Diabetes & Metabolism Journal

Original Articles

Complications
Time-Window Stratified Machine-Learning Risk Prediction Model for Diabetic Retinopathy and Cross-Cohort Study: Jingwen Hui, Zheya Han, Yuxi Bai, Yawen Gong, Quanhong Han, Xuehao Cui; Received November 1, 2025 Accepted December 18, 2025 Published online April 15, 2026; DOI: https://doi.org/10.4093/dmj.2025.1098 [Epub ahead of print]

1,551 View
27 Download

Abstract PDF Supplementary Material PubReader ePub: Background
Diabetic retinopathy (DR) remains a significant cause of vision loss worldwide. Existing risk models rarely account for when DR develops relative to the onset of diabetes, even though early- and late-onset disease may have different clinical implications. We hypothesized that DR occurring within 6 years of diabetes diagnosis (short-term) represents an early-onset phenotype driven mainly by metabolic dysregulation and microvascular injury, whereas DR developing after 6 years (long-term) reflects late-onset disease shaped by cumulative metabolic burden and aging. This study aimed to develop and validate a time-window- stratified DR risk prediction model.
Methods
Data from two large cohorts (UK Biobank and Tianjin Eye Hospital) were analyzed, including 1,943 patients with diabetes but without DR at baseline. Separate Cox models were built for short-term (≤6 years) and long-term (>6 years) DR incidence. Feature selection used least absolute shrinkage and selection operator (LASSO) and Boruta algorithms, and model performance was assessed by area under the curve (AUC), calibration, and decision curve analyses. Machine-learning models were further developed on pooled data for multiclass classification (no DR, short-term DR, long-term DR) and interpretability using SHapley Additive exPlanations (SHAP) analysis.
Results
Both models identified glycosylated hemoglobin and retinal neurostructural measures (retinal nerve fiber layer thickness and retinal ganglion cell layer thickness) as consistent predictors, indicating that neuroretinal degeneration precedes clinical DR. The short-term model emphasized renal and lipid metabolism markers, whereas the long-term model highlighted age and uric acid. Internal validation and cross-cohort replication showed stable discrimination (AUC 0.79–0.84). The pooled XGBoost model improved accuracy (overall approximately 80%) with transparent interpretability.
Conclusion
Time-window-based modeling revealed distinct early- and late-onset DR risk profiles, thereby enhancing prediction precision. Integrating optical coherence tomography imaging with routine clinical variables offers a practical, interpretable framework for individualized risk assessment and early intervention in DR.

Complications
Optimizing Early Detection of Diabetic Kidney Disease through Synergistic Biomarkers and Serum Metabolites in Humans: Xianke Zhou, Yuan Gui, Jia-Jun Liu, Shijia Liu, Dongning Liang, Yuanyuan Wang, Henry Wells Shaffer, Samantha Mae Mallari, Cameron Jones, Priya Gupta, Dier Li, Ke Zhang, Ying Yu, Jianling Tao, Yanlin Wang, Silvia Liu, Dong Zhou, Haiyan Fu; Diabetes Metab J. 2026;50(4):752-769. Published online January 29, 2026; DOI: https://doi.org/10.4093/dmj.2025.0193

3,270 View
149 Download
2 Web of Science
2 Crossref

Abstract PDF Supplementary Material PubReader ePub

Background
Diabetic kidney disease (DKD) often progresses to end-stage renal disease more rapidly than nondiabetic kidney disease because of persistent hyperglycemia and early activation of multiple pathogenic pathways. Early detection of DKD is crucial for identifying subtle kidney damage before clinical symptoms appear.
Methods
This study combined human serum proteomics with public single-cell RNA sequencing and spatial transcriptomics data from diabetic kidneys to identify key biomarkers for DKD diagnosis. These biomarkers were validated in multiple organs of db/db mice at early and advanced stages. In a discovery cohort, sera from 173 healthy adults and 444 patients with type 2 diabetes mellitus (T2DM), with or without kidney disease, were analyzed using metabolomics and enzyme-linked immunosorbent assay (ELISA). Multiple machine learning algorithms were developed to integrate synergistic biomarkers and serum metabolites for early DKD detection, with results validated in 435 participants from four independent clinical cohorts.
Results
Metalloproteinase-7 (MMP-7) and tenascin C (TNC) were elevated in human diabetic kidneys at the single-cell and spatial levels. Proteomics indicated upregulation of serum amyloid A1 (SAA1) and TNC in the serum of patients with DKD. In db/db mice, all three biomarkers increased in multiple organs by 18 weeks of age. In sera from patients with DKD, MMP-7 and TNC levels were consistently elevated across cohorts. The new algorithms combining MMP-7, SAA1, and TNC enhanced early-stage DKD detection, with approximately 13% improvements in accuracy when serum metabolites were included to distinguish progression from early to advanced DKD stages.
Conclusion
Integrating synergistic biomarkers with serum metabolomics enhances the early detection of DKD, potentially improving outcomes by slowing disease progression in patients with T2DM.

Citations

Citations to this article as recorded by

Recent advances in point-of-care colorimetric biosensors for detecting small molecule metabolites
Zhiqiang Zhu, Zhun Gu, Xinxing Cao, Shanshan Zhang, Shao Su
Chemical Communications.2026; 62(48): 12000. CrossRef
Pathology-Anchored Biomarker Research Progress for the Early Diagnosis of Diabetic Kidney Disease: From Pathological Association to Early Validation
Qiu Li, Mei Yang, Yingyu Luo, Nannan Zhang
Biomedicines.2026; 14(7): 1643. CrossRef

Basic and Translational Research
Revealing VCAN as a Potential Common Diagnostic Biomarker of Renal Tubules and Glomerulus in Diabetic Kidney Disease Based on Machine Learning, Single-Cell Transcriptome Analysis and Mendelian Randomization: Li Jiang, Jie Jian, Xulin Sai, Xiai Wu; Diabetes Metab J. 2025;49(3):407-420. Published online January 24, 2025; DOI: https://doi.org/10.4093/dmj.2024.0233

9,982 View
403 Download
4 Web of Science
4 Crossref

Abstract PDF Supplementary Material PubReader ePub

Background
Diabetic kidney disease (DKD) is recognized as a significant complication of diabetes mellitus and categorized into glomerular DKDs and tubular DKDs, each governed by distinct pathological mechanisms and biomarkers.
Methods
Through the identification of common features observed in glomerular and tubular lesions in DKD, numerous differentially expressed gene were identified by the machine learning, single-cell transcriptome and mendelian randomization.
Results
The diagnostic markers versican (VCAN) was identified, offering supplementary options for clinical diagnosis. VCAN significantly highly expressed in glomerular parietal epithelial cell and proximal convoluted tubular cell. It was mainly involved in the up-regulation of immune genes and infiltration of immune cells like mast cell. Mendelian randomization analysis confirmed that serum VCAN protein levels were a risky factor for DKD, while there was no reverse association. It exhibited the good diagnostic potential for estimated glomerular filtration rate and proteinuria in DKD.
Conclusion
VCAN showed the prospects into DKD pathology and clinical indicator.

Citations

Citations to this article as recorded by

Transient versican expression is required for β1-integrin accumulation during podocyte layer morphogenesis in amphibian developing kidney
Isabelle Buisson, Jean-François Riou, Muriel Umbhauer, Ronan Le Bouffant, Valérie Bello
Cells & Development.2026; 185: 204062. CrossRef
m6A Modified VCAN Promotes Glomerular Endothelial Cells Injury and Diabetic Nephropathy by SHH Pathway
Jie Jiang, Jicheng Zhang, Chao Wang, Feng Wang
Applied Biochemistry and Biotechnology.2026; 198(5): 3461. CrossRef
Pathology-Anchored Biomarker Research Progress for the Early Diagnosis of Diabetic Kidney Disease: From Pathological Association to Early Validation
Qiu Li, Mei Yang, Yingyu Luo, Nannan Zhang
Biomedicines.2026; 14(7): 1643. CrossRef
Multi-omics and machine learning identify FN1 and ALDH2 as diagnostic biomarkers and therapeutic targets in early and late diabetic kidney disease
Jingwei Lin, Yingying Zheng, Diman Mai, Fuxiang Fang, Zhuokun Wei, Mengyu Liu, Trung Hieu Pham, Ming Li, Jiawen Zhao
Renal Failure.2025;[Epub] CrossRef

Review

Drug/Regimen
Machine Learning Approach to Drug Treatment Strategy for Diabetes Care: Kazuya Fujihara, Hirohito Sone; Diabetes Metab J. 2023;47(3):325-332. Published online January 12, 2023; DOI: https://doi.org/10.4093/dmj.2022.0349

66,973 View
410 Download
14 Web of Science
13 Crossref

Abstract PDF PubReader ePub

Globally, the number of people with diabetes mellitus has quadrupled in the past three decades, and approximately one in 11 adults worldwide have diabetes mellitus. Since both microvascular and macrovascular diseases in patients with diabetes predispose them to a lower quality of life as well as higher rates of mortality, managing blood glucose levels is of clinical relevance in diabetes care. Many classes of antihyperglycemic drugs are currently approved to treat hyperglycemia in patients with type 2 diabetes mellitus, with several new drugs having been developed during the last decade. Diabetes-related complications have been reduced substantially worldwide. Prioritization of therapeutic agents varies according to national guidelines. However, since the characteristics of participants in clinical trials differ from patients in actual clinical practice, it is difficult to apply the results of such trials to clinical practice. Machine learning approaches became highly topical issues in medicine along with rapid technological innovations in the fields of information and communication in the 1990s. However, adopting these technologies to support decision-making regarding drug treatment strategies for diabetes care has been slow. This review summarizes data from recent studies on the choice of drugs for type 2 diabetes mellitus focusing on machine learning approaches.

Citations

Citations to this article as recorded by

Leveraging AI chatbots in supporting breast and ovarian cancer patients: A scoping review
Vivian Hui, Lidan Tian, Xinyu Feng, Xiaoling Yuan, Janelle Yorke, Young Ji Lee
DIGITAL HEALTH.2026;[Epub] CrossRef
Digital biomarkers in chronic disease management: a systematic review of personalization and ethical challenges
Shrivats Manikandan, Steiv Shore, Daniel Waszczuk, Sumitra Miriyala, Gayathri Sadanala
Personalized Medicine.2026; 23(3): 275. CrossRef
Improving Clinical Preparedness: Community Health Nurses and Early Hypoglycemia Prediction in Type 2 Diabetes Using Hybrid Machine Learning Techniques
Sachin Ramnath Gaikwad, Mallikarjun Reddy Bontha, Seeta Devi, Dipali Dumbre
Public Health Nursing.2025; 42(1): 286. CrossRef
Empowerment in Type 2 diabetes: A patient-centred approach for lifestyle change
Charlotte Björk Ingul, Siri Marte Hollekim-Strand, Mari Mørkeset Sandbakk, Torunn Ingfrid Grønseth, Tone Iren K. Rånes, Lars Tung Dyrendahl, Katarina Eilertsen, Stephan Kristensen, Turid Follestad, Bjarte Bye Løfaldli
Diabetes Research and Clinical Practice.2025; 220: 111998. CrossRef
Artificial intelligence to improve cardiovascular population health
Benjamin Meder, Folkert W Asselbergs, Euan Ashley
European Heart Journal.2025; 46(20): 1907. CrossRef
Enhancing Antidiabetic Drug Selection Using Transformers: Machine-Learning Model Development
Hisashi Kurasawa, Kayo Waki, Tomohisa Seki, Eri Nakahara, Akinori Fujino, Nagisa Shiomi, Hiroshi Nakashima, Kazuhiko Ohe
JMIR Medical Informatics.2025; 13: e67748. CrossRef
Machine learning algorithms to predict stroke in China based on causal inference of time series analysis
Qizhi Zheng, Ayang Zhao, Xinzhu Wang, Yanhong Bai, Zikun Wang, Xiuying Wang, Xianzhang Zeng, Guanghui Dong
BMC Neurology.2025;[Epub] CrossRef
Investigation of insulinotropic mechanism of paracetamol: in silico followed by in vivo approaches for repurposing
Majid Khan, Muhammad Usman Amin, Ashraf Ullah Khan, Sabi Ur Rehman, Safeer Khan
Toxicological & Environmental Chemistry.2025; 107(7): 1547. CrossRef
Artificial intelligence in glycemic management for diabetes: Applications, opportunities and challenges
Zhen Ying, Xiaoying Li, Ying Chen
Journal of Translational Internal Medicine.2025; 13(4): 314. CrossRef
Documentation of social determinants of health for patients with type 2 diabetes in Epic Cosmos
Polina V Kukhareva, Matthew J O’Brien, Daniel C Malone, Kensaku Kawamoto, Ramkiran Gouripeddi, Deepika Reddy, Mingyuan Zhang, Vikrant G Deshmukh, David Danks, Julio C Facelli
JAMIA Open.2025;[Epub] CrossRef
Exploring antioxidant activities and inhibitory effects against α‐amylase and α‐glucosidase of Elaeocarpus braceanus fruits: insights into mechanisms by molecular docking and molecular dynamics
Hong Li, Yuanyue Zhang, Zhijia Liu, Chaofan Guo, Maurizio Battino, Shengbao Cai, Junjie Yi
International Journal of Food Science & Technology.2024; 59(1): 343. CrossRef
3D Convolutional Neural Networks for Predicting Protein Structure for Improved Drug Recommendation
Pokkuluri Kiran Sree, SSSN Usha Devi N
EAI Endorsed Transactions on Pervasive Health and Technology.2024;[Epub] CrossRef
Artificial Intelligence in Plastic Surgery: Advancements, Applications, and Future
Tran Van Duong, Vu Pham Thao Vy, Truong Nguyen Khanh Hung
Cosmetics.2024; 11(4): 109. CrossRef

Original Article

Others
Development of Various Diabetes Prediction Models Using Machine Learning Techniques: Juyoung Shin, Jaewon Kim, Chanjung Lee, Joon Young Yoon, Seyeon Kim, Seungjae Song, Hun-Sung Kim; Diabetes Metab J. 2022;46(4):650-657. Published online March 11, 2022; DOI: https://doi.org/10.4093/dmj.2021.0115

14,374 View
500 Download
18 Web of Science
15 Crossref

Abstract PDF Supplementary Material PubReader ePub

Background
There are many models for predicting diabetes mellitus (DM), but their clinical implication remains vague. Therefore, we aimed to create various DM prediction models using easily accessible health screening test parameters.
Methods
Two sets of variables were used to develop eight DM prediction models. One set comprised 62 easily accessible examination results of commonly used variables from a tertiary university hospital. The second set comprised 27 of the 62 variables included in the national routine health checkups. Gradient boosting and random forest algorithms were used to develop the models. Internal validation was performed using the stratified 10-fold cross-validation method.
Results
The area under the receiver operating characteristic curve (ROC-AUC) for the 62-variable DM model making 12-month predictions for subjects without diabetes was the largest (0.928) among those of the eight DM prediction models. The ROC-AUC dropped by more than 0.04 when training with the simplified 27-variable set but still showed fairly good performance with ROC-AUCs between 0.842 and 0.880. The accuracy was up to 11.5% higher (from 0.807 to 0.714) when fasting glucose was included.
Conclusion
We created easily applicable diabetes prediction models that deliver good performance using parameters commonly assessed during tertiary university hospital and national routine health checkups. We plan to perform prospective external validation, hoping that the developed DM prediction models will be widely used in clinical practice.

Citations

Citations to this article as recorded by

Federated multimodal AI for precision-equitable diabetes care
Bing Bai, Xilin Liu, Hong Li
Frontiers in Digital Health.2026;[Epub] CrossRef
An Explainable Transformer‐Based Model for Predicting Chronic Diseases Risk
Nan Xu, Kaiyuan Zhang, Jiao Tian
Concurrency and Computation: Practice and Experience.2026;[Epub] CrossRef
Machine learning-driven risk prediction for post-hospitalization diabetes case management: Integrating clinical and social determinants of health
Seung-Yup Lee, Mohammad Saleem, Andrew M. Land, Erin W. DeLaney, Alison R. Garretson, Mahee Patel, Allyson G. Hall, Salisa Westrick, Fernando Ovalle, Andrea L. Cherrington, Jane C. Banaszak-Holl
International Journal of Medical Informatics.2026; 220: 106604. CrossRef
Artificial intelligence applied to diabetes complications: a bibliometric analysis
Yukun Tao, Jinzheng Hou, Guangxin Zhou, Da Zhang
Frontiers in Artificial Intelligence.2025;[Epub] CrossRef
The Impact of Balancing Techniques and Feature Selection on Machine Learning Models for Diabetes Detection
Vahid Sinap
Fırat Üniversitesi Mühendislik Bilimleri Dergisi.2025; 37(1): 303. CrossRef
Applications of Artificial Intelligence and Machine Learning in Prediabetes: A Scoping Review
Benjamin Lalani, Rohan Herur, Daniel Zade, Grace Collins, Devin M. Dishong, Setu Mehta, Jalene Shim, Yllka Valdez, Nestoras Mathioudakis
Journal of Diabetes Science and Technology.2025;[Epub] CrossRef
Predictive modeling for the development of diabetes mellitus using key factors in various machine learning approaches
Marenao Tanaka, Yukinori Akiyama, Kazuma Mori, Itaru Hosaka, Kenichi Kato, Keisuke Endo, Toshifumi Ogawa, Tatsuya Sato, Toru Suzuki, Toshiyuki Yano, Hirofumi Ohnishi, Nagisa Hanawa, Masato Furuhashi
Diabetes Epidemiology and Management.2024; 13: 100191. CrossRef
Validation of the Framingham Diabetes Risk Model Using Community-Based KoGES Data
Hye Ah Lee, Hyesook Park, Young Sun Hong
Journal of Korean Medical Science.2024;[Epub] CrossRef
Integrated Embedded system for detecting diabetes mellitus using various machine learning techniques
Rishita Konda, Anuraag Ramineni, Jayashree J, Niharika Singavajhala, Sai Akshaj Vanka
EAI Endorsed Transactions on Pervasive Health and Technology.2024;[Epub] CrossRef
Predicting diabetes in adults: identifying important features in unbalanced data over a 5-year cohort study using machine learning algorithm
Maryam Talebi Moghaddam, Yones Jahani, Zahra Arefzadeh, Azizallah Dehghan, Mohsen Khaleghi, Mehdi Sharafi, Ghasem Nikfar
BMC Medical Research Methodology.2024;[Epub] CrossRef
Use of Machine Learning to Predict the Incidence of Type 2 Diabetes Among Relatively Healthy Adults: A 10-Year Longitudinal Study in Taiwan
Ying-Qiang Liu, Tzu-Wei Chang, Lung-Chun Lee, Chia-Yu Chen, Pi-Shan Hsu, Yu-Tse Tsan, Chao-Tung Yang, Wei-Min Chu
Diagnostics.2024; 15(1): 72. CrossRef
The Present and Future of Artificial Intelligence-Based Medical Image in Diabetes Mellitus: Focus on Analytical Methods and Limitations of Clinical Use
Ji-Won Chun, Hun-Sung Kim
Journal of Korean Medical Science.2023;[Epub] CrossRef
Machine learning for predicting diabetic metabolism in the Indian population using polar metabolomic and lipidomic features
Nikita Jain, Bhaumik Patel, Manjesh Hanawal, Anurag R. Lila, Saba Memon, Tushar Bandgar, Ashutosh Kumar
Metabolomics.2023;[Epub] CrossRef
Retrospective cohort analysis comparing changes in blood glucose level and body composition according to changes in thyroid‐stimulating hormone level
Hyunah Kim, Da Young Jung, Seung‐Hwan Lee, Jae‐Hyoung Cho, Hyeon Woo Yim, Hun‐Sung Kim
Journal of Diabetes.2022; 14(9): 620. CrossRef
Improving Machine Learning Diabetes Prediction Models for the Utmost Clinical Effectiveness
Juyoung Shin, Joonyub Lee, Taehoon Ko, Kanghyuck Lee, Yera Choi, Hun-Sung Kim
Journal of Personalized Medicine.2022; 12(11): 1899. CrossRef

First
Prev
Page of 1
Next
Last

Search