Background: Predicting hypoglycemia while maintaining a low false alarm rate is a challenge for the wide adoption of continuous glucose monitoring (CGM) devices in diabetes management. One small study suggested that a deep learning model based on the long short-term memory (LSTM) network had better performance in hypoglycemia prediction than traditional machine learning algorithms in European patients with type 1 diabetes. However, given that many well-recognized deep learning models perform poorly outside the training setting, it remains unclear whether the LSTM model could be generalized to different populations or patients with other diabetes subtypes. Objective: The aim of this study was to validate LSTM hypoglycemia prediction models in more diverse populations and across a wide spectrum of patients with different subtypes of diabetes. Methods: We assembled two large data sets of patients with type 1 and type 2 diabetes. The primary data set including CGM data from 192 Chinese patients with diabetes was used to develop the LSTM, support vector machine (SVM), and random forest (RF) models for hypoglycemia prediction with a prediction horizon of 30 minutes. Hypoglycemia was categorized into mild (glucose=54-70 mg/dL) and severe (glucose<54 mg/dL) levels. The validation data set of 427 patients of European-American ancestry in the United States was used to validate the models and examine their generalizations. The predictive performance of the models was evaluated according to the sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Results: For the difficult-to-predict mild hypoglycemia events, the LSTM model consistently achieved AUC values greater than 97% in the primary data set, with a less than 3% AUC reduction in the validation data set, indicating that the model was robust and generalizable across populations. AUC values above 93% were also achieved when the LSTM model was applied to both type 1 and type 2 diabetes in the validation data set, further strengthening the generalizability of the model. Under different satisfactory levels of sensitivity for mild and severe hypoglycemia prediction, the LSTM model achieved higher specificity than the SVM and RF models, thereby reducing false alarms. Conclusions: Our results demonstrate that the LSTM model is robust for hypoglycemia prediction and is generalizable across populations or diabetes subtypes. Given its additional advantage of false-alarm reduction, the LSTM model is a strong candidate to be widely implemented in future CGM devices for hypoglycemia prediction.

Background. The aging population is increasingly susceptible to cardiovascular disease (CVD). Visit-to-visit variability in glucose and lipid levels both contributed to CVD risk independent of their mean values. However, whether variability in the triglyceride-glucose (TyG) index is a risk factor for CVD remains unknown. Research Design and Methods. In this retrospective study of electronic health records, 27,520 participants aged over 60 years were enrolled. The visit-to-visit variability of TyG index was calculated from annual health examination data and defined as average real variability (ARV), standard deviation (SD), or the coefficient of variability (CV). CVD events were identified from the chronic disease registry or follow-up database and included myocardial infarction, angina, coronary, and stroke. Multivariate Cox regression was used to examine the correlation between TyG variability and incident CVD. Results. Over a median follow-up of 6.2 years, there were 2,178 CVD events. When participants were divided into four quartiles according to their TyG variability, after adjusting for established CVD risk factors, subjects in the top quartile had (HR = 1.18, 95% CI 1.05–1.34, ) significantly higher CVD risk than those in the bottom quartile. The association remained significant in overweight individuals or those without diabetes ( and , respectively). Conclusions. High variability in TyG was significantly associated with elevated CVD risk in the elderly, independent of average TyG and other risk factors. Close monitoring variability in TyG might be informative to identify old individuals at high risk of CVD.

The R package Continuous Glucose Monitoring Time Series Data Analysis (CGMTSA) was developed to facilitate investigations that examine the continuous glucose monitoring (CGM) data as a time series. Accordingly, novel time series functions were introduced to (1) enable more accurate missing data imputation and outlier identification; (2) calculate recommended CGM metrics as well as key time series parameters; (3) plot interactive and three-dimensional graphs that allow direct visualizations of temporal CGM data and time series model optimization. The software was designed to accommodate all popular CGM devices and support all common data processing steps. The program is available for Linux, Windows, and Mac at GitHub.

Type 2 diabetes is an established risk factor for dementia. However, how its genetic heterogeneity affects different dementia subtypes remains unclear. In this study, we investigate the associations between genetic risk of type 2 diabetes and dementia subtypes among 33,136 older Chinese adults from the KARE cohort. We find that a higher overall polygenic risk score for type 2 diabetes is significantly associated with an increased risk of vascular dementia, but not Alzheimer's disease. Further analyses using cluster-specific partitioned polygenic score show that elevated genetic risk specific to the hyperinsulinemia pathway is strongly associated with increased incidence of vascular dementia. These findings highlight the potential role of insulin-related metabolic abnormalities in the pathogenesis of vascular dementia and provide genetic evidence to support the use of the hyperinsulinaemia pathway as a clinically relevant marker for early risk stratification and precision prevention strategies.

OBJECTIVE: mtDNA copy number (CN) reflects mitochondrial function, but prior studies have reported inconsistent associations with type 2 diabetes risk, ranging from inverse to positive or null findings. We hypothesized that mtDNA-CN is nonlinearly associated with incident type 2 diabetes. RESEARCH DESIGN AND METHODS: We included 34,835 adults without diabetes from the Kunshan Aging Research With E-Health (KARE) cohort and 289,338 from the UK Biobank (UKB). Associations between blood mtDNA-CN and incident type 2 diabetes were evaluated using Cox proportional hazards and restricted cubic spline models stratified by age. RESULTS: A U-shaped association was observed in the KARE cohort (P < 0.001), in which the hazard ratios (95% CIs) across increasing mtDNA-CN quartiles were 1.00 (reference), 0.94 (0.88–1.00), 0.85 (0.79–0.91), and 0.93 (0.87–1.00). In contrast, the UKB cohort exhibited a predominantly inverse linear trend. Age-stratified analyses revealed that this U-shaped association was particularly evident in younger participants (aged <65 years in KARE and <50 years in UKB), indicating elevated diabetes risk at both low and high mtDNA-CN levels. Additionally, mtDNA-CN declined with age in both cohorts, with an accelerated decrease beyond ∼65 years in KARE and ∼50 years in UKB. CONCLUSIONS: Blood mtDNA-CN showed a U-shaped association with incident type 2 diabetes in younger individuals.

Background: Predicting hypoglycemia while maintaining low false alarm rate is a bottleneck for wide adoption of continuous glucose monitoring (CGM) in diabetes management. One small study suggested the long short-term memory (LSTM) network deep learning model had better performance of hypoglycemia prediction than traditional machine learning algorithms in European patients with type 1 diabetes. However, given that many well-recognized deep learning models perform poorly outside the training consideration, whether the LSTM model could be generalized to different populations or patients with other diabetes subtypes are unknown. Methods: We assembled two large datasets of patients with both type 1 diabetes and type 2 diabetes. The primary dataset containing 192 patients from Chinese were used to develop the LSTM, support vector machine (SVM) and random forest (RF) models for hypoglycemia prediction at the prediction horizon of 30 minutes. Hypoglycemia was defined as the mild (54mg/dl <= glucose < 70mg/dl) and severe (< 54mg/dl) hypoglycemic level separately. The validation dataset of 427 patients from European-Americans was used to validate the models and examine their generalizations. The predictive performance of the models was evaluated by sensitivity, specificity and area under the operating curve (AUC). Results: For the difficulty to predict mild hypoglycemia events, the LSTM model always achieved AUC greater than 97.22% in the primary dataset, with less than 3% AUC reduction in the validation dataset, indicating the model was robust and generalizable across populations. AUC higher than 93.49% was also achieved when LSTM was applied to both type 1 diabetes and type 2 diabetes in the validation dataset, further strengthening the generalizability of the model. Under different satisfactory levels of sensitivity for mild and severe hypoglycemia prediction, the LSTM model achieved higher specificity than the SVM and RF models, thereby reducing false alarms. Conclusions: Our results demonstrated that the LSTM model was robust for hypoglycemia prediction and generalizable across populations or diabetes subtypes. Given its extra advantage on false alarm reduction, the LSTM model was a strong candidate to be widely implemented by future CGM devices for hypoglycemia prediction.