Journal of Diabetes Science and Technology最新文献

Objective: Evidence from long-term, real-world use of blood glucose (BG) monitoring technologies is sparse. We investigated whether using a diabetes app with connected meters could support durable diabetes management improvements in people with type 2 diabetes (T2D) over 5-years.

Methods: Anonymized glucose and app analytics from 501 people with T2D were extracted from our server. The first 14 days using the app were compared with the last 14 days of each consecutive year for 5 years, using paired within-subject differences. Subjects had ≥365 BG readings per year.

Results: People with T2D improved BG readings in range (RIR, 70-180 mg/dL) by +6.9 percentage points (%pts, 74.6% to 81.5%) and readings in tight range (RITR, 70-140 mg/dL) by +7.8%pts (49.2% to 57.0%) at year 1. Year 1 improvements in RIR and RITR remained evident at year 5 (+7.5%pts and +7.7%pts, respectively). Reductions in hyperglycemic readings (>180 and >250 mg/dL) explained the improvements in RIR and RITR over the 5-years. Mean BG reduced by -9.1 mg/dL at year 1 (150.2 to 141.1 mg/dL) and this was sustained at year 5 (-10.6 mg/dL, 150.2 to 139.6 mg/dL). Subjects performed BG checks at a consistent level, equivalent to 1.8 to 2.1 checks per day, over 5 years. All these glycemic changes were significant (p<0.001). Higher app engagement (>4 app sessions per week) effected better diabetes management.

Conclusion: Real-world follow-up of people with type 2 diabetes using a diabetes app with connected meters found improvements in glycemia were durable over 5-years.

Aim: We aimed to develop a calculator to determine the probability of having HNF1A-MODY (hepatocyte nuclear factor 1 alpha-maturity-onset diabetes of the young) or HNF4A (hepatocyte nuclear factor 4 alpha)-MODY (the commonest forms of MODY) in Asian Indians using clinical and biochemical criteria.

Methods: We extracted data on individuals with young-onset diabetes aged <30 years (n = 29 191) from electronic records. Genetically confirmed HNF1A- and HNF4A-MODY (n = 55) were selected along with 1000 individuals each of type 1 diabetes (T1D) and type 2 diabetes (T2D). These data sets were used to develop a classification model using logistic regression. The model's performance was evaluated using receiver operating characteristic (ROC) curves in an internal data set and validated in an external data set.

Results: Eight predictive models were constructed, beginning with a basic model that included variables, such as age at diagnosis, body mass index (BMI), parental history, and glycated hemoglobin (HbA1c) (models 1 and 5). High-density lipoprotein (HDL) cholesterol was added in models 2 and 6, stimulated C-peptide in models 3 and 7, and all predictors were combined in models 4 and 8. Models 1 to 4, designed to distinguish MODY from T1D, achieved an ROC-area under the curve (AUC) value ranging from 0.884 to 0.957, while models 5 to 8, aimed at differentiating MODY from T2D, achieved an ROC-AUC value ranging from 0.914 to 0.936. All models demonstrated excellent performance in internal validation, with high five-fold cross-validation c-statistics. An online calculator using these models estimates MODY probability that is accessible at https://mdrf-t1d-calculator.shinyapps.io/MODY/.

Conclusion: We developed an ethnicity-specific calculator to help identify individuals with possible HNF1A-MODY or HNF4A-MODY in Asian Indians. This user-friendly, web-based tool would be helpful to select candidates for genetic testing in this population.

Introduction: Initiation of automated insulin delivery (AID) systems involves a complex multistep process, starting with the shared decision of patient and provider to start an AID, followed by prescription, AID training, and AID start. We aim to assess time taken from decision to start an AID system to actual initiation of AID therapy, termed time to AID (TT-AID) and identify factors that influence the process.

Methods: This retrospective study included insulin pump naive youth with type 1 diabetes, who decided to initiate an AID system after May 2022, at the Johns Hopkins Diabetes Center. Electronic medical records and device portals were reviewed to collect demographics and AID details. Time-to-event analysis was performed.

Results: Participants included 270 youth with T1D (median age = 12.4, 57% male, 58.9% non-Hispanic white, 4.4% Hispanic, and median diabetes duration = 0.3 years). Median TT-AID is 43.5 days, with longest duration observed between prescription and pre-AID training (median = 37.5 days). Time to AID increased significantly for participants with a diabetes duration greater than one year (40 days vs 56 days; P = .0002) and higher area deprivation index (hazard ratio [HR] = 0.95; P = .023). There were no significant differences in TT-AID based on insurance type or type of AID system.

Conclusion: The process of starting an AID system can be lengthy, with factors such as longer diabetes duration and high area deprivation being associated with delays in the process. Future interventions could address these factors by encouraging early AID system discussions and providing additional social support to improve the efficiency of the AID initiation process.

Background: Suboptimal inpatient glycemia is associated with adverse outcomes, including infection, length of stay, and hospitalization costs. Interventions to improve inpatient glycemia may benefit from standardization of in-hospital glycemic measurement and reporting."Glucometrics," as coined by Goldberg et al (2006), proposes models and metrics that allow quantitative inpatient glycemic data analysis. This systematic review investigates the actual use of "glucometric" terminology and its derivations since conception.

Methods: Original research articles on "glucometrics" and its derivations in inpatient contexts, published between 2006 and 2023, were searched in five databases. Studies were screened and extracted through PRISMA-compliant review software (Covidence®) and systematically reviewed.

Results: Of 767 studies identified, 44 were included for final review. Study settings included non-critical care wards (n=19), critical care (n=6), and both (n=13). Of the Goldberg models, "patient-day" was most used (n=33). Most studies (n=30) referred to "glucometrics" per the original description. An increase in the introduction of new metrics (e.g., time-weighted averages, adverse glycemic days, and glucose excursions) was seen over the study period, as well as an increase in the use of "glucometric" to refer to glycemic measurement/reporting in general.Significant variation in thresholds defining hyperglycemia/hypoglycemia existed between studies, where hyperglycemia ranged between 140 and 432 mg/dL (most commonly 300 mg/dL), while the hypoglycemia ranged between 40 and 70 mg/dL (most commonly 70 mg/dL).

Conclusion: This systematic review provides insights into contemporary use of glucometric terminology, highlighting the lack of consensus on a standardized approach toward analyzing inpatient glycemia, and the need for glucometric harmonization to improve inpatient glycemia and diabetes care.

Background: Diabetes is a chronic condition requiring long-term management, and continuous health education is vital for improving disease awareness and self-management. Large language models (LLMs), advanced artificial intelligence systems trained on large text data sets, have shown promise in generating diabetes-related educational materials. While LLMs can generate accurate and readable content, most studies focus on general education based on guidelines, rather than tailoring content to individual patients' clinical profiles. This study addresses these gaps by comparing the performance of three major LLMs (ChatGPT-4o, Doubao 1.5, and DeepSeek R1) in generating health education materials for discharged patients with diabetes.

Methods: Ten de-identified medical records of discharged patients with diabetes were uploaded to the LLMs. Each model generated health education materials based on these records. Experienced diabetes nursing experts evaluated the quality of the generated materials.

Results: The comprehensibility scores pass rates for all models were above 70%, with DeepSeek R1 performing the best (P < .01). The actionability scores pass rates were below 70% for all models, with no significant differences (P > .01). Accuracy scores for all models were ≥98%, and there were no significant differences in accuracy (P > .01). Similarly, no significant differences were observed in personalization and effectiveness scores (P > .01). DeepSeek R1 achieved the highest safety score, while Doubao 1.5 had the lowest safety score (P < .01).

Conclusion: While ChatGPT-4o, Doubao 1.5, and DeepSeek R1 generate accurate and comprehensible materials, concerns remain regarding their actionability and safety. These findings suggest that LLMs should be used as auxiliary tools in diabetes education, requiring further refinement for personalized and actionable content.

Background: Discrepancies between HbA1c and glucose management indicator (GMI) may reflect individual variations in glycation rate, independent of mean glycemia, and could influence complication risk stratification in type 1 diabetes (T1D). We evaluated the phenotype of individuals with T1D using continuous glucose monitoring (CGM), identified as high glycators based on HbA1c/_updatedGMI ratio, and assessed retrospectively their risk of diabetic retinopathy (DR) and the time to DR diagnosis. The secondary aim was to identify clinical correlates of high glycation.

Primary outcome: time to first diagnosis of DR.

Secondary outcomes: clinical factors associated with high glycation.

Methods: A retrospective study of 411 individuals with T1D using CGM and concurrent HbA1c values. Patients with conditions affecting red blood cell (RBC) lifespan were excluded. Participants were divided into 3 subgroups based on current HbA1c/_updatedGMI ratio ≤0.95 (low glycators), >0.95 and <1.05 (concordant glycators), and ≥1.05 (high glycators). Time to diagnosis of DR was retrieved.

Results: High glycation is associated with shorter time to first diagnosis of DR (adjusted hazard ratio 1.60). Non-HDL-C, RBC indices, and metformin were associated with high glycation.

Conclusion: Among individuals with T1D, an HbA1c/_updatedGMI ratio ≥1.05 is associated with higher odds of DR. Non-HDL-C and RBC indices are correlates of high glycation. These results underscore the relevance of HbA1c and _updatedGMI discrepancy in cardiometabolic risk assessment, but cutoffs remain to be set.

A panel of experts in the use of continuous glucose monitoring (CGM) data in the treatment of diabetes met in Burlingame, California on October 27, 2025 to discuss the utility of the glycemia risk index (GRI) for clinical care research and population health management. The GRI composite metric is a single number (on a 0-100 percentile scale-lower is better) based on an expert-determined weighting of the seven individual components in the existing ambulatory glucose profile (AGP). The GRI describes the quality of glycemia based on glucose values collected in a 14-day CGM tracing, thus providing additional insights into CGM profiles beyond the AGP. During the meeting, the mathematical derivation of the GRI metric was presented along with its use for adult and pediatric individuals with diabetes and cancer who require medications that can adversely affect the glucose concentration. Examples where the GRI provided useful insights into the quality of CGM tracings were also discussed by the expert panel. In addition, a new smartphone application, the GRI Calculator, was presented. This app calculates the GRI of a CGM tracing and provides visualization of sequential CGM tracings for a specific individual. The GRI provides a reference measurement for the accuracy of artificial intelligence (AI) models assigning levels of glycemic quality to CGM tracings intended to match the assessments of clinicians. The GRI is now part of the data visualization panel for the Integration of Connected Diabetes Device Data into the Electronic Health Record (iCoDE-2) project, which standardizes both CGM and insulin dosing data. Further exploration of the potential value of the GRI for non-insulin users needs to be undertaken. The panel unanimously recommended that CGM manufacturers and developers of data visualization software for CGMs add the GRI to their data platforms for insulin users.