Journal of Diabetes Science and Technology最新文献

Background: Gestational diabetes mellitus (GDM) often requires pharmacological intervention beyond lifestyle modification to achieve optimal glycemic control. This study aimed to develop machine learning models that integrate clinical and gut microbiome data to predict the need for insulin therapy (IT) in women with GDM.

Methods: We characterized 205 pregnant women with GDM from the Genetic and Epigenetic Mechanisms of Developing Gestational Diabetes Mellitus and its Effects on the Fetus study, collecting clinical parameters, lifestyle questionnaires, self-monitored blood glucose records, and gut microbiome profiles based on 16S rRNA gene sequencing. Gradient-boosting models were trained to predict IT, basal insulin (BI), and prandial insulin (PI) requirements. Model discrimination was assessed using repeated stratified five-fold cross-validated area under the curve-receiver operating characteristic (AUC-ROC) (nested cross-validation). Feature importance and interpretability were evaluated with SHapley Additive exPlanations and permutation analyses. Differential microbial abundance was analyzed by ANCOM-BC2 (analysis of composition of microbiomes with bias correction, version 2), and metabolic pathways were predicted via PICRUSt2.

Results: Women requiring insulin were older and had higher pre-pregnancy body mass index (BMI), fasting plasma glucose, 1-hour oral glucose tolerance test glucose, and glycated hemoglobin than diet-treated women (P < .05 for all). Adding microbiome data improved AUC-ROC for IT prediction from 0.63 (95% CI = 0.43, 0.83) to 0.70 (0.50, 0.89), BI from 0.77 (0.59, 0.95) to 0.82 (0.65, 0.99), and for PI from 0.69 (0.50, 0.88) to 0.70 (0.51, 0.89). Key influential features included glycemic markers, BMI, and microbial taxa, such as Phascolarctobacterium faecium, Alistipes ihumii, Cloacibacillus evryensis, Ruthenibacterium lactatiformans, and Methanosphaera stadtmanae, and the predicted microbial metabolic pathway PWY-5823.

Conclusion: Our findings demonstrate that integrating gut microbiome characteristics with clinical data improves the prediction of insulin treatment needs in GDM, particularly for BI initiation.

Background: Hemoglobin A_1c (HbA_1c) interpretation can be affected by genetic and hematologic factors that alter erythrocyte turnover. This study investigated red blood cell (RBC) profiles and metabolomic alterations linked to glycemic variability in type 2 diabetes (T2D) and evaluated the effects of common RBC genetic disorders on HbA_1c interpretation.

Methods: Participants were recruited in Nakhon Si Thammarat, Thailand. In Phase 1, 244 normoglycemic participants and 447 individuals with T2D were enrolled. In Phase 2, 45 participants from each group were analyzed for hematologic and biochemical parameters. In Phase 3, liquid chromatography-mass spectrometry (LC-MS)-based RBC metabolomics were performed in 10 individuals without diabetes and 14 individuals with diabetes.

Results: Fasting blood glucose, fructosamine, and ferritin showed no significant differences, whereas HbA_1c was significantly lower in those with RBC disorders for both individuals without diabetes (P = .001) and individuals with diabetes (P < .001) groups. Red blood cells with hypochromic microcytosis in β-thalassemia heterozygote (BTH) were used as a model to explore metabolomic changes associated with normal and high HbA_1c levels. Multivariate analyses revealed distinct clustering patterns in high-HbA_1c cases. Interestingly, 5-oxo-L-proline exhibited the highest fold change (FC = 6.90, P = .0004), followed by 5-aminolevulinate and D-gluconic acid, along with increased oxidized/reduced glutathione and decreased NADH and sphingomyelin.

Conclusions: Distinct RBC metabolic signatures were observed in BTHs with elevated HbA_1c, highlighting alterations in redox and heme metabolism. These findings provide a basis for future investigations into RBC-derived metabolites as complementary tools for glycemic assessment in individuals with thalassemia and hemoglobinopathies.

Background: To describe changes in continuous glucose monitoring (CGM)-derived glucose metrics of healthy and physically active participants with mild dysglycemia at baseline (>5% time with glucose levels outside of 70-140 mg/dL) who wore a real-time CGM device (GSB, Glucose Sport Biosensor) integrated with a smartphone mobile application over an eight-week period (four GSB wear periods).

Methods: Two hundred twenty-five participants (51 females and 174 males) aged 45.0 ± 10.1 years with body mass index 23.4 ± 3.9 kg/m² with suboptimal time in tight range (TITR) (ie, <95%) wore a GSB for approximately eight weeks. Linear mixed-effects models (LMEMs) were used to compare the cumulative time in different glycemic zones (% time below range [TBR, <70 mg/dL]; % TITR [70-140 mg/dL]; % time above range [TAR, >140 mg/dL]) with respect to each GSB wear time.

Results: Linear-mixed effects model analysis returned significant effects of sensor on TITR and TBR across four wear periods (both P < .001), with inter-individual variability in baseline values and response slopes. Each day of sensor wear increased TITR by 0.59% (95% confidence interval [CI]: 0.50, 0.69, P < 0.001) and reduced TBR (-0.42 %, 95% CI: -0.50, -0.35, P <.001) and TAR (-0.17 %, 95% CI: -0.24, -0.10, P < .001), with small sensor-dependent differences in daily improvements.

Conclusions: These findings indicate both cumulative and day-to-day gains in glucose control with repeated sensor use for individuals with a TITR <95%. Indeed, CGM use coincided with short-term improvements in glucose metrics. Future studies should directly measure lifestyle behaviors to determine which factors may contribute to improvements in glycemia.

Background: Prediabetes often remains undiagnosed until it progresses to type 2 diabetes mellitus (T2DM), particularly among individuals with obesity or a family history of diabetes. Conventional tests, including fasting plasma glucose, hemoglobin A1c, and the oral glucose tolerance test, provide only static snapshots of glycemia and may fail to capture variability and postprandial excursions. Continuous glucose monitoring (CGM) offers dynamic insight into glucose regulation and may complement traditional diagnostic tools.

Methods: This systematic review and meta-analysis evaluated differences in CGM-derived metrics between individuals with prediabetes and those with normoglycemia. PubMed, EMBASE, and CENTRAL were searched from inception to September 3, 2025 (PROSPERO: CRD42024608658). The primary outcome was the mean amplitude of glycemic excursions (MAGE); secondary outcomes included time above range, 24-hour mean glucose, coefficient of variation, and time in range. Sixteen studies met the inclusion criteria, and ten were included in the quantitative synthesis (n = 1657).

Results: Prediabetes was consistently associated with higher CGM values compared with normoglycemia: MAGE (mean difference [MD] = 9.41 mg/dL, 95% confidence interval [CI]: 4.31, 15.31), TAR% (MD = 5.68%, 95% CI: 1.04, 10.32), and 24-hour mean glucose (MD = 7.91 mg/dL, CI: 6.27, 9.55).

Conclusions: These results provide the first quantitative evidence that CGM can discriminate between prediabetes and normoglycemia, supporting its potential as a complementary tool for refined metabolic risk assessment. Further prospective studies are needed to determine its predictive value for progression to T2DM.

Background: 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.

Objective: 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.

Methods: 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.

Results: 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.

Conclusion: 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.

The adoption of diabetes technology for the management of type 1 and insulin-treated type 2 diabetes has greatly increased. The annual volume of discarded continuous glucose monitoring (CGM) devices, considering only Dexcom and Freestyle Libre brands, totals more than 153 million units and Omnipod^® contributes an additional estimated 43.8 million units.Although these technologies are clinically effective, their environmental impact is unknown. Batteries are a pivotal, yet often overlooked, component in diabetes technologies and can exert a detrimental impact on the environment.In this commentary article, we explore the environmental impact of batteries used in diabetes devices. Furthermore, we highlight various strategies, including recycling of used batteries and alternative design approaches, that may reduce the environmental burden, as they become the ubiquitous standard of care for people with diabetes.

Background: Diabetic peripheral neuropathy (DPN) is a prevalent and debilitating complication of diabetes, often leading to severe neuropathic pain. Although other diabetes-related complications have witnessed a surge of emerging treatments in recent years, DPN has seen minimal progression. This stagnation stems from various factors, including insensitive diagnostic methods and inadequate treatment options for neuropathic pain.

Methods: In this comprehensive review, we highlight promising novel diagnostic techniques for assessing DPN, elucidating their development, strengths, and limitations, and assessing their potential as future reliable clinical biomarkers and endpoints. In addition, we delve into the most promising emerging pharmacological and mechanistic treatments for managing neuropathic pain, an area currently characterized by inadequate pain relief and a notable burden of side effects.

Results: Skin biopsies, corneal confocal microscopy, transcutaneous electrical stimulation, blood-derived biomarkers, and multi-omics emerge as some of the most promising new techniques, while low-dose naltrexone, selective sodium-channel blockers, calcitonin gene-related peptide antibodies, and angiotensin type 2 receptor antagonists emerge as some of the most promising new drug candidates.

Conclusion: Our review concludes that although several promising diagnostic modalities and emerging treatments exist, an ongoing need persists for the further development of sensitive diagnostic tools and mechanism-based, personalized treatment approaches.

Background: Continuous glucose monitoring (CGM) systems allow detailed assessment of postprandial glucose responses (PPGR), offering new insights into food choices' impact on dysglycemia. However, current approaches to analyze PPGR using a CGM require manual meal logging, limiting the scalability of CGM-driven applications like personalized nutrition and at-home diabetes risk assessment.

Objective: We propose a machine learning (ML) framework to automatically identify and characterize breakfast-related PPGRs from CGM profiles in adults at risk of or living with noninsulin-treated type 2 diabetes (T2D).

Methods: Our PPGR estimation framework uses a random forest ML algorithm trained on 15 adults without diabetes who wore a CGM for up to four weeks. The algorithm performance was evaluated on a held-out subset of the participants' CGM data as well as on an external validation data set of 36 individuals at risk for or with noninsulin-treated T2D.

Results: Our algorithm's estimations of breakfast PPGRs displayed no statistically significant differences to annotated PPGRs, in terms of incremental area under the curve and glucose rise (P > .05 for both data sets), while a small difference in prebreakfast glucose was found in the nondiabetes data set (P = .005) but not in the validation T2D data set (P = .18).

Conclusions: We designed an ML framework to automatically estimate the timing of meal events from CGM data in individuals without diabetes and in individuals at risk or with T2D. This could provide a more scalable approach for analyzing postprandial glycemia, increasing the feasibility of CGM-based precision nutrition and diabetes risk assessment applications.