Postbariatric hypoglycemia (PBH) after Roux-en-Y gastric bypass predisposes to health risks and impaired quality of life. Given limited therapeutic options, we evaluated the efficacy of a continuous glucose monitoring (CGM)-guided forecasting algorithm to reduce PBH. In this randomized trial, 59 participants underwent a standardized meal test and were assigned to receive 5 g of glucose either upon the algorithm's predictive alert (intervention, n = 32) or when plasma glucose declined below 3.0 mmol/L (control, n = 27). Hypoglycemia incidence (<3.0 mmol/L) was 31% in the intervention group and 44% in the control group (P = 0.30). Nadir glucose and time spent below 3.9 and 3.0 mmol/L did not differ significantly between groups. Extrapolation based on previously published glucose dose-response data suggests that increasing the preventive glucose dose to 10 g could reduce hypoglycemia incidence to 9%. While a 5 g preventive dose was insufficient, these simulations indicate that CGM-guided hypoglycemia forecasting warrants evaluation as an approach for reducing postprandial hypoglycemia in individuals with PBH.

IntroductionContinuous glucose monitoring (CGM) sensors are increasingly used to identify and manage post-bariatric hypoglycemia (PBH) and to support decision support systems (DSSs) for proactive glucose management. Despite meal timing being key information for these systems, automated, wearable-based meal detection remains an unmet clinical need.MethodsWe present a real-time meal detection algorithm for individuals with PBH that combines CGM and heart rate (HR) signals. The algorithm is a heuristic decision-tree model based on four individualized features extracted from CGM (rate of change, glucose relative excursion, glucose peak value) and HR (peak value). It was developed and tested using a dataset of 40 PBH patients monitored for up to 50 days with a Dexcom G6 CGM and a Garmin Venu Sq smartwatch, and its performance was evaluated in both controlled and free-living conditions, benchmarked against state-of-the-art CGM-only meal detection methods for the PBH population.ResultsThe algorithm achieved 100% recall in the controlled setting and, in free-living conditions, an average precision of 85% and recall of 78%. It also reduced false positives compared with CGM-only algorithms (one every 2.3 days vs. one every 1.3 days).DiscussionEliminating the need for manual meal announcement, the proposed algorithm overcomes a key barrier to fully automated glucose management, reducing patient burden while maintaining reliable detection performance even in unstructured, free-living conditions. These results support the integration of the algorithm into DSSs for PBH and other populations, where timely and accurate meal detection is critical.

Introduction and Objective: Empagliflozin, a highly selective inhibitor of the SGLT2 transporter, lowers blood glucose by inducing glucosuria. However, methods to dynamically quantify this effect in response to individual glycemic profiles are lacking. To address this gap, a mathematical model was developed. Methods: Data from a double-blind, randomized, two-period crossover trial (NCT05057819) were used. Twenty-two non-diabetic adults with postbariatric hypoglycaemia following RYGB surgery (age=48 [34, 51] y, median [IQR]; BMI= 25.4 [23.0, 27.0] kg/m2; GMI= 5.89 [5.74, 5.99] %; eGFR≥90 mL/min/1.73m²) completed two 20 day-trial periods, receiving once daily empagliflozin 25mg (E) first and placebo (P) second, or vice versa. Glycaemic responses were evaluated via continuous glucose monitoring (CGM) throughout the study, a 2h solid mixed meal test (MMT), and a 24h urinary glucose excretion (UGE) measurement via a 24h urine collection prior to the MMT. The oral glucose minimal model was adapted to incorporate a term describing the rate of UGE, assumed to be proportional to glucose concentration exceeding a threshold (Th) for glucose spillage into the urine. Model parameters, including Th and insulin sensitivity (IS), were estimated from plasma glucose and UGE data, with plasma insulin and CGM used as inputs. Results: Fasting glucose levels were lower in E vs P (88.3 [84.7, 91.9] vs. 92.8 [86.5, 96.4] mg/dl, p=0.002), while incremental area under the glucose curve was similar (5.3 [3.1, 7.5] vs. 5.0 [3.0, 8.3] g/dl·min). The model predicted a reduction of Th in E vs P (72.1 [55.8, 85.3] vs. 173.1 [156.4, 179.8] mg/dl, p<0.001), and similar IS. In E, daily glucosuria in the fasting state was 14.6 [8.32, 22.1] g/day. Conclusion: The proposed model provides a quantitative, individualized estimate of empagliflozin's effect on UGE, which mechanistically explains the observed reduction in fasting glucose and can be used to dynamically assess treatment response over time. Disclosure A. Brunasso: None. C. Dalla Man: Other - Webinar provider; Ended; Sanofi. Other - Joint research project; Current; Sanofi-Aventis Deutschland GmbH. D. Herzig: None. L. Bally: Research Support; Current; Dexcom, Inc., Ypsomed AG, Novo Nordisk, Eli Lilly and Company. Advisory Panel; Ended; Ypsomed AG, Novo Nordisk, Eli Lilly and Company. M. Schiavon: Research Support; Current; Sanofi. Funding Swiss National Science Foundation (PCEGP3_186978); MIUR (Italian Minister for Education, 39-411-19-DOT1319501-6721) (Brunasso)

Background:Dietary counseling is an essential complement to the growing use of glucagon-like peptide-1 receptor agonists (GLP-1 RAs) to achieve sustainable weight loss. Diverse patient needs, the demand for ongoing support, and limited resources underscore the potential of using digital support tools. Objective:This study aimed to evaluate the usability, treatment satisfaction, and weight effects of a personalized nutrition advisor (PNA), a prototype decision-support tool designed to facilitate personalized weight management counseling. Methods:We assessed usability, treatment satisfaction, and weight effects of a web-based PNA prototype dashboard designed to help dietitians deliver personalized remote dietary counseling in overweight adults (BMI ≥28 kg/m² at pharmacotherapy initiation) who had achieved ≥5% weight loss with GLP-1 RA therapy over 12-weeks. The PNA dashboard generated tailored nutritional recommendations to help participants achieve personalized weight loss goals based on an energy balance model that estimated daily caloric intake and nutritional quality using frequent body weight measurements, dietary logs, and physical activity data from a lifestyle-tracking app. To evaluate the PNA's usability, we used a convergent parallel mixed-methods approach, combining a validated quantitative usability survey and qualitative focus-group interviews and feedback questionnaires with both dietitians and patients, alongside patient-reported satisfaction metrics for PNA-guided consultations (treatment satisfaction, general self-efficacy, and perceived health impact). Thematic analysis with a hybrid deductive-inductive methodology was applied to qualitative data, using 4 a priori themes derived from this study's conceptual framework regarding prior expectations, perceived utility and satisfaction, usability challenges, and suggestions for clinical integration. Percent weight loss during the intervention was assessed as an exploratory outcome. Results:Among 78 participants (58/78, 74% women), on GLP-1 RA (mean weight loss of -13.3%, SD 6.7%, after 10, SD 5, months of treatment), between 3 and 5 remote nutrition consultations were delivered with PNA support. After the 3-month intervention period, weight loss was sustained in all participants (-0.2%, SD 2.5%), with 21% (16/78) suspending the GLP-1 RA due to drug shortages. Dietitians rated the PNA's usability as moderate (62%) on the Healthcare Systems Usability Scale. Patients evaluated the perceived health impact of the PNA-assisted consultations at 19.1 (SD 3.5) of 25 on the user version of the Mobile Application Rating Scale. No significant differences were observed in patient-reported treatment satisfaction (P=.78) or self-efficacy (P=.57) before and after PNA-assisted care. Qualitative analysis identified convergence with quantitative findings across themes of usability, treatment satisfaction, and clinical integration, with divergence observed for self-efficacy and tracking motivation. Both patients and dietitians found the PNA valuable for generating actionable nutritional insights. Conclusions:While effective implementation of tools such as the PNA prototype demands dietitian training and workflow integration, they offer scalable, personalized nutrition or lifestyle guidance-as a therapeutic decision-support tool within weight loss interventions, with or without adjunct GLP-1 RA pharmacotherapy.

OBJECTIVE:Although continuous glucose monitors (CGMs) are increasingly used to detect and manage postbariatric hypoglycemia (PBH) and associated glucose variability, data on their accuracy in this population remain scarce. RESEARCH DESIGN AND METHODS:We retrospectively assessed the accuracy of the Dexcom G6 CGM system in adults with PBH after Roux-en-Y gastric bypass (RYGB) surgery (n = 70). Glucose excursions were induced using a standardized solid mixed-meal test, and reference blood glucose (BG) values were obtained through repeated venous whole-blood sampling. CGM accuracy was analyzed separately during stable and dynamic postmeal glucose periods, with dynamic phases stratified according to magnitude and direction of the rate of change (RoC). We further estimated the lag time for each sensor and examined predictive factors affecting CGM accuracy. RESULTS:A total of 1,822 CGM-BG pairs obtained with 70 individuals were included in the analysis. Mean absolute relative differences at stable and dynamic levels were 9.6% and 16.4%, respectively. After the meal test, 67.6% of pairs had CGM values within 15%, or 15 mg/dL of the reference BG; 78.0% within 20%, or 20 mg/dL; and 90.8% within 30%, or 30 mg/dL. Performance was worse at rapid plasma glucose decline (>1.5 mg/dL/min), and CGM values at the time of plasma glucose nadir were systematically higher (bias, 8.2 mg/dL). Plasma-interstitium time delay was estimated at 9.8 min. No participant or sensor characteristic had a significant impact on CGM accuracy. CONCLUSIONS:Meal-induced glucose dynamics, particularly rapid declines, challenge CGM accuracy in people with PBH and must be carefully considered when diagnosing or managing the condition.

Subcutaneous insulin delivery in individuals with insulin-deficient type 1 diabetes bypasses the portal circulation, disrupting the physiological porto-systemic insulin gradient and affecting postprandial hepatic glucose regulation. However, direct, non-invasive measurement of these liver-specific dynamics and their deviation from normal physiology in individuals with type 1 diabetes is challenging. To address this, we integrated metabolic imaging with whole-body tracer dilution to map postprandial glucose metabolism in both the liver and systemically in adults with type 1 diabetes and healthy control individuals. In this cross-sectional study, ten adults with type 1 diabetes and ten healthy control individuals with similar age, BMI and gender distributions were enrolled. After an overnight fast, participants ingested 60 g [6,6′-2H2]-glucose (D-Glc); subcutaneous insulin was administered to type 1 diabetes participants according to their carbohydrate-to-insulin ratio. Interleaved deuterium metabolic imaging (DMI) and 13C-magnetic resonance spectroscopy (13C-MRS) at 7 T were performed from pre-ingestion to 150 min post-ingestion to quantify hepatic D-Glc and glycogen concentrations. Blood samples were collected to measure plasma glucose, insulin and glucagon. Postprandial glucose–insulin dynamics were quantified using the single tracer oral minimal model, accounting for non-steady-state insulin exposure. At baseline, individuals with type 1 diabetes had significantly higher plasma glucose concentrations than control individuals (10.7±2.3 and 5.2±0.4 mmol/l, respectively; p<0.001), while preprandial glycogen levels did not differ significantly. Following D-Glc administration, hepatic D-Glc increased more markedly in the individuals with type 1 diabetes compared with the control group (peak values 4.7±2.0 and 3.0±0.8 mmol/l, respectively; p=0.02). In the postprandial period, glycogen levels did not significantly rise at 150 min in type 1 diabetes, whereas a clear increase was observed in control individuals (iAUC0–180=2.4 mol/l × min). Despite similar systemic insulin exposure and no significant differences in postprandial glucagon concentrations between groups, individuals with type 1 diabetes demonstrated significantly reduced suppression of endogenous glucose production (p=0.001) but similar insulin-dependent glucose disposal. Hierarchical clustering identified two distinct type 1 diabetes subgroups: Subgroup 1 exhibited a steeper increase in both hepatic and systemic D-Glc profiles, while subgroup 2 showed a divergent D-Glc trajectory and net glycogen depletion relative to accumulation in subgroup 1 (iAUC0–180=−3.0 vs 2.5 mol/l × min, p=0.04), despite no overt clinical differences between subgroups. By integrating DMI/13C-MRS liver imaging with systemic stable-isotope modelling, this comparative study demonstrates significantly altered hepatic glucose metabolism in adults with well-managed type 1 diabetes vs control individuals, together with substantial phenotypic heterogeneity within the type 1 diabetes cohort. These findings highlight the potential of non-invasive metabolic phenotyping to resolve metabolic alterations and inter-individual variation in type 1 diabetes, which are essential steps towards the provision of precision medicine.

Background:Cardiovascular disease (CVD) prediction models recommended by guidelines are developed using different populations, predictors, and outcome definitions. The implications of this heterogeneity for risk estimation are unclear, and direct comparisons remain limited. Objectives:Head-to-head comparison of the performance and transportability of three guideline-endorsed CVD risk prediction models, focusing on their sex-specific performance. Methods:We evaluated models recommended by the American Heart Association (PREVENT), European Society of Cardiology (SCORE2), and the National Institute for Health and Care Excellence (QRISK3). Risk of bias was assessed using the PROBAST tool. External validation was performed using the UK Biobank (UKBB) in a primary analysis including all participants with complete data for all models, enabling direct comparison, and in a secondary analysis applying each model to participants meeting its original eligibility criteria. Model performance was assessed using Brier scores, Area Under the Receiver Operating Characteristic Curve (AUC), and calibration across original and alternative outcome definitions, stratified by sex. Results:The PREVENT, SCORE2, and QRISK3 models varied substantially in terms of predictors, populations, and outcome definitions. We used data from 502,157 UKBB participants for the external validation in the primary analysis. Overall predictive performance (discrimination & calibration), as measured by Brier scores, was generally better in females. The AUC (95% CI) ranged from 0.7092 (0.7090-0.7094) to 0.7468 (0.7465-0.7471) for female and 0.6813 (0.6812-0.6814) to 0.6946 (0.6945-0.6946) for male populations. Calibration was suboptimal, particularly for older individuals, with systematic overestimation of risk. The models showed consistent performance when applied to different outcomes. All models were at high risk of bias. Conclusion:Despite heterogeneity in populations, predictors, and outcome definitions, PREVENT, SCORE2, and QRISK3 showed similar performance in the UKBB. Future studies should focus on prospective and standardized definitions and assessment of candidate predictors and outcomes.

Despite extensive research on liver metabolism, mathematical models describing hepatic glucose kinetics are currently limited due to the lack of organ-level data. Here, we propose a model of postprandial hepatic glucose kinetics exploiting liver deuterium metabolic imaging (DMI) data combined with plasma isotope dilution analysis in humans. We used data from 10 individuals who had previously undergone Roux-en-Y gastric bypass surgery (RYGB) and 10 healthy controls (HCs). The experimental setting included a labeled oral glucose tolerance test comprising 60 g of [6,6'-2H2]-glucose in combination with liver DMI at 7 T. The hepatic glucose tracer signal was frequently quantified over 150 min, whereas peripheral plasma insulin and glucose tracer concentrations were measured in venous blood. The model was able to describe both liver and peripheral glucose tracer data well and provided estimates of postprandial glucose appearance and disposal in both the liver and the systemic circulation. The model predicted that almost all the ingested glucose had appeared in the liver in RYGB, but not in HC (89.0% vs. 64.0%, P = 0.008) after 150 min, whereas total hepatic disposal (RYGB = 26.4% vs. HC = 29.7%) and first-pass extraction (RYGB = 10.7% vs. HC = 11.4%) were similar between populations. The fraction of glucose eliminated in the periphery was greater in RYGB (49.9% vs. 25.3%, P = 0.003). Finally, no differences were observed in hepatic blood flow and GLUT2 transport rates. Although further studies are needed to validate and extend the model to include endogenous glucose production and disposal, it can be used to quantify parameters, and possibly reveal defects, of hepatic glucose handling.NEW & NOTEWORTHY The proposed hepatic model allows, for the first time, to describe postprandial liver glucose tracer kinetics in humans, allowing to estimate exogenous glucose appearance and disposal in the liver, as well as glucose transport and hepatic blood flow rate. The model may become a useful tool in clinical research by supporting the identification of metabolic defects at the hepatic level without requiring invasive procedures.

Background:Image-based automated meal analysis using smartphones has the potential to facilitate meal management in type 1 diabetes. We evaluated the glycaemic efficacy of SNAQ-an image-based automated meal analysis app-in adults using hybrid automated insulin delivery (AID) systems requiring carbohydrate entry for prandial insulin dosing. Methods:In this single-centre trial (NCT05671679) adults with type 1 diabetes on AID therapy were randomly assigned to using SNAQ-a commercial mobile app recognizing and quantifying food from images for meal management support-or continuing their usual meal management (control group) for 3 weeks. The primary endpoint was the change in the %time in range (TIR, 3.9-10.0 mmol/L). Following the first three weeks, SNAQ was also provided to the control group for evaluating the sustainability of benefits and usage across all participants. Findings:Twenty-two participants were randomized between March 14 and November 23 2023 to using SNAQ and 22 to control. At baseline, TIR was 75.4 ± 13.7% and 74.3 ± 12.7% in the intervention and control group, respectively. After three weeks, the baseline-adjusted difference in TIR between SNAQ (used 1.6 ± 0.8 per day) and control was 6.6 percentage points in favour of SNAQ (95% CI 2.9 to 10.3, P < 0.001). SNAQ further improved mean glucose (-0.54 mmol/L, CI -0.9 to -0.2, P = 0.004) and time above range (-6.3%, CI -10 to -2.7, P = 0.001). Time below range, total daily insulin dose, bolus frequency, nor carbohydrate entered into the pump did not significantly differ between groups. Post-discontinuation, the glycaemic benefits of SNAQ were not sustained. No study-related serious adverse events occurred. Interpretation:Short-term use of the automated meal analysis app SNAQ improved glucose control in adults with type 1 diabetes treated with AID. Funding:The study was supported by the EFSD/EUDF Digitalisation on Diabetes Care Research Grant and by the Diabetes Center Berne.

BackgroundPost bariatric hypoglycaemic (PBH) is a late complication of weight loss surgery, characterised by critically low blood glucose levels following meal-induced glycaemic excursions. The disabling consequences of PBH underline the need for the development of a decision support system (DSS) that can warn individuals about upcoming PBH events, thus enabling preventive actions to avoid impending episodes. In view of this, we developed various algorithms based on linear and deep learning models to forecast PBH episodes in the short-term.MethodsWe leveraged a dataset obtained from 50 patients with PBH after Roux-en-Y gastric bypass, monitored for up to 50 days under unrestricted real-life conditions. Algorithms' performance was assessed by measuring Precision, Recall, F1-score, False-alarms-per-day and Time Gain (TG).ResultsThe run-to-run forecasting algorithm based on recursive autoregressive model (rAR) outperformed the other techniques, achieving Precision of 64.38%, Recall of 84.43%, F1-score of 73.06%, a median TG of 10 min and 1 false alarm every 6 days. More complex deep learning models demonstrated similar median TG but inferior forecasting capabilities with F1-score ranging from 54.88% to 64.10%.ConclusionsReal-time forecasting of PBH events using CGM data as a single input imposes high demands on various types of prediction algorithms, with CGM data noise and rapid postprandial glucose dynamics representing the key challenges. In this study, the run-to-run rAR yielded most satisfactory results with accurate PBH event predictive capacity and few false alarms, thereby indicating potential for the development of DSS for people with PBH.

Post-bariatric hypoglycemia (PBH) is a metabolic complication of individuals with obesity who have undergone bariatric surgery, characterized by rapid glycemic excursions followed by hypoglycemic events usually occurring 1-3 h post-meal. Without an approved pharmacotherapy, dietary modifications are essential for managing PBH, with continuous glucose monitoring (CGM) devices emerging as crucial tools for capturing postprandial glucose responses that can guide intervention strategies to prevent PBH. The effectiveness of such interventions is based on the availability of rich datasets, containing both CGM and meal data. However, meal information is often incomplete, being its manual recording burdensome and prone to user-related errors. In response, we proposed a template match algorithm (TMA) for the retrospective identification of unreported meals using CGM data only. TMA relies on a similarity score calculated between a post-prandial glycemic curve template and the glycemic trace of interest. Our study demonstrates promising results: TMA correctly identifies 1237 out of 1340 meals, generating 208 false positives within a dataset of 20 PBH subjects monitored in free-living conditions for nearly 50 days, yielding a median F1-score of 0.90. The effectiveness of TMA enables its use to enhance data quality in long-term studies involving PBH patients, facilitating the development of new approaches to manage PBH.

Introduction Postbariatric hypoglycaemia (PBH) is a complex medical condition with a significant impact on patients’ quality of life. The underlying mechanisms remain to be elucidated. We have shown that food ingestion increases IL-1β and subsequently stimulates insulin secretion. We therefore hypothesised that overactivation of the IL-1β pathway could lead to PBH by promoting excessive insulin secretion after a meal. In a proof-of-concept study, we have shown that acute treatment with the IL-1 receptor antagonist anakinra can attenuate PBH after a single liquid mixed meal. This study aims to validate this therapeutic approach over a longer period of time using the long-acting anti-IL-1β antibody canakinumab.Methods and analysis In this prospective, randomised, double-blind, placebo-controlled, multicentre trial, we plan to enrol 62 adult patients after bariatric surgery with frequent, postprandial hypoglycaemia (ie, <3.0 mmol/L and at least five hypoglycaemic episodes per week). Eligible subjects will be randomised to receive either single-dose 150 mg canakinumab (Ilaris, Novartis) subcutaneously (s.c.) or matched placebo (1.0 mL physiologic saline). For 28 days, patients are required to wear a blinded continuous glucose monitoring device (CGMS, Dexcom G6) and use a diary to track their hypoglycaemic episodes. Primary outcomes include health-related quality of life, measured by the SF-36, as well as postprandial hypoglycaemic events (glucose <3.0 mmol/L). A significant improvement in any one of these outcomes will be considered sufficient to demonstrate the clinical superiority of canakinumab over placebo. Secondary outcomes include patient-oriented measures such as postprandial hypoglycaemic symptoms, hypoglycaemia unawareness, fear of hypoglycaemia, as well as metabolic measures and safety assessments.Ethics and dissemination The trial was approved by the Cantonal Ethics Committee ‘Ethikkommission Nordwest- und Zentralschweiz’ in January 2022 (#2021–02325), as well as by Swissmedic in April 2022 (#701280). Current, approved protocol version 1.3 of 28.03.2023. The study is actively recruiting. Results will be published in a relevant scientific journal and communicated to participants and relevant institutions through dissemination activities. Individual data are accessible on request.Trial registration The study is registered with the www.clinicaltrials.gov registry (NCT05401578) and the Swiss National Clinical Trials Portal (SNCTP) on www.kofam.ch (SNCTP000004838).

Type 2 diabetes is a wide-spread chronic condition in which blood glucose and body weight management constitute essential therapeutic targets. Emerging technologies have the potential to aid complex therapeutic pharmacotherapy choices that are optimally tailored to individual needs. Here we propose an artificial intelligence combining guidelines with clinical features and continuous glucose monitoring (CGM) to optimize therapeutic decision-making. Therapeutic guidelines are first encoded using a rule-based model and trained on a neural network. Relying on real world evidence outcomes of a specialist outpatient clinic, transfer learning is used to optimize for glucose-lowering therapies that led to successful treatment outcomes defined as an absolute 0.3% reduction in glycated (HbA1c) over 6.5% without increasing body weight for a BMI over 28. Recommendations that deviate from guidelines are described with Shapley values and tested in digital twins for statistical significance. Four CGM-derived glucose-insulin response dynamic factors serve as additional biomarkers. Dual glycemic & weight targets were achieved in actual clinical practice in 51% cases, increasing to 54% when clinical guidelines were followed. Selecting outcomes in the test set that follow individualized recommendations, this increases further to 56% when using only phenotypic markers and to 64% when adding CGM-derived dynamics factors. Tested on the limited patient number available, our findings show that AI can outperform guidelines in complex type 2 diabetes cases by integrating multiple data sources, drawing on experiential clinical insights, and selecting treatments most effective for each patient’s glucose and weight control. A neural network is first trained on guidelines and subsequently on real-world evidence outcomes, performing dual glycaemic/weight optimization to improve the management of type 2 diabetes with/out gluco-dynamic parameters extracted from Continuous Glucose Monitoring.

Background Post-bariatric hypoglycemia (PBH) is a severe and often overlooked complication of bariatric surgery (BS), characterized by dangerously low blood glucose levels after meals, particularly those high in carbohydrates. Unlike in Type 1 and Type 2 diabetes (T1D, T2D), where decision support systems (DSS) and continuous glucose monitoring (CGM) tools aid blood glucose management, no dedicated DSS exists for PBH. This leaves individuals vulnerable to recurrent, unpredictable hypoglycemia, posing significant health risks. To address this gap, we propose Glu4, an open-source software package designed to predict and notify users of impending PBH events using CGM data. Methods Glu4 employs a two-step approach to predictPBH. A run-to-run algorithm forecasts future glucose levels using past CGM data, identifying potential hypoglycemic events 30 min in advance. An intelligent alarm system alerts users when glucose levels are predicted to drop below a critical threshold, prompting preventive action. A pilot study involving three PBH patients collected real-time glucose data to validate the system’s predictive performance. Results The pilot study demonstrated that Glu4 reliably predicted impending hypoglycemia in all participants, providing timely alerts 30 min before glucose drops. The system showed a high specificity, with no false alarms being triggered during the monitoring period. The proactive notifications enabled participants to manage their glucose levels more effectively by taking preventive actions such as consuming rescue carbohydrates before the onset of severe hypoglycemia. Conclusions Glu4 represents a promising tool for managing PBH, leveraging CGM data to deliver accurate, timely alerts that enable proactive intervention. By improving safety and quality of life for individuals with PBH, Glu4 addresses a critical unmet need. Future work will focus on enhancing system capabilities and conducting larger-scale studies to validate its effectiveness and refine its usability for clinical adoption.