Journal of Diabetes Science and Technology

This study evaluated the impact of continuous glucose monitoring (CGM) on health outcomes and workplace absenteeism among people living with type 2 diabetes (T2D) who are not using insulin.

This was a pre–post observational study using real-world data from a large employer health plan. Glycated hemoglobin (HbA1c), body mass index (BMI), and sick days were measured 360 days before (baseline period) and 360 days after (follow-up period) CGM initiation. Persistence was defined as refilling CGM supplies at least once every 90 days.

In total, 71 patients were included (mean age, 52.3 years; 56.3% female); 33 (46.5%) were persistent to CGM. Overall, mean HbA1c levels decreased from 7.6% to 7.1% after the initiation of CGM (mean reduction, 0.52%; N = 55; P = .042), while mean BMI decreased by 0.95 kg/m² (N = 34; P = .075). Compared with the nonpersistent group, persistent patients achieved larger mean reductions in HbA1c levels (−0.89% vs. −0.17%; mean difference, 0.72%; P = .152) and BMI (−2.26 kg/m² vs. +0.37 kg/m²; mean difference, 2.62 kg/m²; P = .009). Absenteeism decreased in the persistent group but increased in the nonpersistent cohort (−8.41 vs. +6.95 mean sick days; mean difference, 15.36 days, P = .002).

In this population, CGM was associated with improvements in clinical outcomes and absenteeism, although the economic implications require further evaluation. BMI and absenteeism were statistically significantly improved among persistent patients, compared with those of the nonpersistent cohort. Employers could benefit from broader coverage of CGM in this population.

Continuous glucose monitoring (CGM) has come a long way and is standard for patients with type 1 diabetes and many with type 2 diabetes. Several attempts to establish noninvasive glucose monitoring, that is, measuring glucose without puncturing the skin, have not been successful yet.

A different approach is the monitoring of volatile organic compounds in breath.

This addresses a number of limitations of current invasive glucose monitoring techniques. This should enhance compliance, adherence, clinical outcomes, and quality of life. It might also reduce costs associated with CGM.

A recent publication in this journal indicates the clinical value of this approach by presenting data from a clinical study. The respective pros and cons will be discussed briefly.

Quantifying the effect of meal composition (MC) on postprandial glucose excursions would allow optimizing insulin therapy, accounting for fat and protein that can affect gastric retention (GR), glucose rate of appearance (R_a), and insulin sensitivity (S_I). Such variables can be estimated from continuous glucose monitor (CGM) and continuous subcutaneous insulin infusion (CSII) data using the Minimally-Invasive Oral Minimal Model (MI-OMM). In this work, we aim to quantify the effect of MC on those variables by applying the MI-OMM on a data set of prandial CGM and CSII profiles where MC information was available.

A total of 120 individuals with type 1 diabetes (age = 15.5 ± 11.5 years, weight = 51.3 ± 28.0 kg) were monitored under free-living conditions while using CGM and CSII, and MC was carefully recorded. We extracted 353 CGM and CSII traces using predefined criteria and classified them into low or high fat content and low or high protein content. Finally, the MI-OMM was used to estimate GR, R_a, and S_I in each meal.

MI-OMM was able to fit CGM profiles and provided precise and physiologically plausible parameter estimates. Comparison among different classes of meals showed that a high content of fat and protein in the meal significantly slowed both GR (P < .01) and R_a (P < .01), and reduced S_I (P < .05).

In this work, the effect of MC on postprandial glucose excursion was quantified in real-life conditions with the help of a model-based methodology. These results are usable for redesigning current insulin therapies, accounting for the presence of fat and protein in meals.

Advances in diabetes technologies have transformed the management of diabetes in recent decades. The widespread adoption of continuous glucose monitoring (CGM) systems has revolutionized diabetes care in Saudi Arabia. Despite rapid technological integration, the clinical landscape lacks unified, Gulf-specific protocols tailored for diabetes educators. Structured educational diabetes technology programs are essential for providing diabetes educators with the knowledge and confidence to support people with diabetes who are using different types of diabetes technologies.

An expert panel of 10 diabetes educators from across Saudi Arabia was formed. Draft statements on an optimal educational pathway for diabetes educators for training people with diabetes on CGM in Saudi Arabia were generated using insights gathered from a survey and a face-to-face meeting. Consensus was reached using the Delphi methodology.

Consensus was reached on all 7 consensus statements. A structured educational pathway including face-to-face training and interactive e-learning modules, a mentorship program, detailed clinical evidence for CGM, case-based scenarios, hands-on experience, and in-field training was recommended by the experts. Overcoming barriers such as institutional support or funding, availability of training programs, and adequate time for participation are instrumental to the success of a structured educational pathway for educators.

This consensus will provide the foundations for the creation of an educational pathway for diabetes educators in Saudi Arabia. Such a pathway will aim to increase diabetes educators’ knowledge, skills, and confidence in CGM and improve patient outcomes, including glycemic control and quality of life.

This study aimed to investigate the decline over time in the proportion of total daily insulin delivered as boluses in newly diagnosed youth with type 1 diabetes using a hybrid closed-loop system.

A secondary analysis was conducted using data from the CLOuD study, an open-label, multicenter, randomized, parallel hybrid closed-loop trial to investigate bolus patterns in youth with newly diagnosed type 1 diabetes.

Over the 48-month trial period, the proportion of total daily insulin delivered as carbohydrate-related boluses decreased from 58% to 34%. There was a decreasing trend in the median (interquartile range) amount of carbohydrates entered per day from 236 (204, 253) g to 184 (127, 232) g, and the number of carbohydrate-related boluses per day from 5.5 (4.6, 6.5) to 3.7 (2.9, 5.2) over the 48 months. Mean ± SD daily carbohydrate-related bolus insulin increased from 15.1 ± 6.6 to 22.0 ± 9.0 units/d, and the amount of insulin delivered per 10 g of carbohydrate more than doubled from 0.6 (0.5, 0.8) units to 1.3 (0.9, 1.5) units. The postprandial change in glucose (measured as the difference between peak glucose 30 to 180 minutes post carbohydrate-related bolus and glucose on carbohydrate-related bolus delivery) changed from 49 (45, 54) to 59 (53, 66) mg/dL.

The decline in the proportion of total daily insulin delivered for as bolus is likely attributable to a combination of missed boluses and under-bolusing, while the closed-loop algorithm compensates for the missed or insufficient carbohydrate-related insulin delivery by increasing basal insulin delivery.

Automated insulin delivery (AID) improves glycemia in people with type 1 diabetes (T1D). However, concern remains about early worsening of diabetic retinopathy (EWDR) following rapid and large glycemic improvements. This study evaluated diabetic retinopathy (DR) outcomes in adolescents and young adults with T1D (aged 10-30 years) following AID initiation.

This retrospective observational study included adolescents and young adults with T1D and hemoglobin A1c (HbA1c) ≥ 8.5% (69 mmol/mol) prior to AID initiation. Clinical data, continuous glucose monitoring (CGM) metrics, and retinopathy grading were collected from research and clinical databases before and after at least three months of AID initiation. Statistical analyses assessed outcomes.

A total of 95 adolescents/young adults (mean age 17.8 years, diabetes duration 9.7 years, 54.7% female) with a baseline HbA1c of 10.3% (89.5 mmol/mol) were included. Mean HbA1c improved by 2.1 percentage points (22.6 mmol/mol) following AID initiation. Retinopathy remained stable or improved in 72/95 (75.8%), while 23/95 (24.2%) experienced EWDR inclusive of diabetic macular edema (DME). While no one required treatment for DME, proliferative DR requiring treatment developed in three participants (3.2%); all had preexisting retinopathy and ≥1 additional diabetes-related complication/risk factor. Logistic regression identified age >18 years and preexisting retinopathy at AID initiation as the only significant risk factors for EWDR.

Automated insulin delivery is associated with substantial glycemic improvement in adolescents and young adults with T1D. Despite these large glycemic improvements, diabetic retinal disease remains stable or improves in most cases. Risk factors for deterioration include age >18 years and preexisting DR.

Older adults with type 2 diabetes experience high risk of hypoglycemia, yet clinicians often lack actionable glucose data to guide individualized treatment decisions. This study examined how standardized continuous glucose monitoring (CGM) reports inform clinicians’ decision-making using Endsley’s situation awareness (SA) framework.

We conducted semi-structured interviews. Thirty clinicians reviewed three simulated older adult type 2 diabetes cases before and after reviewing CGM reports. Data saturation was achieved, consistent with qualitative research standards. Using an SA informed framework, we analyzed the proportion of clinicians referencing specific data elements and describing changes in diagnostic reasoning and treatment decisions.

After CGM data review, clinicians frequently revised assessments and plans. Most clinicians (90%) incorporated time-in-range metrics, while 93% identified previously unrecognized hypoglycemia or nocturnal patterns. Continuous glucose monitoring data led 86% of clinicians to modify their original treatment plan, including deprescribing high-risk agents (eg, sulfonylureas), adjusting insulin timing or dosing, or initiating safer alternatives (eg, SGLT-2 inhibitors, GLP-1 receptor agonists). Clinicians also identified multiple barriers to treatment implementation, including cost, medication access, housing instability, and limited food security.

Continuous glucose monitoring data—compared with A1C alone—provided meaningful, actionable information that improved individualized treatment decisions for older adults with type 2 diabetes. Continuous glucose monitoring data enhanced clinicians’ perception, comprehension, and projection across the SA continuum, fostering more confident diagnostic reasoning and safer treatment strategies. These findings inform the design of SA-based clinical decision-support tools to better integrate CGM data, address contextual barriers, and optimize management of older adults with type 2 diabetes.

Hemoglobin A_1C (HbA_1C) is the gold standard for assessing long-term glycemic control in people with diabetes. Increasing use of continuous glucose monitoring (CGM) has led to adoption of the glucose management indicator (GMI) as a CGM‑based HbA_1C estimate, but GMI often differs from laboratory HbA_1C, especially in type 2 diabetes. This discordance may be associated with the fact that GMI, as a measure of central tendency, fails to capture temporal glycemic trends and variability that relate to HbA_1C formation.

To evaluate whether combining CGM-derived metrics capturing variability, excursions, and temporal trends improves estimation of laboratory-measured HbA_1C in type 2 diabetes.

A machine learning framework was applied to CGM data from a three-month randomized trial, including 159 participants with type 2 diabetes. Participants had ≥70% CGM data coverage and valid end-of-trial HbA_1C. From a standardized 90-day CGM window, 51 metrics were extracted. Benchmark models (mean glucose and GMI) were compared with models developed using forward and exhaustive feature selection with threefold cross-validated multiple linear regression.

Benchmark models yielded R-squared = 0.53. A forward selection model including five metrics (GMI at night, night-to-overall mean glucose ratio, glycemic risk assessment diabetes equation, time in tight range [3.0-7.8 mmol/L], time above range [13.9 mmol/L] at night) improved R-squared to 0.60. The best-performing model (substituting GRADE at night for GMI at night) achieved a similar R-squared (0.61). Nighttime and hyperglycemia‑related metrics were consistently selected.

Continuous glucose monitoring‑based HbA_1C estimation improves when variability and temporal patterns are included. Nighttime hyperglycemia adds notable predictive value, though further validation is needed.

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.

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.

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).

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.

As type 2 diabetes mellitus (T2DM) becomes an increasingly urgent global health concern, interest has grown in how screen-based behaviors contribute to its risk. Excessive screen exposure is often associated with sedentary lifestyles, poor sleep quality, and circadian disruption—all potential contributors to T2DM. Yet, how screen time interacts with specific sleep characteristics in shaping diabetes risk remains underexplored.

This study investigates the relationship between screen exposure and T2DM risk, with particular focus on sleep duration and diagnosed sleep disorders as potential effect modifiers. We also explored variation by age, sex, and racial/ethnic groups.

We analyzed data from 23 023 US adults in the 2007 to 2016 National Health and Nutrition Examination Survey. Screen exposure was dichotomized using age-specific thresholds (≥2 vs <2 hours/day for ages 3 to 18; ≥3 vs <3 hours/day for adults). Type 2 diabetes mellitus was defined by self-reported physician diagnosis. Sleep duration and diagnosed sleep disorders were examined as modifiers. Missing data were handled using multiple imputation by chained equations, and survey-weighted multinomial logistic regression was applied.

High screen exposure was associated with increased odds of T2DM in fully adjusted models (odds ratio [OR] = 3.47, 95% confidence interval [CI]: 2.74, 4.36). Sleep duration was not independently associated with T2DM, whereas sleep disorders were linked to approximately twofold higher odds (OR = 2.21, 95% CI: 1.17, 4.18). The screen-T2DM association was stronger among females than males, with variation observed across sleep and racial/ethnic subgroups.

Excessive screen time is linked to elevated T2DM risk, particularly among females and individuals with sleep disorders. Longitudinal research is needed to assess causality and inform targeted interventions.

Hypoglycemia is a critical challenge for insulin-dependent people with diabetes using multiple daily injections (MDI), who rely on reactive responses to continuous glucose monitoring (CGM) alerts. To meet the need for a proactive safety tool, we evaluated the performance of the Low Glucose Predict (LGP) feature in the Accu-Chek SmartGuide Predict App.

This retrospective analysis pooled data from three prospective trials, including 85 subjects over 2709 recording days. The LGP feature uses a XGBoost model to predict low glucose events up to 30 minutes in advance. Performance was assessed rigorously against both capillary blood glucose (BG) and CGM values, including an analysis with “close-call” predictions (+10 mg/dL above the threshold). Metrics included sensitivity, specificity, and ROC-AUC.

Against the stringent capillary BG reference, LGP showed high performance: sensitivity of 87.13% and specificity of 97.43% (ROC-AUC 0.9787). Including close-call events improved sensitivity to 91.89% and specificity to 98.09%. Referenced against CGM, sensitivity was 94.40% and specificity was 98.25%. The system provided an actionable mean lead time of 14.71 ± 8.30 minutes (CGM reference), with a low average daily true notification rate of 1.31 (2.60 including close-calls).

The LGP feature is an accurate, highly sensitive, and specific tool for timely, proactive low glucose prediction, validated against both capillary BG and CGM. This predictive intelligence is a crucial mechanism for people with diabetes to safely mitigate hypoglycemia risk, addressing a significant clinical gap and potentially reducing fear of hypoglycemia and diabetes distress.

To identify diurnal glycemic patterns in adults with type 2 diabetes (T2D) using continuous glucose monitoring (CGM)–based machine learning and examine their association with diabetes distress, a key psychosocial outcome.

In this observational study, 137 adults with T2D wore blinded CGM (FreeStyle Libre Pro), yielding 1657 days of data. Glycemic patterns were identified using unsupervised machine learning via Gaussian mixture modeling, validated with Bayesian information criterion and silhouette scores. Diabetes distress was assessed with the 17-item Diabetes Distress Scale and analyzed through analysis of covariance (ANCOVA), adjusting for age, sex, body mass index, diabetes duration, and glucose management indicator.

Clustering identified four distinct glycemic profiles: Cluster 1 (suboptimal control, nocturnal hypoglycemia; 15.8%), Cluster 2 (suboptimal control, nocturnal hyperglycemia; 27.1%), Cluster 3 (poorly controlled, prolonged hyperglycemia; 21.1%), and Cluster 4 (well controlled; 36.1%). Diabetes distress scores varied significantly: participants in Cluster 3 reported the highest distress (mean = 2.37, 95% CI = 1.99-2.76), while Cluster 4 reported the lowest (mean = 1.67, 95% CI = 1.48-1.86; P = .03). Effect sizes indicated differences corresponded to clinically meaningful categories of “little or no distress” vs “moderate distress.”

CGM-based machine learning identified physiologically distinct glycemic phenotypes that were also associated with psychosocial burden. This work demonstrates the added value of integrating CGM-derived profiles with patient-reported outcomes. These findings highlight the potential of CGM phenotyping to support precision diabetes care by enabling early identification of high-risk subgroups, guiding tailored behavioral and psychosocial interventions, and informing technology-enabled decision tools that connect physiological monitoring with emotional well-being in T2D management.

The hemoglobin glycation index (HGI), defined as the difference between HbA1c and the glucose management indicator (GMI) derived from continuous glucose monitoring (CGM), has emerged as a tool to evaluate discordance between laboratory and sensor-based measures. The impact of sodium–glucose cotransporter 2 inhibitors (SGLT2i) on these markers remains unclear.

We retrospectively analyzed CGM data from 143 individuals with type 2 diabetes, stratified by SGLT2i use. Both HGI and glycated albumin-to-HbA1c (GA/HbA1c) ratio were compared. A restricted dataset (n = 117) excluding individuals with anemia or advanced renal dysfunction was also examined.

SGLT2i users exhibited higher hematologic parameters and significantly greater HGI (full dataset: 0.3 vs 0.1, P = .0297; restricted: 0.4 vs 0.1, P = .0206), whereas the GA/HbA1c ratio was significantly lower (full: 2.27 vs 2.54, P = .0117; restricted: 2.18 vs 2.38, P = .0178). The HbA1c–GMI relationship showed significantly different slopes between SGLT2i users and non-users, while the GA–GMI relationship was consistent across groups. Multivariate analyses identified SGLT2i use as an independent determinant of both higher HGI and lower GA/HbA1c ratio, even after adjustment for age, sex, body mass index (BMI), hemoglobin, albumin, renal function, mean glucose, and glycemic variability.

SGLT2i therapy alters the interpretation of glycemic markers by elevating HGI and lowering GA/HbA1c, independent of hematologic and renal factors. These findings emphasize the need for individualized assessment of glycemic control using CGM-derived metrics and complementary biomarkers.

Stability of insulin varies on storage conditions, handling, and types of formulations, and in Tanzania, unreliable refrigeration and climate fluctuations can compromise storage, which can reduce potency. Real-world data are crucial for optimizing diabetes management in developing countries. Previous studies have reported mixed findings, linking insulin degradation to temperature fluctuations, formulation issues, and distribution challenges. This study assessed the impact of storage conditions on insulin content during usage among diabetic patients at Bugando Medical Centre (BMC).

This cross-sectional study, conducted from April to July 2024 in Mwanza, assessed the quality of 16 batches of insulin using the high-performance liquid chromatograph (HPLC) method. Insulin samples were collected from 48 diabetic patients attending the diabetic clinic at BMC and analyzed for physical appearance, identity, and concentration, while storage conditions were monitored using a Bluetooth temperature data logger.

Of the 16 insulin batches analyzed, 7 (43.8%) had low active insulin content, while 4 batches (25%) contained no active insulin at all. Among the 48 patients sampled, 24 (50%) were using substandard insulin. Of the 26 patients who used 2 separate vials, 15 had at least 1 vial that failed to meet standard specifications. Among the 20 patients using a single combined vial, 7 (35%) had low insulin content.

All patients stored insulin within recommended temperature conditions; however, several samples were substandard, with some batches failing specifications. While 56.2% met United States Pharmacopeia (USP) requirements, the remaining showed variability and reduced concentrations.

Ultrasound frequently detects hyperechogenic tissue at recent insulin infusion sites in youth using automated insulin delivery, but its short-term clinical significance is unclear.

In this post hoc paired analysis of a prior 4-week prospective study, participants were included if they had both hyperechogenic and normoechogenic findings at the most recently used infusion site across 3 visits. Insulin dose and continuous glucose monitoring metrics from days 1 to 2 after infusion-set placement were compared within participants.

Seventeen participants met the inclusion criteria. Insulin dose and continuous glucose monitoring–derived outcomes did not differ significantly between tissue categories. Time since infusion-set removal differed between categories (1.6 vs 2.1 days, P = .049).

In this paired analysis, binary ultrasound classification alone did not explain short-term glycemic variation.

As a standalone parameter, the wound surface area can be used to describe a wound in medical records; however, changes in the wound surface area over time can be used in chronic wounds to assess treatment efficacy and predict successful healing. Recent technological advances in mobile devices and artificial intelligence algorithms have enabled the development of new methods for measuring the surface area of wounds. These new methods require clinical validation or comparison with previously used methods.

Wound surface areas were measured with the Planimator app (PA), Silhouette Mobile device (SMD), and AutoPlanimator service (APS) in 77 patients from the outpatient clinic. One hundred forty-two ulcers were measured. The Passing-Bablok regression analysis was used to assess differences between each pair of methods used.

There were significant differences in proportional differences between two pairs of methods, ie, between the SMD and the PA and between the SMD and the APS. The Passing-Bablok regression analysis revealed the regression equation APS = −0.0145 + 0.987PA for the APS and the PA. There were nonsignificant differences between systematic and proportional differences for the PA and the APS; thus, the inequality between the APS and the PA cannot be confirmed and the two methods can be interchangeably used, which cannot be concluded for the SMD and the APS or the SMD and the PA. This study confirmed that the PA and the APS can be valuable tools in assessing the healing progress of chronic wounds.

Patch insulin pumps are often treated as a single class, although fully disposable and semi-reusable designs differ in cost and waste. We evaluated real-world clinical, pharmacoeconomic, and environmental outcomes of a semi-reusable tubeless insulin pump (Microtech Equil™, referred as SR-TIP).

Prospective, multicentre, open-label real-world study in adults with type 1 or type 2 diabetes transitioning from continuous subcutaneous insulin infusion (CSII) or multiple daily injections (MDI). Follow-up was 3 months. Primary endpoints were HbA1c non-inferiority and change in hypoglycaemic event frequency. Secondary endpoints included safety, device deficiencies, patient-reported outcomes (Diabetes Treatment Satisfaction Questionnaire, Device Assessment Questionnaire), monthly disposable treatment costs, and waste based on disposable component counts and material composition.

Ninety-seven participants completed follow-up (CSII n = 79; MDI n = 18). HbA1c changes met non-inferiority criteria in both groups. Hypoglycaemic events decreased by 45% in CSII users and 86% in MDI users. Device-related adverse events were infrequent and mainly mild. Among participants previously using fully disposable tubeless pumps, mean monthly disposable treatment costs decreased by €108.6 (p < 0.001). The semi-reusable architecture eliminated routine disposal of batteries/electronic components and reduced overall waste.

In routine care, a semi-reusable tubeless pump maintained glycaemic control while reducing hypoglycaemia and disposable costs versus fully disposable patch pumps, with measurable reductions in electronic waste. These data support value-based, sustainable diabetes technology adoption.