/PAID scores (SMD=-0.266; 95% CI [-0.369, -0.163], I²=0.0%).

Conclusions: Diabetes technologies improve some PROMs outcomes, particularly fear of hypoglycaemia, treatment satisfaction, and diabetes distress. These non-glycaemic benefits should be considered in technology-related management decisions.

THE IMPACT OF CHATGPT USE ON THE PERCEPTION OF AI IN HEALTHCARE AMONG INDIVIDUALS WITH DIABETES

Anders Isaksen ¹, Adam Hulman^1,2, Jonas Schaarup^1,2

¹Aarhus University Hospital, Steno Diabetes Center Aarhus, Aarhus, Denmark, ²Aarhus University, Public Health, Aarhus, Denmark

Background and Aims: With the growing interest in applying artificial intelligence (AI) in the treatment of diabetes, it is important to understand how first-hand experiences with AI, such as ChatGPT, influence perceptions of benefits and risks of applying AI in healthcare. This study aimed to investigate how perceptions of AI in healthcare changed between 2022 and 2024 among individuals with diabetes and the impact of ChatGPT use.

Methods: Data on 2,917 individuals with diabetes in the 2022 and 2024 rounds of the Health in Central Denmark survey were analyzed. Participants reported their perception of AI as either “benefits outweigh risks” (B>R), “benefits equal risks” (B=R), “risks outweigh benefits” (R>B), or “don't know” (DK). ChatGPT use in 2024 was categorized into users and non-users.

Results: ChatGPT-use was reported by 22% of participants. At baseline, perception of AI was reported as: B>R: 39%, B=R: 18%, R>B: 3%, and DK: 40%. Among those reporting B>R in 2022, 69% of ChatGPT users maintained this view in 2024, compared to 49% of non-users. During follow-up, more ChatGPT users than non-users shifted to B>R from other baseline perceptions (DK: 24% vs. 13%, B=R: 41% vs. 24%, R>B: 23% vs. 11%). In the rest of the DK baseline perception group, 35% vs. 65% maintained this perception at follow-up, while 14% vs. 23% shifted to reporting B=R, and 18% vs. 8% shifted to R>B.

Conclusions: ChatGPT use was associated with a reduction of uncertainty in the diabetes population's perception of AI in healthcare, mainly due to more positive perceptions, but also awareness of risks.

CAREGIVER FACTORS RELATED TO USE OF DIABETES TECHNOLOGY IN ADOLESCENTS WITH T1D

Sarah Jaser

Vanderbilt University Medical Center, Pediatrics, Nashville, United States of America

Background and Aims: Adolescents are the age group least likely to use diabetes technology, and few studies have examined caregivers, who are closely involved in adolescents’ medical decisions. The current study aimed to identify caregivers’ demographic and psychosocial factors associated with device use in adolescents with type 1 diabetes (T1D).

Methods: Maternal caregivers of adolescents ages 11-17 with T1D (n=151, 56% female, 83% non-Hispanic White) reported on symptoms of depression, diabetes distress and diabetes-related family conflict. HbA1c and device use were collected from adolescents’ medical records.

Results: 78% of adolescents used continuous glucose monitors (CGMs), 62% used insulin pumps, and mean A1C was 9.0±2.1%. Non-Hispanic White adolescents were more likely to use insulin pumps and CGMs than adolescents of color (both p<.05). Adolescents from lower-income families were less likely to use insulin pumps (52%) than those from higher-income families (82%, χ²=13.1, p<.001), but there was no income-related difference in CGM use, and no differences in device use related to caregivers’ education or marital status. Caregivers of adolescents using injections reported significantly higher levels of diabetes distress and diabetes-related family conflict than those using insulin pumps (t=-2.1, -.2.2, respectively, both p<.05), but no significant differences in caregivers’ depressive symptoms emerged. No psychosocial differences were observed between caregivers of adolescents who were and were not using CGMs. After adjusting for adolescent race/ethnicity and family income, caregivers’ diabetes distress was significantly related to use of injections (β=.53, p<.05).

Conclusions: Caregiver factors are important to consider when optimizing use and reducing disparities related to diabetes technology, particularly insulin pump use.

ASSOCIATION BETWEEN NEW TECHNOLOGIES AND COPING STRATEGIES IN PATIENTS WITH TYPE 1 DIABETES MELLITUS: A CROSS-SECTIONAL ANALYSIS

Nuno Jesus ^1,2, Patrícia Tavares¹, Henrique Alexandrino¹, Margarida Gonçalves¹, Ana Sousa¹, Diogo Ramalho¹, Gustavo Rocha¹, Mariana Martinho¹, Marta Ferreira¹, Sara Correia¹, Maria João Oliveira¹

¹ULS Gaia e Espinho, Endocrinology, Vila Nova de Gaia, Portugal, ²i3S, University of Porto, Cancer Signalling & Metabolism, Porto, Portugal

Background and Aims: Despite technological advancements, managing Type 1 Diabetes Mellitus (T1DM) is intricate and significantly relies on the patient. Coping strategies are essential for navigating challenges related to T1DM. This study evaluates the association between coping strategies in people with T1DM and the use of Continuous Glucose Monitoring (CGM) and Continuous Subcutaneous Insulin Infusion (CSII).

Methods: T1DM patients were invited to answer the Brief Coping Orientation to Problems Experienced Inventory, evaluating 14 dimensions of coping, which can be categorized into Problem-Focused Coping (PFC), Emotion-Focused Coping (EFC), and Avoidant Coping (AC), or Adaptive and Maladaptive Coping. The data were analyzed using SPSS, with α<0.05.

Results: A total of 146 participants responded validly to the questionnaire, 50% of whom were women, with an average age of 37.5 years, an average disease duration of 17.2 years, and an average HbA1c of 7.9%. HbA1c <7% was associated with higher active coping scores (76.8% vs 63.2%; p<0.01) and lower EFC scores (41.5% vs 50.2%; p=0.025). CGM users (64%) scored higher in the PFC dimension than nonusers (59.6% vs. 46.2%, p<0.01), mainly due to active coping, positive reframing, and planning. CSII users (22%) showed lower AC scores (9.9% vs. 18.2%, p<0.01), primarily from lower reliance on self-distraction and denial. Maladaptive strategies were significantly lower among CGM users (15.3% vs 21.2%, p=0.028) and CSII users (12.3% vs 20.1%, p=0.016).

Conclusions: CGM was associated with greater use of PFC, while CSII was associated with lower use of AC. The use of new technologies was associated with less use of maladaptive strategies.

DIABETES TECH: WHO OPTS-IN AND HITS THE TARGET? HOW MANY CHOOSE DIABETES TECHNOLOGY WHEN GIVEN ACCESS, AND HOW MANY REACH THE HBA1C TARGET?

Herborg Líggjasardóttir Johannesen ¹, Jens Andreassen¹, Amar Nikontovic², Rasmus Borðoy³

¹The National Hospital of The Faroe Islands, Steno Diabetes Centre Faroe Islands, Tórshavn, Faroer Islands, ²Steno Diabetes Center North Denmark, Gistrup, Denmark, ³National Hospital of the Faroe Islands, Department Of Research, Tórshavn, Faroer Islands

Background and Aims: Despite recommendations for a target HbA1c of <7% (53 mmol/mol) in nonpregnant adults with type 1 diabetes (ADA 2023), fewer than 30% of individuals achieve this goal. The adoption of advanced diabetes technologies, including insulin pumps and continuous glucose monitoring systems, holds the potential to enhance glycemic control. Ensuring full reimbursement for these technologies, as we have on the Faroe Islands, is essential for equitable access to optimal diabetes care.

Methods: This cross-sectional nationwide study includes all individuals over 18 years old with type 1 diabetes living in the Faroe Islands from 2006 to 2024. We excluded persons with Latent Autoimmune Diabetes in Adults (LADA). Using data from electronic medical records, laboratory results, and ICD codes extracted on August 14, 2024, comparing hybrid closed loop (HCL) pump usage with the remaining levels of technology, which are non-hybrid pumps and MDI – with and without sensors and focuses on HbA1c levels (mmol/mol) as the primary outcome.

Results: Of the participants (n=217), 46% (n=100) used an HCL insulin pump, and 54% (n=117) did not use an HCL pump. 42% of participants achieved HbA1c <53 mmol/mol (7%), with distribution across HCL pump and remaining groups being 48% and 38% (p=0.12).

Conclusions: When given a free choice, 46% of adults preferred HCL systems, with the majority opting for CGM, highlighting a trend towards integrated diabetes management solutions. With 46% on hybrid closed-loop systems, the rates of achieving HbA1c <53 mmol/mol (7%) suggest that both advanced technologies and the remaining effectively support glycemic control when access is equitable.

PRETREATMENT AND MANAGEMENT OF SKIN REACTIONS DUE TO DIABETES DEVICES IN YOUTH AND YOUNG ADULTS (YYA) WITH TYPE 1 DIABETES (T1D)

Angela J. Karami, Valeria Fadhel Hernandez, Kasserine Taylor, Emily E. Fivekiller, Cari Berget, Samantha Lange, Erin Cobry

Barbara Davis Center for Diabetes, University of Colorado, Aurora, United States of America

Background and Aims: People with T1D often achieve improved glycemic control with diabetes devices (continuous glucose monitors [CGM] and insulin pumps). However, device users frequently experience skin irritation and reactions that disrupt use of these devices. We sought to understand which on-body locations experience the most skin reactions as reported in YYA and the skincare regimen or treatments used to alleviate these reactions.

Methods: Youth and young adults (2-25 years) with T1D at a diabetes clinic using diabetes devices were recruited over 2 months. The study consisted of an in-house developed survey assessing device location, skin reactions, and skincare surrounding device insertion and removal.

Results: Seventy-seven YYA enrolled (14.2±6.1 yrs, T1Ddur 5.9±4.9 yrs, HbA1c 7.0±1.0, 60% Female, 88% NHW, 100% CGM users, 90.1% pump users). Reported pertinent medical history included a diagnosed skin condition (9%) and environmental allergies (25%). The most common reaction sites were the upper arm (73%-CGM) and abdomen (53%-pump). Most participants reported some skincare regimen (62%-prior to device insertion; 34% post-removal; 85% prepped both pump and CGM sites). Skin prep was rated using a 1-10 scale with a higher score indicating greater satisfaction (6.8±2.5). The most common preparation included adhesive wipes (65%) followed by alcohol wipes or washing (25%).

Conclusions: Skin reactions to diabetes devices are common among YYA with T1D, and many use skincare routines to alleviate these reactions. Understanding the most effective techniques and skincare practices is crucial for guiding clinicians in providing better care and supporting device use for their patients.

IMPROVING HBA1C ON 3RD TRIMESTER TYPE 1 DIABETES PREGNANCY WITH CAMAPS FX: CLINICAL AND ECONOMIC OUTCOMES

Maria-Eleni Syleouni¹, Aesha Khan Mirón¹, Tara Lee², Eleanor Scott³, Helen Murphy ²

¹Ypsomed AG, Burgdorf, Switzerland, ²University of East Anglia, Norwich Medical School, Norwich, United Kingdom, ³Leeds Institute of Cardiovascular and Metabolic Medicine, University of Leeds, Leeds, United Kingdom

Background and Aims: Optimized glucose levels in pregnant women living with type 1 diabetes (PWwT1D) are associated with reduced risk of pregnancy and neonatal complications. The aim of this study was to assess the clinical and economic impact of improving HbA1c during third trimester in PWwT1D.

Methods: Leveraging the AiDAPT study, a health economics model was constructed. AiDAPT HbA1c distribution and associated clinical outcomes probabilities (obstetric: preeclampsia, method of delivery; neonatal: length of stay in normal/critical/intensive care) were used together with UK healthcare costs extracted from NHS National Schedule and other published sources. Total costs to the UK healthcare system were estimated using an incremental 0.3% HbA1c reduction achieved with CamAPS FX over the standard of care (between group mean adjusted difference).

Results: The analysis showed improved obstetric and neonatal outcomes, such as reduced risk of preeclampsia (-13%), reduced hospitalization days for neonatal critical care of high dependency (-16%) and neonatal intensive care unit (-12%). Cost savings for the UK healthcare system deriving from improved obstetric and neonatal outcomes were estimated at £0.9M per 1,000 PWwT1D using CamAPS FX instead of standard of care.

Conclusions: This rather conservative analysis showed the impact of further optimizing maternal HbA1c in PWwT1D. An additional 0.3% HbA1c drop achieved with CamAPS FX vs standard of care lead to an 8% cost reduction. CamAPS FX is the only AID system approved during pregnancy and offers HCPs a way to help PWwT1D achieve tighter glycemic goals while helping reduce clinical and economic burden for healthcare systems.

TESTS IN TIGHT RANGE IN GESTATIONAL DIABETES MELLITUS - HOW CAN AN MHEALTH APP HELP?

Michael Steinhauser-Mitter ¹, Johanna Kober¹, Ezgi Cinar¹, Josip Zivkovic¹, Stephan Silbermann²

¹mySugr GmbH, Vienna, Austria, ²Roche Diabetes Care, Mannheim, Germany

Background and Aims: Gestational diabetes mellitus (GDM) is a common complication of pregnancy with stringent euglycemic ranges to reduce the risk of fetal complications. This puts a significant burden on patients and creates an opportunity for mHealth apps to help simplify diabetes self-management. We aim to assess the impact of using the mySugr® mHealth application on the glycemic control of people with GDM, with a specific focus on tests in tight range (70-140 mg/dl) and tests above tight range (> 140 mg/dl) after three month of app use.

Methods: We define a study cohort of users with GDM that have used the application for at least 90 days with a bluetooth connected BG-meter and quantified change in glycemic metrics as compared to baseline. Statistical testing was done by performing a two-sided, one-sample t-test.

Results: The cohort of 1377 users showed a statistically significant change in eHbA1c (Δ eHbA1c[%]=-0.09 ± 0.01; p<.001), while being already well controlled at baseline. Tests in tight range increased by 1.33 ± 0.24 percentage points (p<.001), and tests above tight range, dropped by -1.59 ± 0.24 percentage points (p<.001) over the study period, falling below the recommended threshold of 10%.

Conclusions: These results suggest that usage of an mHealth app in a population with GDM can help improve pregnancy specific glycemic targets such as tests in tight range and tests above tight range over a time frame of three months. The results are encouraging and warrant further investigation in a randomized control trial.

OLDER ADULTS WITH TYPE 1 DIABETES AND HYPOGLYCEMIA: A RANDOMIZED CONTROLLED STUDY USING CGM COMBINED WITH GERIATRIC PRINCIPLES

Medha Munshi ^1,2,3, Christine Slyne⁴, Atif Adam⁴, Colin Conery⁴, Adeolu Oladunjoye⁵, Simon Neuwahl⁶, David Wypji^4,5, Elena Toschi^2,3

¹Beth Israel Deaconess Medical Center, Boston, United States of America, ²Harvard Medical School, Boston, United States of America, ³joslin diabetes center, Boston, United States of America, ⁴Joslin Diabetes Center, Clinical And Behavioral Research, Boston, United States of America, ⁵Boton Children's hospital, Boston, United States of America, ⁶Research Triangle Institute, chapel hill, United States of America

Background and Aims: CGM use in older adults have shown benefits in improving glycemic outcomes. However, the impact of CGM use, coupled with geriatric principles that are better suited for vulnerable older adults with multiple comorbidities and hypoglycemia, is not known. In this study, we evaluated the effectiveness of eCGM (CGM combined with geriatric principles that included adjusting goals and simplification of regimens based on overall health) in older adults with T1D having hypoglycemia.

Methods: Older adults (≥65 years) with T1D who were CGM-naïve or CGM-users with hypoglycemia (≥2 episodes <70 mg/dL for ≥20 minutes over 2 weeks) were randomized to intervention or control groups. The intervention consisted of eCGM use over a 6-month period. Control group received usual care. The primary endpoint was change in time <70 mg/dL from baseline to 6-months. Cost-effectiveness was calculated using a healthcare sector perspective.

Results: We randomized 131 older adults (age 71±5 years; 21% ≥75 years) to intervention (n=68; CGM-users=33) or control (n=63; CGM-users=40) groups. The median change in hypoglycemia from baseline to 6-months was –2·6% in the intervention and –0.3% in the control group (median difference –2.3%,95%Cl [–3.7%,–1.3%];P<0.001). This improvement was seen in both CGM-naïve (–2.8%,[–5.6%,–0.8%]) and CGM-users (–1.2%,[–2.7%,–0.1%]). The HbA1C did not differ between the groups (7.5% vs 7.3%). The intervention was cost-effective (ICER, $71,623/QALY).

Conclusions: In older adults with T1D and hypoglycemia, use of eCGM lowered hypoglycemia in CGM-naïve as well as CGM-user cohort, without worsening glycemic control in a cost-effective fashion. Acknowledgement: NIH DP3 Grant (1DP3DK112214-01); NIH P30 Grant (P30DK036836); partical material support by Dexcom.

SIMPLIFICATION OF MEALTIME INSULIN REGIMEN IN OLDER ADULTS WITH TYPE 1 DIABETES AND HYPOGLYCEMIA

Elena Toschi^1,2, Molly Savory³, Colin Conery³, Christine Slyne³, Sam Carl³, Medha Munshi ^1,2,4

¹Joslin Diabetes Center, Boston, United States of America, ²Harvard Medical School, Boston, United States of America, ³Joslin Diabetes Center, Clinical And Behavioral Research, Boston, United States of America, ⁴Beth Israel Deaconess Medical Center, Boston, United States of America

Background and Aims: In our recent study we showed that continuous glucose monitoring (CGM) coupled with simplified treatment regimens and personalized glycemic goals (eCGM) reduces hypoglycemia in older adults with type 1 diabetes (T1D). In this study, we describe the impact of simplification of mealtime insulin regimen on post-prandial hyperglycemia.

Methods: Older adults ≥65 years with T1D with hypoglycemia were randomized to intervention (eCGM) and control groups. The intervention included simplification strategies around meal insulin boluses based on CGM and patient characteristics and comprised of avoiding carbohydrate counting; using fixed insulin doses for small or large meals, avoiding correction doses between meals, for snacks, or at bedtime. Control group received usual care. Changes in post-prandial hyperglycemia episodes (defined as ≥ 15 minutes of time >180 mg/dL occurring between 3-5 hours from rise of glucose of>50 mg/dL) from baseline to 6-months were assessed along with glycemic variability (CV %).

Results: Sixty-eight older adults with T1D (age 71±4 years; diabetes duration 38±17 years) were randomized to intervention (n=34) and control (n=34) groups. The average number of post-prandial episodes were similar in intervention and control groups at baseline. At 6-months, a reduction of episodes of post-prandial hyperglycemia were observed in the intervention group (Figure). In addition, CV decreased in the intervention group at 6-month (CV 39 vs 33%; p<0.0001) and did not change in the control group (39 vs 38%; p=0.4).

Conclusions: Simplification of insulin dosing strategies associated with mealtime boluses in older adults with T1D may improve post-prandial hyperglycemia and glucose excursion.

COVALENT-111: EVALUATING LONG-TERM EFFICACY AND SAFETY OF SHORT-TERM ICOVAMENIB TREATMENT IN PERSONS WITH TYPE 2 DIABETES

Alexander Abitbol¹, Jose Rodriguez², Sanchita Mourya³, Jeffrey Kim³, Thomas Butler³, Juan Frias ³

¹LMC Diabetes & Endocrinology, Oakville, Canada, ²South West General Healthcare Center, Fort Myers, United States of America, ³Biomea Fusion, Redwood City, United States of America

Background and Aims: T2D is characterized by a progressive decline in beta-cell function. Menin, a scaffold protein, has emerged as a promising target for T2D treatment due to its role in beta-cell proliferation and insulin secretion. In the pancreas, menin interacts with transcriptional partners that drive the expression of cell cycle protein regulators, including those that prevent the replication and expansion of insulin producing beta-cells. Icovamenib, a novel oral covalent menin inhibitor, has shown potential in early T2D studies and is currently being investigated in the Phase 2b Expansion of the COVALENT-111 trial (NCT05731544). Herein we describe the design of this study.

Methods: This 52-week, randomized, double-blind, placebo-controlled study enrolls adults with T2D (duration ≤7 years, HbA1c 7-10.5%). Participants are allocated to three icovamenib treatment arms: 1) 100mg daily for 8 weeks, then 4 weeks placebo; redosing if required after Week 22 for an additional 4 weeks 2) 100mg daily for 12 weeks; 3) 100mg daily for 8 weeks, then 200mg daily for 4 weeks. Each arm (n=72) is randomized 3:1 (icovamenib:placebo). Key outcomes include changes in glycemic parameters and durability of response at Week 26 and Week 52. Safety and tolerability are continuously monitored.

Results: This trial is ongoing in the US and Canada. Twenty-six-week follow-up efficacy and safety data, which are currently pending, will be presented.

Conclusions: Icovamenib addresses the root cause of T2D, beta-cell dysfunction, with a novel mechanism of action. COVALENT-111 aims to evaluate the long-term efficacy and safety of short-term icovamenib therapy, potentially addressing a significant unmet need in T2D management.

MYDIAMATE FOR REMISSION OF DIABETES DISTRESS IN TYPE 1 DIABETES: A MULTI-NATIONAL RANDOMISED-CONTROLLED TRIAL

Theresa Mohr ¹, Maartje De Wit², Jiska Embaye¹, Dominic Ehrmann³, Norbert Hermanns³, Gina Lehmann³, Kirsty Winkley⁴, Alexandra Famiglietti⁴, Sarah Williams⁴, María Teresa Anarte Ortiz⁵, Laura Torreblanca Murillo⁵, Francois Pouwer⁶, Frank Snoek²

¹Amsterdam UMC, Amsterdam, Netherlands, ²Amsterdam Public Health Research Institute, Medical Psychology, Amsterdam, Netherlands, ³FIDAM - Research Institute Diabetes-Academy Bad Mergentheim, Bad Mergentheim, Germany, ⁴King's College London, London, United Kingdom, ⁵Universidad de Málaga, Málaga, Spain, ⁶Odense University Hospital, Steno Diabetes Center Odense, Odense, Denmark

Background and Aims: Up to 40% of persons with Type 1 Diabetes (PwT1D) experience significant diabetes-distress, with limited psychological support available. We developed MyDiaMate, an online unguided psychological self-help application. MyDiaMate consists of 6 modules on healthy coping, mood, fatigue, eating, hypoglycaemia, and social network. Following pilot studies with encouraging results, we launched the multinational RCT “MyREMEDY”, sponsored by JDRF, to investigate effectiveness of MyDiaMate in reducing diabetes-distress in PwT1D.

Methods: In spring 2024 we started recruiting 660 adults with T1D in the Netherlands, Germany, Spain, and the United Kingdom. Participants randomised into the intervention group receive access to MyDiaMate for 6 months, those in the control group receive access after 3 months, for 3 months. Measurements are scheduled at 0, 3, and 6 months. Primary outcome is diabetes-distress (PAID-20). Secondary outcomes are: Mood (WHO-5), fatigue (CIS fatigue subscale), social network (LSNS-6), diabetes self-efficacy (DES-SF), and glycaemic control (self-report).

Results: A Dutch pilot study (n=55) confirmed feasibility and acceptability of MyDiaMate along with positive user experiences. In a second, larger study (n=1008), MyDiaMate was made freely available for Dutch adults with T1D to test uptake and app usage, which showed consistent interest and different user behaviours, i.e. completion rates, driven by individual needs and preferences. By September 2024, we have recruited almost 50% of participants in the multi-national MyREMEDY RCT.

Conclusions: MyDiaMate is an innovative, accepted, and feasible self-guided online application addressing the mental health of PwT1D. If shown efficacious in the MyREMDY trial, MyDiaMate is ready to be disseminated to a large population of PwT1D.

DIVERSITY IN TYPE 1 DIABETES (T1D) CLINICAL TRIALS: A SIGNIFICANT GAP IN REPRESENTATION

Ananta Addala ¹, Lauren Von Zabern², Lauren Figg¹, Wendy Espinoza¹, Diya Pillai¹, Ricardo Medina Penaranda¹, Susanne Cabrera²

¹Stanford University, Stanford, United States of America, ²Medical College of Wisconsin, Milwaukee, United States of America

Background and Aims: Clinical trials set the precedent for care in clinical settings. Inadequate representation of diverse populations can lead to inaccurate or incomplete conclusions, compromising treatment effectiveness and quality. The goal of this review was to quantify disparities by sociodemographic factors in T1D clinical trials over the last two decades, focusing on trials of diabetes technology and immunotherapies.

Methods: Prespecified search criteria queried PubMed to identify all clinical trials in diabetes technology and therapeutics from 2004-2023 in the US. The initial search yielded 962 studies. Title review narrowed the search to 231 studies and abstract review further narrowed to 121 studies. After further exclusion during the data abstraction phase, 78 studies were included in this analysis. For each sociodemographic variable, the percentages were averaged for categories presented in each study and scaled to 0-100%.

Results: Clinical trials evaluated technology (68.8%), immunotherapies (28.6%), or both (2.6%). Studies included adult and pediatric populations (58.4%) while a smaller portion only included pediatrics (23.4%) or adults (18.2%). The sociodemographic characteristics captured varied across the 78 studies (gender/sex=74, race/ethnicity=50, education=21, income=17, insurance=12). Studies primarily recruited White/Caucasian (81.0%), and privately-insured (82.9%) individuals (Table). The diversity of the cohort only marginally improved over the 20-year period, with a 5% improvement in diversity of race and ethnicity.

Conclusions: Clinical trials have not consistently captured sociodemographic traits and consistently underrecruit minoritized populations. The next steps to equitable clinical trials include the documentation of minoritized identities and strategies to ensure that clinical trials are as diverse as the individuals living with T1D.

CARDIOVASCULAR RISK EVALUATION IN PEOPLE WITH CYSTIC FIBROSIS

Irene Aguilera García ¹, Paula Vázquez Pérez¹, Bricia López Plaza², Samara Palma Milla¹, Ruth Gaspar Lafuente¹, Ester Zamarrón De Lucas³, M Concepción Prados Sánchez³, Noemí González Pérez De Villar¹

¹Hospital Universitario La Paz, Endocrinology, Madrid, Spain, ²IdiPaz, Madrid, Spain, ³Hospital Universitario La Paz, Pneumology, Madrid, Spain

Background and Aims: Life expectancy in people living with Cystic Fibrosis (CF) has been increasing in the last decades thanks to advances in treatment. CF patients are reaching and passing middle age, but the cardiovascular risk (CVR) of this population has been understudied. This study aimed to evaluate the presence of CVR factors in patients with CF.

Methods: A prospective study was conducted on 39 patients affected by CF (with and without diabetes) who had a clinical and biochemical evaluation in a third-level hospital and wore a FreeStyle Libre 2 sensor for 14 days.

Results: Patient’s mean age was 33.7±11.6 years, ranging from 18 to 58 years old. 36% were females. Every patient had a low CVR calculated by SCORE2. However, older age was correlated to higher weight and Body Mass Index (BMI; r=0.573 and r=0.545 respectively, p<0.001), as well as higher waist circumference (WC) and waist-to-height ratio (r=0.761 and r=0.744 respectively, p<0.001). Weight, BMI, WC, and waist-to-height ratio correlated with elevated fasting glycemia (r=0.354, p=0.027; r=0.393, p=0.013; r=0.508, p=0.009 and r=0.498, p=0.011). WC and waist-to-height ratio are also associated with worse lipidic profile: higher total cholesterol (r=0.517, p=0.008), LDL-cholesterol (r=0.396, p=0.049), and triglycerides (r=0.528, p=0.007). Lower HDL-cholesterol levels were related to the diagnosis of diabetes and hyperglycemia >180 mg/dL (r=-0.338, p=0.038).

Conclusions: In the present cohort, patients with CF showed classical CVR factors such as altered lipid profile, elevated fasting glycemia, overweight, and high WC, which increased with age, but not well evaluated by SCORE2. Studies are needed to propose an adapted CVR score for patients with CF.

TYPE 1 DIABETES AND SPATIAL COGNITION

Hussein Zaitoon^1,2, Liat Perl^1,2, Asaf Oren^1,2, Eyal Cohen-Sela¹, Irina Laurian¹, Anna Dorfman¹, Efrat Chorna¹, Hagar Interator¹, Yael Lebenthal^1,2, Avivit Brener ^1,2

¹Dana-Dwek Children's Hospital, Tel Aviv Sourasky Medical Center, Pediatric Endocrinology, Tel Aviv, Israel, ²Tel Aviv University, Faculty Of Medicine, Tel Aviv, Israel

Background and Aims: Glycemic excursions in patients with type 1 diabetes (T1D) raise concerns about potential cognitive impairment. We investigated spatial cognition in youth with T1D, and its relationship with glycemic control and weight status.

Methods: This cross-sectional study included patients aged 10-20 years with T1D duration of >1 year. The adapted Moris Water Maze, a validated computer-based virtual reality tool, was used to assess spatial learning and memory retrieval of the study participants. A control group of healthy sex- and age-matched volunteers was used to establish Water Maze task performance for comparison. Data collected from medical records included: diabetes duration, HbA1c at diagnosis, current HbA1c, time-in-range, and BMI z-score. A sex- and age-adjusted multiple linear regression stepwise model was applied to identify potential contributors to longer time-to-platform.

Results: Sixty-six participants (68.2% males) with T1D completed the task at age 15.6±3.0 years, median diabetes duration of 6.1[3.6,10] years, current HbA1c level of 7.1 ± 0.9% and 63.4 ± 18.3% time-in-range. Water Maze task performance did not differ between patients with T1D and controls. Boys with T1D completed the task faster than girls with T1D (p<0.001). Positive associations were found between time-to-platform and BMI z-score (r=0.31, p=0.012), and diabetes duration (r=0.40, p=0.002). Diabetes duration was the only significant contributor for time-to-platform: for each year increase in diabetes duration, time-to-platform increased by 0.423 seconds (β=0.423, p=0.002).

Conclusions: In youth with T1D, despite showing similar performance to controls, longer diabetes duration and higher weight status were related to worse spatial cognition. These concerning findings call for additional research.

THE ROLE OF DIABETES TECHNOLOGY IN REDUCING EMERGENCY DEPARTMENT ATTENDANCES WITHIN BART'S HEALTH NHS TRUST

Rachel Flynn ¹, Peter Jacob¹, Christopher Garrett¹, Meera Ladwa², Kirun Gunganah², Kalpita Majumdar³, Charlotte Waters⁴, Belinda Krishek⁴, Bobby Huda¹

¹Royal London Hospital, Bart's Health NHS Trust, Diabetes & Endocrinology, London, United Kingdom, ²Diabetes Department, Newham Hospital, Bart's Health Trust, London, United Kingdom, ³Diabetes Department, Whipps Cross Hospital, Bart's Health NHS Trust, London, United Kingdom, ⁴Pharmacy and Medicines Optimisation Team, NHS North East London, London, United Kingdom

Background and Aims: To assess the impact of using diabetes technology in patients with type 1 diabetes (PwT1D) on attendances to emergency departments (ED) within Bart’s Health NHS Trust.

Methods: A retrospective analysis of ED attendance data from the business intelligence team covering financial years 2021-2024 was reviewed in adult PwT1D with a diagnosis of diabetic ketoacidosis (DKA), hyperglycaemia or hypoglycaemia. Blueteq funding request data was extracted, detailing the different diabetes technologies used -continuous glucose monitoring (CGM), open loop insulin pumps (OL) & hybrid closed loop systems (HCL).

Results: 602 PwT1D (18-81 years; median 35, [IQR 27.8-44] used diabetes technologies (CGM 184, OL 212, HCL 206) between 2021-2024. 12% (72/602) of our cohort had one or more ED attendances. Of ED attendees, 79.2% (57/72) attended prior to the initiation of technology, with no attendances afterwards. 15 patients attended despite use of diabetes technologies, detailed in figure 1. It is of note that of those established or starting on diabetes technologies during our collection period, only 4.35% (8/184) of CGM users, 1.89% (4/212) of OL users and 1.46% (3/206) of HCL users attended ED for DKA, hypoglycaemia or hyperglycaemia.

Conclusions: Our findings demonstrate that the use of technology positively impacts diabetes related ED attendances with implications for health and psychosocial outcomes. This is especially relevant given the recent expansion in eligibility for diabetes technologies. Increased healthcare professional input accompanying technology initiation may also be an important factor. For those attending despite technology use, factors including health and tech literacy, compliance and connectivity may play a role.

COMORBIDITY TRAJECTORIES AMONG DIABETIC PATIENTS AND NON-DIABETICS WITH VARYING RISK LEVELS

Puja Goswami ¹, Dilip Tr², Yogesh Shejul³, Anjali Kulkarni⁴

¹International Institute for Population sciences, Population Sciences, Mumbai, India, ²International Institute for Population Sciences, Population Sciences, Mumbai, India, ³Bhabha Atomic Research Centre, Medical Division, Mumbai, India, ⁴Homi Bhabha National Institute, Medical Division, Mumbai, India

Background and Aims: Diabetes mellitus is a growing global health threat, with 537 million people affected in 2021, projected to rise to 784 million by 2045. The majority of patients live in low- and middle-income countries, with India contributing significantly to this burden. Diabetes is linked to microvascular and macrovascular complications, as well as conditions like arthritis and depression. While Indian studies focus on comorbidity prevalence, there is a need for longitudinal research to better understand comorbidity progression and risk across different populations.

Methods: This retrospective cohort study analyzed data from 2,505 participants in the CHIPS cohort from 2011 to 2020, divided into diabetic, prediabetic, and non-diabetic groups. Kaplan-Meier survival curves and log-rank tests assessed comorbidity progression. Incidence rates and ratios were calculated, stratified by sex and age groups (30–44, 45–59, and 60+), using electronic health records for follow-up.

Results: Kaplan-Meier analysis showed that hypertension incidence was highest in diabetics, followed by prediabetics, with rates increasing with age. Women in non-diabetic groups had lower hypertension rates than men, but this difference disappeared in diabetics. Hypothyroidism appeared earlier in prediabetics, particularly in women aged 45-59. Low HDL incidence was similar in diabetics and prediabetics, with women showing higher rates. High LDL, STC, and TGL incidences were most common in prediabetics, especially in women and those aged 45-59. Gender disparities were noted, with women having higher incidences in diabetics, particularly for HDL and TGL levels.

Conclusions: Prediabetes increases the risk of metabolic and cardiovascular conditions, highlighting the need for early detection and targeted interventions.

ADVANCED HYBRID CLOSED LOOP SYSTEMS FOR THE MANAGEMENT OF INSULIN THERAPY IN POST-TRANSPLANT DIABETES MELLITUS: PRELIMINARY AND ENCOURAGING DATA FROM A SINGLE CENTER POPULATION

Valeria Grancini, Irene Cogliati, Alessia Gaglio, Veronica Resi, Emanuela Orsi

Fondazione IRCCS Ca' Granda - Ospedale Maggiore Policlinico di Milano, Endocrinology, Milan, Italy

Background and Aims: Post-transplant diabetes mellitus (PTDM) is one of the most common complications of solid organ transplantation and has been related to a poorer long-term prognosis. Actually, given the lack of evidence concerning the therapeutic algorithms and the glycemic targets in this population, the same guidelines for type 2 diabetes have been proposed. Among these individuals, a non-insulin glucose lowering therapy can’t be considered in a small minority, due to concomitant comorbidities.

Methods: Among people with PTDM followed at our Center, 5 was proposed to manage their insulin therapy with an AHCL system. Glucose control parameters, Ambulatory Glucose Profiles (AGP) data, insulin requirements, device settings, and questionnaire outputs (DTSQ and ADDQoL) were collected before and after switching to the AHCL system.

Results: We observed an improvement in diabetes management (HBA1c levels from 7.7±0.7 to 6.5±0.4, %; mean glucose from 182.0±45.2 to 143.0±12.0, mg/dl; GMI from 7.78±0.77 to 6.72±0.28, %; CV from 37.2±3.3 to 28.2±1.8 and TIR from 58.4±14.1 to 83.8±0.70, %). All patients referred an improvement in quality of life and diabetes therapy satisfaction (DTSQ total score from 26.6±1.4 to 30.6±0.9, mean±SD; AWI score from -2.34±1.17 to -1.05±0.2, mean±SD).

Conclusions: These data could be considered a first insight on the use of AHCL systems in solid organ transplanted patients in whom insulin is the only option to manage PTDM and in whom an MDI scheme is not sufficient to reach therapeutic targets. In this small but fragile population, these devices could be a valid therapeutic strategy, considering their positive impact on glucose control and quality of life.

THREE DECADES OF GLUCOSE-LOWERING THERAPY IN PATIENTS AT HIGH CARDIOVASCULAR RISK– A REAL-WORLD ANALYSIS

Magdalena Neyer^1,2, Johannes Vogel^1,2, Pascal Elsner^1,2, Heike Kuehrer¹, Christoph Saely^1,2, Axel Muendlein¹, Alexander Vonbank^1,3, Arthur Mader^1,3, Andreas Festa¹, Heinz Drexel^1,2, Andreas Leiherer ^1,2

¹Vorarlberg Institute for Vascular Investigation & Treatment (VIVIT), Feldkirch, Austria, ²Private University in the Principality of Liechtenstein, Triesen, Liechtenstein, ³Academic Teaching Hospital Feldkirch, Feldkirch, Austria

Background and Aims: Over recent years, therapy options and strategies for type 2 diabetes mellitus (T2DM) have developed substantially. This study investigated glucose-lowering treatment in patients with high cardiovascular risk over three decades.

Methods: A total of 2158 patients undergoing elective coronary angiography at a tertiary care hospital in Europe were included in three sequential observational studies (OS): OS1 (1999–2000; n=672), OS2 (2005–2008; n=1005), and OS3 (2022–2023; n=481). Sociodemographic data, patient-reported medication, medical histories and blood samples were analyzed.

Results: A clear trend towards more complex glucose-lowering therapies was found. A wider array of glucose-lowering drugs was used over time (OS1: 11; OS2: 21; OS3: 25) and the number of different drugs used in combination therapy in a single patient increased to a maximum of five in OS3. Furthermore, substantial differences in applied medication regimens were observed: Sodium-glucose cotransporter-2 inhibitors were the most frequently reported substance class (34.0% of total reported glucose-lowering drugs) in OS3 while metformin remained a key component (OS1: 33.9%; OS2: 41.8%; OS3: 32.0%). Other drug classes like sulfonylureas were largely replaced. A total of 69.2% of patients in OS3 achieved an HbA1c level of <7% (vs. OS1: 51.9%, OS2: 54.7%; ptrend<0.001). Over 25% of patients with T2DM were newly diagnosed at admission (OS1: 43.8%, OS2: 29.7%, OS3: 27.2%; ptrend<0.001) and had therefore no diabetes-related medication.

Conclusions: These real-world data emphasize a marked shift in T2DM treatment towards novel substance classes. However, the use of incretin mimetics remained low. Significantly more patients reached HbA1c targets in the most recent cohort.

DEFECTIVE BLOOD GLUCOSE CONTROL IS ASSOCIATED WITH HIGH FREQUENCY OF TR3-56 IMMUNE CELLS DURING TYPE 1 DIABETES PROGRESSION

Enza Mozzillo ¹, Francesca Di Candia¹, Alessandro Fierro¹, Ludovica Fedi¹, Irene Cuccurullo², Sara Bruzzaniti³, Erica Piemonte³, Francesco Rosanio¹, Mario Galgani³

¹Università degli Studi di Napoli Federico II, Naples, Italy, Department Of Translational Medical Science, Section Of Pediatrics, Naples, Italy, ²Federico II University Hospital, Naples, Italy, ³Istituto per l’Endocrinologia e l’Oncologia Sperimentale, Consiglio Nazionale delle Ricerche,Naples, Italy, Department Of Molecular Medicine And Medical Biotechnology, University Of Naples “federico Ii”, Naples, Italy, Naples, Italy

Background and Aims: CD3+CD56+ regulatory T cells (Tr3-56) may be involved in the progression of beta cell destruction in subjects with type 1 diabetes (T1D). Aim of our study was to evaluate the association between glucometabolic control with frequency of Tr3-56 and their characteristics both at the onset (T0) of T1D and after 1 year (T1).

Methods: We enrolled 52 children with T1D (10.2 ± 2.1 23M) and 36 healthy controls (8.5 ± 3.15 years, 16 M). At T0 and T1 we evaluated: frequency of Tr3-56 by flow cytometry; glucometabolic control assessed by mean glycosylated hemoglobin (HbA1c %), and time in range (TIR), time above range (TAR) of continuous glucose monitoring (CGM) of the last 14 days.

Results: The frequency of Tr3-56 in T1D was reduced at T0 (P<0.005), and increased at T1 (P<0.0001) compared to healthy controls; while at T1 was positively correlated with HbA1c levels (r=0.4107; P=0.0158) and TAR (r=0.4384; P=0.0084); negatively correlated with TIR (r=-0.3632; P=0.0320); and higher in subjects with TIR<70% respect to subjects with TIR>70% (P=0.034). Finally, citotoxicity markers such as CD16 and NKG2C, were highly expressed on Tr3-56 at T1 in subjects with T1D (P=0.05 and P=0.0019, respectively).

Conclusions: These data showed at T1 an association between both hyperglycemia and the frequency of Tr3-56, and between hyperglycemia and the cytotoxic function. The identification of specific targets of cells involved in the pathogenesis of T1D could lead to the control of the progression of beta cell destruction in T1D but further investigation is needed.

PARTICIPANT EXPERIENCES USING SEMAGLUTIDE AS ADJUNCT TO AUTOMATED INSULIN DELIVERY: QUALITATIVE ANALYSES AND QUALITY-OF-LIFE MEASURES

Melissa-Rosina Pasqua ¹, Joelle Doumat², Michael Tsoukas¹, Ahmad Haidar¹

¹McGill University Health Centre - Royal Victoria Hospital, Endocrinology & Metabolism, Montreal, Canada, ²McGill University Health Centr, MONTREAL, Canada

Background and Aims: Interest in GLP1-RA's for type 1 diabetes is increasing, but data lacks regarding quality-of-life. The aim of this substudy was to assess participant-reported outcomes and experiences with semaglutide (vs. placebo) with AID.

Methods: This is a sub-analysis from our randomized, placebo-controlled, cross-over trial to assess semaglutide vs. placebo as add-on therapy to AID. The following quality-of-life questionnaires were performed after each intervention: diabetes distress scale (DDS), hypoglycemia fear survey (HFS), diabetes treatment satisfaction questionnaire (DTSQ), INSPIRE, and diabetes bowel symptom questionnaire (DBSQ). Semi-structured interviews were performed at the end of the trial. Interviews were recorded, transcribed, and coded by research personnel for themes using an inductive-deductive approach.

Results: Twenty-three participants completed questionnaires, while 24 performed interviews. For semaglutide vs. placebo, there were no statistically significant differences in the mean scores for the DDS, HFS, INSPIRE, and DTSQ. The DBSQ demonstrated semaglutide increased symptom frequency and severity, in particular nausea/vomiting, constipation, and diarrhea. Nausea was a barrier for carb counting and pre-bolus. Synergy was noted between the drug and AID for glucose control due to appetite suppression. Forty-two percent were certainly interested in semaglutide use, with reasons being lower insulin requirements, weight loss, and improved glycemic control. Many of these qualities, including complication risk reduction, were attributes of the ideal adjunctive therapy per participants.

Conclusions: GLP1-RA therapy is of interest to those with type 1 diabetes. Attention to bolus experience should be given during dose titration. Quality-of-life surveys did not capture differences between semaglutide and placebo outside of increased gastrointestinal side effects, but interviews did.

IMPROVED PATIENT- AND PHYSICIAN-REPORTED OUTCOMES WITH IGLARLIXI WHEN SWITCHING FROM PREMIXED INSULIN IN PEOPLE WITH TYPE 2 DIABETES (T2D): THE SOLI-SWITCH STUDY

William Polonsky ¹, Martin Haluzík², Katarzyna Cypryk³, Rachid Malek⁴, Hakan Korkmaz⁵, Soo Lim⁶, Awadhesh Singh^7,8, Samia Bokhari⁹, Agustina Alvarez¹⁰, Felipe Lauand¹¹, Yan Ren¹², Vivian Fonseca¹³

¹Behavioral Diabetes Institute, San Diego, United States of America, ²Diabetes Centre, Institute for Clinical and Experimental Medicine (IKEM), Prague, Czech Republic, ³Department of Internal Diseases and Diabetology, Medical University of Lodz, Lodz, Poland, ⁴Internal Medicine Department, CHU Mohamed Saadna Abdennour, Sétif, Algeria, ⁵Division of Endocrinology and Metabolism, Department of Internal Medicine, Faculty of Medicine, Süleyman Demirel University, Isparta, Turkey, ⁶Department of Internal Medicine, Seoul National University College of Medicine and Seoul National University Bundang Hospital, Seoul, Korea, Republic of, ⁷G. D Hospital & Diabetes Institute, Kolkata, India, ⁸Sun Valley Hospital & Diabetes Research Centre, Guwahati, India, ⁹Department of Endocrinology and Diabetes, King Fahd Armed Forces Hospital, Jeddah, Saudi Arabia, ¹⁰Sanofi, Madrid, Spain, ¹¹Sanofi, Paris, France, ¹²Aixial on behalf of Sanofi, Gentilly, France, ¹³Tulane University Medical Center, New Orleans, United States of America

Background and Aims: iGlarLixi is a fixed-ratio combination of insulin glargine 100 U/mL plus lixisenatide. Soli-Switch demonstrated that switching from premixed insulin (PI) to iGlarLixi resulted in improved glycaemic outcomes, body weight benefits, without increase in hypoglycaemia risk. This exploratory analysis assessed patient- and physician-reported outcomes in people with T2D switching from PI to iGlarLixi.

Methods: In 24-week, single-arm Soli-Switch (EudraCT number 2021-003711-25) study, 162 participants switched from once- or twice- daily PI to once-daily iGlarLixi. Participants had to be aged ≥18 years, diagnosed with T2D for ≥6 months and had received PI for ≥3 months. Patient-reported outcomes were assessed using the Treatment-Related Impact Measure-Diabetes (TRIM-D) questionnaire from baseline to Week 24. Patient- and physician-reported outcomes were assessed using Global Treatment Effectiveness Evaluation (GTEE) questionnaire from Week 12 (end of titration period) to Week 24.

Results: Over 24 weeks, in addition to HbA1c reduction of –1.2%, switching from PI to iGlarLixi led to improvements in the total TRIM-D score, with an increase in least squares (LS) mean of 16.1 (95% CI 14.1, 18.0). Improvements were noted across all five TRIM-D domains, in particular treatment burden (21.4; 95% CI 18.5, 24.3) and diabetes management (23.6; 95% CI 20.7, 26.6). GTEE scores (mean; 95% CI) increased from Week 12 to Week 24 for both patients (2.6; −0.2, 5.5) and physicians (3.6; 0.9, 6.4) (Table).

Conclusions: Switching from PI to iGlarLixi was associated with improvements in patient- and physician-reported outcomes, particularly in reducing treatment burden and enhancing treatment effectiveness, following 24 weeks of treatment.

REAL-WORLD PREVALENCE AND RISK FACTORS OF DIABETES KETOACIDOSIS IN ADULTS WITH TYPE 1 DIABETES (INPHORM, USA)

Alexandria Ratzki-Leewing ¹, Jason Black², Bridget Ryan³, Guangyong Zou³, Stewart Harris²

¹University of Maryland Institute for Health Computing, Department Of Epidemiology And Public Health, University Of Maryland School Of Medicine Baltimore, Baltimore, United States of America, ²Western University, Department Of Family Medicine, London, Canada, ³Western University, Epidemiology And Biostatistics, London, Canada

Background and Aims: Diabetes ketoacidosis (DKA) remains a significant public health challenge for US adults with type 1 diabetes (T1D). Nevertheless, there exists minimal real-world evidence on its risk factors and, thus, targets for preventive intervention. We analyzed iNPHORM data to address this gap.

Methods: Adults (18-90 years old) with T1D were recruited from a US-wide probability-based internet panel. An online screener, baseline, and 12 monthly questionnaires captured data on DKA and participant characteristics. Regularized logistic regression identified factors predictive of DKA for respondents with ≥1 completed follow-up(s).

Results: N=164 were analysed (male: 34.8%; age: 44.8 [SD: 14.0] years; T1D duration: 26 [IQR: 21] years). Pumps or continuous glucose monitoring devices were used by 43.3% and 53.1% of participants, respectively. One in ten (95%CI: 6.6–16.0)% participants reported ≥1 DKA event(s) over follow-up (10.3 [SD: 2.9] months; retention rate: 85.3%). The adjusted odds of DKA increased with antibiotic use (yes vs. no) (OR: 2.98 [95%CI: 1.07–8.27, p-value=0.04]); higher A1C values (OR: 3.34 [95%CI: 0.6–18.48] for 7.1%–8% vs. ≤7%; OR: 5.41 [95%CI: 0.94–31.12] for 8.1%–9% vs. ≤7%; OR: 11.73 [95%CI: 2.19–62.91] for ≥9.1% vs. ≤7%; p-value=0.04); increased number of diabetes complications (excluding DKA) (OR: 2.68 [95%CI: 1.73–4.16], p-value<0.001); increased number of diabetes visits (past 30 days) (OR: 1.52 [95%CI: 0.99–2.32], p-value=0.05); and increased number of past severe hypoglycemia requiring hospital care (OR: 1.89 [95%CI: 1.35–2.66], p-value<0.001).

Conclusions: The observed incidence proportion of DKA in our study aligns with other global investigations. Real-world DKA prevention strategies should focus on T1D adults with above-recommended A1C values, possible infection, and complex health statuses.

DIABETIC KETOACIDOSIS AND SEVERE HYPOGLYCEMIA IN ADULTS WITH TYPE 1 DIABETES AND CHRONIC KIDNEY DISEASE: DATA FROM T1D EXCHANGE REAL-WORLD OBSERVATION STUDY

Tae Won Yi ¹, Lucas Hernandez², Slaven Sikirica³, Saketh Rompicherla⁴, Susan Thapa⁴, Michael Davies³, Francesco Vendrame⁵, David Lam⁶, Nirali Shah⁶, Guillermo Umpierrez², Osagie Ebekozien^4,7, David Cherney¹

¹University of Toronto, Toronto, Canada, ²Emory University, School Of Medicine, Atlanta, United States of America, ³Lexicon Pharmaceuticals, The Woodlands, United States of America, ⁴T1D Exchange, Boston, United States of America, ⁵Unversity of Miami, Miller School Of Medicine, Miami, United States of America, ⁶5. Icahn School of Medicine at Mt. Sinai, New York, United States of America, ⁷University of Mississippi School of Public Health, Jackson, United States of America

Background and Aims: Diabetic ketoacidosis (DKA) and severe hypoglycemia (SH) are serious complications associated with type 1 diabetes (T1D). The impact of comorbid chronic kidney disease (CKD) on these complications remains poorly understood.

Methods: A retrospective cohort analysis using the T1D Exchange Quality Improvement Electronic Health Record database was conducted to determine rates of DKA and SH in people with T1D (PwT1D) and CKD. CKD, was defined by ICD-10 and SNOMED codes or laboratory (eGFR) values in adults (age ≥18) with ≥1 annual clinic visit from 2015 - 2023. DKA and SH diagnoses were confirmed with ICD-10 codes. A propensity score matched (PSM, 1:3 CKD versus no CKD) analysis controlled for age, sex, race/ethnicity, insurance type, continuous glucose monitor use, and insulin pump use.

Results: Of 49,178 PwT1D, 1558 (3.2%) had CKD. PwT1D with CKD were older (56.7y versus 30.4y), with a higher proportion of non-Hispanic Black PwT1D (23% versus 12%), lower rates of CGM and insulin pump use, and longer duration of diabetes vs PwT1D without CKD. Figure 1 demonstrates the unadjusted and PSM rates of DKA and SH in T1D PwT1D with or without CKD. PwT1D and CKD showed a statistically significant increase in SH rates in the unadjusted analysis, and DKA rates in the PSM analysis.

Conclusions: In this PSM analysis, PwT1D and CKD have a small, but statistically significant increase in DKA rates, but similar SH rates compared to those without CKD. There were significant disparities in sociodemographics, CGM, and insulin pump use between groups.