CoChat
Sign Up
Cited by 4Artificial Intelligence in Healthcare and EducationArtificial Intelligence in HealthcareMachine Learning in Healthcare

Generative artificial intelligence in predictive analysis of diabetes and its complications: a narrative review

Open access
Osonuga, Ayokunle, Olawade, David B., Gore, Manisha, Rotifa, Oluwayomi B., Egbon, Eghosasere, Chinwah, Viviane, Boussios, Stergios·2025·Annals of Translational Medicine
DOI 10.21037/atm-25-62PMC PMC12592000
Open this paper in CoChat — free account

Abstract

Background and Objective: Diabetes mellitus (DM), particularly type 2 diabetes (T2D), represents a significant global health crisis, often complicated by severe and progressive conditions such as retinopathy, neuropathy, and cardiovascular disease. Traditional diagnostic approaches frequently detect these complications at advanced stages, limiting the opportunity for early, effective intervention. This review aims to examine how recent advancements in generative artificial intelligence (AI), particularly large language models (LLMs), can transform diabetes management by enabling earlier detection and more personalized interventions. Methods: A narrative review was conducted to evaluate the current literature on the application of generative AI and LLMs in diabetes care. The review focused on how these technologies analyse multi-dimensional datasets, including medical imaging, electronic health records (EHRs), genetic profiles, and lifestyle factors, and how they process both structured and unstructured data to enhance predictive analytics and risk stratification for diabetes complications. Key Content and Findings: Generative AI models have demonstrated significant promise in detecting hidden trends and early risk factors for complications such as diabetic retinopathy and neuropathy, often before clinical symptoms manifest. LLMs enhance predictive performance by synthesising unstructured data sources, such as physician notes and patient-reported outcomes, with clinical datasets. Despite limitations concerning data quality, model transparency, and ethical concerns surrounding data privacy, these technologies offer powerful tools for proactive disease monitoring and personalized care. Conclusions: Generative AI and LLMs are poised to redefine diabetes management by enabling earlier detection of complications and personalised treatment strategies. Their integration into clinical decision support systems (CDSS) and precision medicine frameworks may reduce the global burden of diabetes, improve patient outcomes, and shift care from reactive to preventative.

AI snapshot

Written from the paper’s full text.

TL;DR

This narrative review synthesizes evidence that generative AI and large language models can improve predictive analysis and early detection of diabetes and its complications by integrating structured and unstructured data. It argues these tools may shift diabetes care from reactive to proactive, while noting data quality, transparency, ethical, and validation limitations.

Key findings

  • Generative AI and LLMs can uncover hidden trends and early risk factors for diabetic retinopathy, neuropathy, and cardiovascular disease before clinical symptoms manifest, using sources such as medical imaging, EHRs, genetic profiles, lifestyle factors, physician notes, and patient-reported outcomes.
  • In cited supervised machine learning examples, De Silva et al. identified 25 predictors of prediabetes with ≥70% AUROC, outperforming the CDC prediabetes screening tool; Buccheri et al. reported 75.3% AUC using only age and waist circumference, with sensitivity 0.65 and specificity 0.73, and stratified AUC of 0.69–0.78 across sex and ethnic groups.
  • A UK Biobank study by Dorraki et al. achieved 85.13% accuracy for cardiovascular disease prediction incorporating psychological factors (n=375,145), and AI-enabled single-lead ECG analysis predicted heart failure risk.
  • The review highlights future integration into clinical decision support systems, personalized medicine, drug discovery, and patient self-management tools, but notes persistent concerns about data quality, model transparency, and data privacy/ethics.

Methods

  • Narrative review with a literature search in PubMed, Scopus, Web of Science, IEEE Xplore, and Google Scholar; search conducted 24 March 2025, covering 1 January 2018 to 31 December 2024, using a Boolean string combining generative AI/LLM terms with diabetes and predictive analysis terms.
  • Inclusion criteria were English-language peer-reviewed articles, reviews, and relevant grey literature discussing generative AI or LLMs in diabetes management, early detection, prediction, or complications; exclusions included non-generative AI without diabetes relevance, commentaries without data-backed discussion, duplicates, and outdated sources.
  • Data were extracted and thematically grouped; no formal quality assessment was conducted due to source heterogeneity, and Table 1 states formal dual independent screening was not undertaken.

Limitations

  • Stated: as a narrative review, the methodology lacks a rigorous systematic framework, introducing potential selection bias because inclusion was based on relevance and thematic alignment rather than formal quality appraisal or protocol-driven screening.
  • Stated: the search was limited to English-language articles accessible through major databases, so relevant non-English or non-indexed studies may have been overlooked; many included studies are emerging, preprints, or preliminary, with limited long-term clinical validation and uncertain generalizability/reproducibility; the rapidly evolving field may have produced newer developments after writing; and no formal risk-benefit analysis or stakeholder perspectives were included.
  • Evident: Table 1 explicitly notes that formal dual independent screening was not undertaken and no formal quality appraisal was performed, so conclusions rest on heterogeneous sources.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.21037/atm-25-62

Cites

0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

56 of 132 cited sources checked · 13 supported · 1 flagged · accuracy not yet measured

  • Premisecitation

    A study demonstrated that LLMs can predict diabetic neuropathy by analyzing a combination of EHR data, physician notes, and patient-reported symptoms.

    “A study demonstrated the use of LLMs in predicting diabetic neuropathy by analyzing a combination of EHR data, physician notes, and patient-reported symptoms (88).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [88] Ding JE Thao PNM Peng WC Large language multimodal models for new-onset type 2 diabetes prediction using five-year cohort electronic health records. Sci Rep 2024; 14: 20774. 10.1038/s41598-024-71020-2 39237580 PMC11377777 source

      The Introduction and Conclusions identify the predicted outcome as new-onset type 2 diabetes, not diabetic neuropathy. The study does use EHR clinical notes and laboratory data, but its Data collection and study design section does not identify patient-reported symptoms as a separate input. (LLM, conf=0.89)

      Checked by a reasoning model — may be wrong.

      Experimental — accuracy not yet measured
  • Premisecitation

    In diabetes care, LLMs are capable of analyzing diverse information sources such as electronic health records, clinical notes, medical literature, and patient-reported outcomes.

    “In the context of diabetes care, LLMs are capable of analyzing diverse information sources, such as electronic health records (EHRs), clinical notes, medical literature, and even patient-reported outcomes (12).”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [12] Mukherjee S, Gamble P, Ausin MS, et al. Polaris: A safety-focused llm constellation architecture for healthcare. arXiv 2024. arXiv:2403.13313. source

      The Healthcare specialization section describes clinical-note and EHR processing, while §§4.4 and 4.4.6 describe a specialist using EHR data and patient-reported lab values, including a diabetes-related A1C example. The paper does not establish that LLMs analyze medical literature or patient-reported outcomes as information sources specifically in diabetes care; its mention of medical literature concerns a suggestion for improving training. (LLM, conf=0.86)

      Experimental — accuracy not yet measured
  • Premisecitation

    Diabetic retinopathy can go undetected until it has caused irreversible vision damage.

    “For instance, diabetic retinopathy can go undetected until it has caused irreversible vision damage (6), and diabetic neuropathy might only be diagnosed after patients experience substantial nerve impairment (7).”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [6] Kropp M Golubnitschaja O Mazurakova A Diabetic retinopathy as the leading cause of blindness and early predictor of cascading complications-risks and mitigation. EPMA J 2023; 14: 21- 42. 10.1007/s13167-023-00314-8 36866156 PMC9971534 source
  • Premisecitation

    LLMs can identify early signs of diabetic retinopathy by cross-referencing a patient's clinical history with emerging trends in large datasets, providing clinicians a tool to intervene before irreversible damage occurs.

    “For instance, Tan et al. worked on how LLMs can identify early signs of diabetic retinopathy by cross-referencing a patient’s clinical history with emerging trends in large datasets, providing clinicians with a powerful tool to intervene before irreversible damage occurs (23).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [23] Tan TF, Elangovan K, Jin L, et al. Fine-tuning large language model (llm) artificial intelligence chatbots in ophthalmology and llm-based evaluation using GPT-4. arXiv 2024. arXiv:2402.10083. source

      The ‘How is diabetic retinopathy diagnosed?’ passage says that diagnosis and treatment can prevent progression, supporting the importance of timely intervention. But the Introduction describes this study as evaluating LLM responses to patients’ ophthalmology questions; our copy does not show LLMs identifying early signs by cross-referencing clinical histories with trends in large datasets. (LLM, conf=0.82)

      Experimental — accuracy not yet measured
  • Premisecitation

    Recent advancements in AI, particularly predictive analytics, have significantly improved the ability to monitor and manage diabetes.

    “However, recent advancements in AI, particularly through the development of predictive analytics, have significantly improved the ability to monitor and manage diabetes (31).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [31] Guan Z Li H Liu R Artificial intelligence in diabetes management: Advancements, opportunities, and challenges. Cell Rep Med 2023; 4: 101213. 10.1016/j.xcrm.2023.101213 37788667 PMC10591058 source

      The Blood glucose monitoring section says AI-based prediction can provide early warnings and that predicting near-future glucose levels enables better diabetes management. It reports predictive performance, but does not establish that these advances have already significantly improved diabetes monitoring and management in clinical practice. (LLM, conf=0.82)

      Experimental — accuracy not yet measured
  • May combine two claims. 50
  • Classified as probably not a substantial claim. 32

  • May combine two claims.

    Premisecitation

    De Silva et al.'s models achieved at least 70% area under the receiver operating characteristic curve, significantly outperforming the Centers for Disease Control and Prevention prediabetes screening tool.

    “Their models achieved ≥70% area under the receiver operating characteristic curve (AUROC), significantly outperforming the Centers for Disease Control and Prevention (CDC) prediabetes screening tool (34).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [34] De Silva K Jönsson D Demmer RT. A combined strategy of feature selection and machine learning to identify predictors of prediabetes. J Am Med Inform Assoc 2020; 27: 396- 406. 10.1093/jamia/ocz204 31889178 PMC7647289 source

  • May combine two claims.

    Premisecitation

    AI algorithms can predict the onset of type 2 diabetes and its complications with remarkable accuracy.

    “Various studies have shown that AI algorithms can predict the onset of T2D and its complications with remarkable accuracy (13-15).”
    Quote in the paperSubstantial claim2 of 3 sources checked
    • cites [13] Deberneh HM Kim I. Prediction of Type 2 Diabetes Based on Machine Learning Algorithm. Int J Environ Res Public Health 2021; 18: 3317. 10.3390/ijerph18063317 33806973 PMC8004981 source

      The Results (§4.2) report that the model predicted T2D occurrence in the following year, with a best test-set accuracy of 73%. The paper does not report predicting complications; its Conclusions (§6) mention them only as outcomes patients might mitigate using the prediction. It describes performance as “reasonably good,” not remarkable. (LLM, conf=0.89)

      Experimental — accuracy not yet measured
    • cites [14] Kopitar L Kocbek P Cilar L Early detection of type 2 diabetes mellitus using machine learning-based prediction models. Sci Rep 2020; 10: 11981. 10.1038/s41598-020-68771-z 32686721 PMC7371679 source

      The Results (Model performance; page unavailable) report AUCs up to 0.859 for predicting undiagnosed T2D, supporting the claim that the models predict diabetes with substantial accuracy. The paper does not report prediction accuracy for complications, and its study concerns detecting undiagnosed disease rather than prospectively predicting onset. (LLM, conf=0.86)

      Experimental — accuracy not yet measured
    • cites [15] Nguyen BP Pham HN Tran H Predicting the onset of type 2 diabetes using wide and deep learning with electronic health records. Comput Methods Programs Biomed 2019; 182: 105055. 10.1016/j.cmpb.2019.105055 31505379 source

  • May combine two claims.

    Premisecitation

    AI models can identify hidden trends not evident to human clinicians and alert healthcare providers to emerging complications before they progress, allowing earlier targeted interventions.

    “By identifying hidden trends that may not be evident to human clinicians, AI models can alert healthcare providers to emerging complications before they progress, allowing for earlier, more targeted interventions (21).”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [21] Giordano C Brennan M Mohamed B Accessing Artificial Intelligence for Clinical Decision-Making. Front Digit Health 2021; 3: 645232. 10.3389/fdgth.2021.645232 34713115 PMC8521931 source

  • May combine two claims.

    Premisecitation

    De Silva et al. used NHANES data, feature selection on 156 exposure variables, and four machine learning algorithms to identify 25 predictors of prediabetes.

    “De Silva et al. conducted a comprehensive analysis using the National Health and Nutrition Examination Survey (NHANES) data, employing feature selection algorithms on 156 exposure variables and applying four machine learning algorithms to identify 25 predictors of prediabetes (34).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [34] De Silva K Jönsson D Demmer RT. A combined strategy of feature selection and machine learning to identify predictors of prediabetes. J Am Med Inform Assoc 2020; 27: 396- 406. 10.1093/jamia/ocz204 31889178 PMC7647289 source

  • May combine two claims.

    Interpretationcitation

    This precision-medicine approach ensures that patients receive the most effective therapy, minimizing the risk of adverse effects and improving overall outcomes.

    “This precision-medicine approach ensures that patients receive the most effective therapy, minimizing the risk of adverse effects and improving overall outcomes (55).”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [55] Florez JC Pearson ER. A roadmap to achieve pharmacological precision medicine in diabetes. Diabetologia 2022; 65: 1830- 8. 10.1007/s00125-022-05732-3 35748917 PMC9522818 source

      In “Caveats: reality vs the ideal” (page not provided), the authors propose that selecting the best drug for an individual will lead to better outcomes. But “What is needed? > Evidence that such prediction leads to better clinical outcomes” says improved outcomes still need to be demonstrated. The claim overstates this proposed approach as one that already ensures the most effective therapy and improved outcomes; the cited body passages also do not establish that it minimizes adverse effects. (LLM, conf=0.88)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    Generative AI models are designed to predict and generate new information, offering unique insights into disease progression and complication risks, unlike traditional AI models focused on classification and regression.

    “Unlike traditional AI models that are focused on classification and regression tasks, some studies mentioned that generative models are designed to predict and generate new information, offering unique insights into disease progression and complication risks (44,45).”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites [44] Xie S Yu Z Lv Z. Multi-disease prediction based on deep learning: a survey. Computer Modeling in Engineering & Sciences 2021; 128: 489- 522.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites [45] Xie Y Lu L Gao F Integration of Artificial Intelligence, Blockchain, and Wearable Technology for Chronic Disease Management: A New Paradigm in Smart Healthcare. Curr Med Sci 2021; 41: 1123- 33. 10.1007/s11596-021-2485-0 34950987 PMC8702375 source

      The opening overview (section and page not supplied) describes AI analyzing wearable-device data to inform diagnosis and treatment, but does not discuss generative models. Searches for generative models, classification and regression, disease progression, and complication risks found no passage in our copy supporting the claimed comparison. (LLM, conf=0.75)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    A generative model trained on longitudinal data can predict the likelihood of a heart attack or stroke in a diabetic patient, offering clinicians insights that can inform preventative treatment plans.

    “A generative model trained on this data can predict the likelihood of a heart attack or stroke in a diabetic patient, offering clinicians valuable insights that can inform preventative treatment plans (50,51).”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites [50] Cichosz SL Johansen MD Hejlesen O. Toward Big Data Analytics: Review of Predictive Models in Management of Diabetes and Its Complications. J Diabetes Sci Technol 2015; 10: 27- 34. 10.1177/1932296815611680 26468133 PMC4738225 source

      The Predictive Models section says models can estimate an individual’s risk and can use time-series data; Heart Disease describes risk scores for patients with diabetes, and the references include stroke-risk models. The paper does not identify a generative model trained on longitudinal data for these outcomes. Heart Disease also cautions that most such models have not been proven valuable in daily care, so their use to inform preventive treatment plans remains unestablished. (LLM, conf=0.86)

      Experimental — accuracy not yet measured
    • cites [51] Khalifa M Albadawy M. Artificial intelligence for clinical prediction: exploring key domains and essential functions. Computer Methods and Programs in Biomedicine Update 2024; 5: 100148.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Premisecitation

    NHANES data has enabled the development of sophisticated machine learning algorithms for prediabetes prediction, with studies demonstrating superior performance compared to traditional screening tools through feature selection techniques.

    “NHANES data has enabled the development of sophisticated machine learning algorithms for prediabetes prediction, with studies demonstrating superior performance compared to traditional screening tools through feature selection techniques applied to comprehensive exposure variables (34-36).”
    Quote in the paperNot a substantial claim1 of 3 sources checked
    • cites [34] De Silva K Jönsson D Demmer RT. A combined strategy of feature selection and machine learning to identify predictors of prediabetes. J Am Med Inform Assoc 2020; 27: 396- 406. 10.1093/jamia/ocz204 31889178 PMC7647289 source
    • cites [35] Buccheri E Dell'Aquila D Russo M. Artificial intelligence in health data analysis: The Darwinian evolution theory suggests an extremely simple and zero-cost large-scale screening tool for prediabetes and type 2 diabetes. Diabetes Res Clin Pract 2021; 174: 108722. 10.1016/j.diabres.2021.108722 33647331 source
    • cites [36] Buccheri E Dell'Aquila D Russo M. Stratified analysis of the age-related waist circumference cut-off model for the screening of dysglycemia at zero-cost. Obesity Medicine 2022; 31: 100398.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Classified as probably not a substantial claim.

    Premisecitation

    Generative AI and LLMs are emerging as critical tools in revolutionizing diabetes care.

    “Generative AI and LLMs are emerging as critical tools in revolutionizing diabetes care (68).”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [68] Vehi J Mujahid O Beneyto A Generative artificial intelligence in diabetes healthcare. iScience 2025; 28: 113051. 10.1016/j.isci.2025.113051 40703444 PMC12284359 source

  • May combine two claims.

    Premisecitation

    NHANES data has supported the creation of innovative AI-driven screening tools for sarcopenia and muscle mass assessment, demonstrating that anthropometric measurements can effectively substitute costly radiological examinations whilst maintaining high diagnostic accuracy.

    “Furthermore, NHANES data has supported the creation of innovative AI-driven screening tools for sarcopenia and muscle mass assessment, demonstrating how anthropometric measurements can effectively substitute costly radiological examinations whilst maintaining high diagnostic accuracy (66,67).”
    Quote in the paperNot a substantial claim1 of 2 sources checked
    • cites [66] Buccheri E Dell'Aquila D Russo M Can artificial intelligence simplify the screening of muscle mass loss? Heliyon 2023; 9: e16323. 10.1016/j.heliyon.2023.e16323 37251872 PMC10208931 source

      Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

      Experimental — accuracy not yet measured
    • cites [67] Buccheri E Dell'Aquila D Russo M Appendicular Skeletal Muscle Mass in Older Adults Can Be Estimated With a Simple Equation Using a Few Zero-Cost Variables. J Geriatr Phys Ther 2024; 47: E149- 58. 10.1519/JPT.0000000000000420 39079022 source

  • May combine two claims.

    Interpretationcitation

    Real-time analysis by LLMs ensures clinicians work with up-to-date knowledge and can tailor treatment plans accordingly.

    “This type of real-time analysis ensures that clinicians are working with the most up-to-date knowledge and can tailor their treatment plans accordingly (85).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [85] Parthasarathy VB, Zafar A, Khan A, et al. The ultimate guide to fine-tuning llms from basics to breakthroughs: An exhaustive review of technologies, research, best practices, applied research challenges and opportunities. arXiv 2024. arXiv:2408.13296. source

      Section 12.4.1 (p. 104) says LLMs can use real-time data to provide personalised healthcare recommendations, supporting the idea of tailoring care. Section 9.3 (p. 76) says LLM knowledge can become outdated and requires updating. Neither passage establishes that real-time analysis ensures clinicians have the most up-to-date knowledge or tailor treatment plans accordingly. (LLM, conf=0.84)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    LLMs such as GPT and BERT are trained on massive datasets and can understand both structured and unstructured text.

    “LLMs, such as GPT and BERT, are trained on massive datasets and have the unique capability of understanding both structured and unstructured text (11,81).”
    Quote in the paperNot a substantial claim1 of 2 sources checked
    • cites [11] Raiaan MAK, Mukta MSH, Fatema K, et al. A Review on Large Language Models: Architectures, Applications, Taxonomies, Open Issues and Challenges. IEEE Access 2024;12:26839-74.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites [81] AlSaad R Abd-Alrazaq A Boughorbel S Multimodal Large Language Models in Health Care: Applications, Challenges, and Future Outlook. J Med Internet Res 2024; 26: e59505. 10.2196/59505 39321458 PMC11464944 source

      The Introduction (page unavailable) says ChatGPT was trained on a vast collection of text data, and the Applications > Text section describes pretrained BERT variants. But Applications > Text says that understanding structured textual data requires additional skills in M-LLMs; it does not establish that GPT and BERT generally understand both structured and unstructured text. (LLM, conf=0.87)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Interpretationcitation

    This integration enhances the ability of clinicians to detect early signs of complications, offering potential for more precise and personalized interventions.

    “This integration enhances the ability of clinicians to detect early signs of complications, offering the potential for more precise and personalized interventions (87).”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [87] Ahmed Z Mohamed K Zeeshan S Artificial intelligence with multi-functional machine learning platform development for better healthcare and precision medicine. Database (Oxford) 2020; 2020: baaa010. 10.1093/database/baaa010 32185396 PMC7078068 source

      The Introduction (page not provided) says integrating patient data can support earlier interventions, advanced diagnostics, and personalized treatment. Section 3.2 also describes using EHR patterns to help physicians anticipate future events in high-risk patients. Neither passage specifically establishes that the integration enhances clinicians’ detection of early signs of complications. (LLM, conf=0.83)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Interpretationcitation

    By providing real-time predictions and personalized treatment recommendations, LLMs have the potential to drastically improve patient outcomes and reduce the burden of diabetes-related complications.

    “By providing real-time predictions and personalized treatment recommendations, LLMs have the potential to drastically improve patient outcomes and reduce the burden of diabetes-related complications (31).”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [31] Guan Z Li H Liu R Artificial intelligence in diabetes management: Advancements, opportunities, and challenges. Cell Rep Med 2023; 4: 101213. 10.1016/j.xcrm.2023.101213 37788667 PMC10591058 source

      The Conclusion and perspective section says LLMs have the potential to personalize treatment and improve outcomes for patients with diabetes. It does not attribute real-time predictions or a reduced burden of diabetes-related complications to LLMs; the complication-prediction examples in the paper concern other AI methods. It also does not characterize the potential improvement in outcomes as drastic. (LLM, conf=0.87)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    These models can analyze retinal images in greater detail than human specialists, identifying patterns that precede visible symptoms of retinopathy.

    “These models can analyze retinal images in greater detail than human specialists, identifying patterns that precede the visible symptoms of retinopathy (68).”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [68] Vehi J Mujahid O Beneyto A Generative artificial intelligence in diabetes healthcare. iScience 2025; 28: 113051. 10.1016/j.isci.2025.113051 40703444 PMC12284359 source

      The “Density estimation using generative artificial intelligence” section (page unavailable) describes models extracting retinal-image features, detecting anomalies, and aiding diabetic-retinopathy diagnosis. It does not say they analyze images in greater detail than human specialists or identify patterns before visible symptoms appear. (LLM, conf=0.86)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    The AI model was able to predict the likelihood of disease progression, improve image collection quality, provide clinical reference, and facilitate diabetic retinopathy screening with an accuracy rate that rivaled expert ophthalmologists.

    “The AI model was able to predict the likelihood of disease progression, improve image collection quality, provide clinical reference, and facilitate diabetic retinopathy screening with an accuracy rate that rivaled expert ophthalmologists, offering the potential for earlier intervention.”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites [92] Dai L Sheng B Chen T A deep learning system for predicting time to progression of diabetic retinopathy. Nat Med 2024; 30: 584- 94. 10.1038/s41591-023-02702-z 38177850 PMC10878973 source

      The Results ('DeepDR Plus predicts time to DR progression'; page unavailable) report that the model predicted progression risk, and the Discussion describes its potential to guide personalized screening and clinical management. The Methods ('Image quality control and grading procedure'; page unavailable) describe ophthalmologists grading images and the exclusion of ungradable images, not AI improving image-collection quality. The paper does not report an accuracy comparison with expert ophthalmologists, and the Discussion says the system had not been applied in clinical practice. (LLM, conf=0.88)

      Experimental — accuracy not yet measured
    • cites [93] Dai L Wu L Li H A deep learning system for detecting diabetic retinopathy across the disease spectrum. Nat Commun 2021; 12: 3242. 10.1038/s41467-021-23458-5 34050158 PMC8163820 source

      The Discussion (page not provided) says DeepDR can improve image collection quality, provide clinical reference, and facilitate diabetic retinopathy screening; the Results report high grading performance. But the Discussion says further studies are needed to evaluate prediction of disease progression, so the claim that the model could already predict it is unsupported. The passages found do not establish that its accuracy rivaled expert ophthalmologists. (LLM, conf=0.90)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Diabetic retinopathy is one of the leading causes of blindness among people with diabetes, and its progression can be asymptomatic until irreversible damage has occurred.

    “Diabetic retinopathy is one of the leading causes of blindness among people with diabetes, and its progression can be asymptomatic until irreversible damage has occurred (6).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [6] Kropp M Golubnitschaja O Mazurakova A Diabetic retinopathy as the leading cause of blindness and early predictor of cascading complications-risks and mitigation. EPMA J 2023; 14: 21- 42. 10.1007/s13167-023-00314-8 36866156 PMC9971534 source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Diabetic neuropathy is a common complication of diabetes, characterized by nerve damage resulting from prolonged high blood glucose levels.

    “Diabetic neuropathy is a common complication of diabetes, characterized by nerve damage resulting from prolonged high blood glucose levels (94).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [94] Pang L Lian X Liu H Understanding Diabetic Neuropathy: Focus on Oxidative Stress. Oxid Med Cell Longev 2020; 2020: 9524635. 10.1155/2020/9524635 32832011 PMC7422494 source

  • May combine two claims.

    Interpretationcitation

    Generative AI enables predictive modeling that enhances early diagnosis, personalized treatment plans, and proactive healthcare interventions in chronic disease management.

    “It also revolutionizes chronic disease management by enabling predictive modeling that enhances early diagnosis, personalized treatment plans, and proactive healthcare interventions (25).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [25] Alapati NK Valleru V. Leveraging AI for Predictive Modeling in Chronic Disease Management. Innovative Computer Sciences Journal 2023; 9: 5.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationreasoning

    This proactive approach improves patient outcomes and reduces the long-term healthcare burden by preventing severe complications from developing.

    “This proactive approach not only improves patient outcomes but also reduces the long-term healthcare burden by preventing severe complications from developing.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Study design and conductnone given

    Preference was given to high-impact journals and studies with robust methodologies or significant relevance to real-world clinical practice.

    “However, preference was given to high-impact journals and studies with robust methodologies or significant relevance to real-world clinical practice.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    There is a gap in early detection tools, which currently rely heavily on standard biomarkers and observable symptoms, potentially missing subtle yet critical changes that signal the onset of complications.

    “The statement of the problem focuses on the gap in early detection tools, which currently rely heavily on standard biomarkers and observable symptoms, potentially missing subtle yet critical changes that signal the onset of complications.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    AI applications in diabetes management range from supervised machine learning models that predict disease onset to tools that enhance patient self-management.

    “AI applications in this field range from supervised machine learning models that predict disease onset to tools that enhance patient self-management (30).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [30] Chaki J Ganesh ST Cidham SK Machine learning and artificial intelligence based Diabetes Mellitus detection and self-management: A systematic review. Journal of King Saud University-Computer and Information Sciences 2022; 34: 3204- 25.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationreasoning

    Despite significant advancements in diabetes care, complications such as diabetic retinopathy, neuropathy, and cardiovascular diseases often remain undetected until advanced stages, leading to irreversible damage and increased morbidity.

    “Despite significant advancements in diabetes care, complications such as diabetic retinopathy, neuropathy, and CVDs often remain undetected until they have reached advanced stages, leading to irreversible damage and increased morbidity.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationreasoning

    Integration of generative AI and LLMs into clinical decision support systems and precision medicine frameworks may reduce the global burden of diabetes, improve patient outcomes, and shift care from reactive to preventative.

    “Their integration into clinical decision support systems (CDSS) and precision medicine frameworks may reduce the global burden of diabetes, improve patient outcomes, and shift care from reactive to preventative.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    Quality assessment was not formally conducted due to heterogeneity of the sources included.

    “Quality assessment was not formally conducted due to the heterogeneity of the sources included (e.g., primary studies, technical papers, and reviews).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Managing diabetes complications effectively requires early detection and timely interventions because they progress silently before severe clinical outcomes.

    “Managing these complications effectively requires early detection and timely interventions, as they tend to progress silently before manifesting into more severe clinical outcomes (4).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [4] Lim LL Chow E Chan JCN. Cardiorenal diseases in type 2 diabetes mellitus: clinical trials and real-world practice. Nat Rev Endocrinol 2023; 19: 151- 63. 10.1038/s41574-022-00776-2 36446898 source

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

    Interpretationreasoning

    Generative AI and LLMs are poised to redefine diabetes management by enabling earlier detection of complications and personalised treatment strategies.

    “Generative AI and LLMs are poised to redefine diabetes management by enabling earlier detection of complications and personalised treatment strategies.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    AI algorithms are capable of identifying patients at risk long before they display clinical symptoms, enabling earlier interventions and personalised care strategies.

    “These algorithms are capable of identifying patients at risk long before they display clinical symptoms, enabling earlier interventions and personalised care strategies.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Recent studies utilising UK Biobank data have demonstrated that AI-powered electrocardiogram analysis can predict heart failure risk from single-lead recordings, showing potential for scalable community-based risk assessment using portable devices.

    “Recent studies utilising UK Biobank data have demonstrated how AI-powered electrocardiogram analysis can predict heart failure risk from single-lead recordings, showcasing the potential for scalable community-based risk assessment using portable devices (65).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [65] Dhingra LS Aminorroaya A Pedroso AF Artificial Intelligence-Enabled Prediction of Heart Failure Risk From Single-Lead Electrocardiograms. JAMA Cardiol 2025; 10: 574- 84. 10.1001/jamacardio.2025.0492 40238120 PMC12004248 source

  • May combine two claims.

    Premisecitation

    Generative AI models can analyze multi-dimensional datasets specific to an individual, unlike traditional risk calculators that rely on generalized population data.

    “While traditional risk calculators rely on generalized population data, generative AI models can analyze multi-dimensional datasets specific to an individual (19).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [19] Liu Y Yang Z Yu Z Generative artificial intelligence and its applications in materials science: Current situation and future perspectives. Journal of Materiomics 2023; 9: 798- 816.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Premisecitation

    Continuous glucose monitors, smart insulin pens, and AI-driven smartphone apps allow patients to monitor blood sugar levels, activity, diet, and medication adherence in real-time.

    “Continuous glucose monitors (CGMs), smart insulin pens, and AI-driven smartphone apps allow patients to monitor their blood sugar levels, activity, diet, and medication adherence in real-time (38).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [38] Bergoeing M, Neyem A, Jofré P, et al. Exploring the Potential of an AI-Integrated Cloud-Based mHealth Platform for Enhanced Type 2 Diabetes Mellitus Management. In: Bravo J, Urzáiz G. editors. Proceedings of the 15th International Conference on Ubiquitous Computing & Ambient Intelligence (UCAmI 2023). Cham: Springer Nature Switzerland; 2023:100-111. source

  • May combine two claims.

    Premisecitation

    Continuous glucose monitors such as Dexcom G6 and Abbott FreeStyle Libre continuously measure interstitial glucose levels and use AI algorithms to predict blood sugar trends, allowing users to take preventative action to avoid hypo- or hyperglycemia.

    “For example, CGMs such as the Dexcom G6 and Abbott FreeStyle Libre continuously measured interstitial glucose levels and use AI algorithms to predict blood sugar trends, allowing users to take preventative action to avoid hypo- or hyperglycemia (39,40).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [39] Swapna G, Soman KP. Diabetes detection and sensor-based continuous glucose monitoring–a deep learning approach. In: Chakraborty C, Ghosh U, Ravi V, et al. editors. Efficient Data Handling for Massive Internet of Medical Things: Healthcare Data Analytics. Cham: Springer International Publishing; 2021:245-68. source
    • cites [40] Mian Z Hermayer KL Jenkins A. Continuous Glucose Monitoring: Review of an Innovation in Diabetes Management. Am J Med Sci 2019; 358: 332- 9. 10.1016/j.amjms.2019.07.003 31402042 source

  • May combine two claims.

    Premisecitation

    NHANES collects comprehensive health and nutrition data from diverse communities across the United States through interviews, health examinations, and laboratory tests, providing nationally representative samples essential for developing generalisable AI models.

    “NHANES, which collects comprehensive health and nutrition data from diverse communities across the United States through interviews, health examinations, and laboratory tests, provides researchers with nationally representative samples essential for developing generalisable AI models (62).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [62] Centers for Disease Control and Prevention. National Health and Nutrition Examination Survey (NHANES). Available online: https://www.cdc.gov/nchs/nhanes/index.html

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Premisecitation

    The age-related waist circumference model achieved consistent area under the curve values of 0.69-0.78 across sex and ethnic groups, demonstrating applicability as a universal screening tool for dysglycemia in diverse populations.

    “Further validation through stratified analysis across different demographic groups confirmed the robustness of the age-related waist circumference model, achieving consistent AUC values of 0.69–0.78 across sex and ethnic groups, demonstrating its applicability as a universal screening tool for dysglycemia in diverse populations (36).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [36] Buccheri E Dell'Aquila D Russo M. Stratified analysis of the age-related waist circumference cut-off model for the screening of dysglycemia at zero-cost. Obesity Medicine 2022; 31: 100398.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Large-scale, population-representative datasets serve as the cornerstone for developing robust AI models in healthcare, with NHANES and UK Biobank as exemplary resources.

    “Large-scale, population-representative datasets serve as the cornerstone for developing robust AI models in healthcare, with the NHANES and UK Biobank standing as exemplary resources that have revolutionised AI-driven medical research (62,63).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [62] Centers for Disease Control and Prevention. National Health and Nutrition Examination Survey (NHANES). Available online: https://www.cdc.gov/nchs/nhanes/index.html

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites [63] UK Biobank. UK Biobank: enabling scientific discoveries that improve human health. Available online: https://www.ukbiobank.ac.uk/

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    UK Biobank's extensive biomedical database, encompassing half a million participants since 2006, offers researchers worldwide access to imaging data, biomarkers, genetic information, healthcare records, and lifestyle data through its secure Research Analysis Platform.

    “Similarly, UK Biobank’s extensive biomedical database, encompassing half a million participants since 2006, offers researchers worldwide access to imaging data, biomarkers, genetic information, healthcare records, and lifestyle data through its secure Research Analysis Platform (63).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [63] UK Biobank. UK Biobank: enabling scientific discoveries that improve human health. Available online: https://www.ukbiobank.ac.uk/

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationreasoning

    The longitudinal nature of UK Biobank's follow-up data enables researchers to track disease progression over time, providing insights into temporal patterns that AI models can learn to recognise early warning signs of cardiovascular complications in diabetic patients.

    “The longitudinal nature of UK Biobank’s follow-up data enables researchers to track disease progression over time, providing crucial insights into the temporal patterns that AI models can learn to recognise early warning signs of cardiovascular complications in diabetic patients.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Diabetes is influenced by various factors, including genetics, lifestyle, metabolic changes, and environmental exposures.

    “Diabetes is influenced by various factors, including genetics, lifestyle, metabolic changes, and environmental exposures, making it difficult for conventional models to predict outcomes accurately (72).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [72] Nazirun NN Wahab AA Selamat A Prediction Models for Type 2 Diabetes Progression: A Systematic Review. IEEE Access. 2024; 12: 161595- 619.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    The ability of generative AI and LLMs to analyze diverse and complex data sources, ranging from structured datasets like medical records to unstructured data like patient notes, sets them apart from traditional AI models.

    “Their ability to analyze diverse and complex data sources, ranging from structured datasets like medical records to unstructured data like patient notes, sets them apart from traditional AI models.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    LLMs' ability to synthesize vast amounts of patient data into actionable insights reduces the cognitive load on clinicians, allowing them to focus on more complex tasks.

    “Their ability to synthesize vast amounts of patient data into actionable insights reduces the cognitive load on clinicians, allowing them to focus on more complex tasks that require human judgment (90).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [90] Zhuhadar LP Lytras MD. The Application of AutoML Techniques in Diabetes Diagnosis: Current Approaches, Performance, and Future Directions. Sustainability 2023; 15: 13484.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationreasoning

    By recognizing early patterns, clinicians were able to implement preventative measures such as medication adjustments and lifestyle interventions, potentially slowing the progression of the disease.

    “By recognizing these early patterns, clinicians were able to implement preventative measures, such as medication adjustments and lifestyle interventions, potentially slowing the progression of the disease.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationcitation

    Generative AI predictions of retinal changes can be made well before the onset of symptoms, allowing for earlier interventions that could prevent vision loss.

    “These predictions were made well before the onset of symptoms, allowing for earlier interventions that could prevent vision loss.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Generative AI can integrate multiple data points and produce highly specific and personalized predictions for each patient.

    “However, studies showed that generative AI can integrate these data points and produce highly specific and personalized predictions for each patient (73).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [73] Zhang P Kamel Boulos MN. Generative AI in medicine and healthcare: promises, opportunities and challenges. Future Internet 2023; 15: 286.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationreasoning

    An LLM could cross-reference a patient's genetic profile with the latest findings from genetic research to predict their risk for specific diabetes-related complications, such as nephropathy or CVD.

    “In clinical practice, this means that an LLM could, for example, cross-reference a patient’s genetic profile with the latest findings from genetic research to predict their risk for specific diabetes-related complications, such as nephropathy or CVD.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Generative AI models are designed to analyze vast amounts of data and generate meaningful outputs by learning from underlying patterns.

    “Generative AI models are designed to analyze vast amounts of data and generate meaningful outputs by learning from underlying patterns (70).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [70] Bandi A Adapa PVSR Kuchi YEVPK. The Power of Generative AI: A Review of Requirements, Models, Input–Output Formats, Evaluation Metrics, and Challenges. Future Internet 2023; 15: 260.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationreasoning

    Without foundational datasets like NHANES and UK Biobank, current advances in generative AI for diabetes care would not be possible, as these models require vast amounts of high-quality, representative data to learn complex patterns that enable accurate prediction and personalised interventions.

    “Without such foundational datasets, the current advances in generative AI for diabetes care would not be possible, as these models require vast amounts of high-quality, representative data to learn the complex patterns that enable accurate prediction and personalised interventions.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationreasoning

    The comprehensive nature and rigorous data collection protocols of NHANES and UK Biobank ensure that AI models developed from them possess the statistical power and external validity necessary for clinical implementation across diverse populations.

    “The comprehensive nature and rigorous data collection protocols of these datasets ensure that AI models developed from them possess the statistical power and external validity necessary for clinical implementation across diverse populations.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Interpretationreasoning

    By including genetic information related to susceptibility to nerve damage, AI can further refine predictions, allowing clinicians to tailor treatment plans that mitigate risk of severe nerve impairment.

    “By including genetic information related to a patient’s susceptibility to nerve damage, AI can further refine predictions, allowing clinicians to tailor treatment plans that mitigate the risk of severe nerve impairment.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    CVD is the leading cause of death among individuals with diabetes, with complications such as heart attacks and strokes being major contributors to diabetes-related morbidity and mortality.

    “CVD is the leading cause of death among individuals with diabetes, with complications such as heart attacks and strokes being major contributors to diabetes-related morbidity and mortality (98).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [98] Almourani R Chinnakotla B Patel R Diabetes and Cardiovascular Disease: an Update. Curr Diab Rep 2019; 19: 161. 10.1007/s11892-019-1239-x 31828525 source

  • May combine two claims.

    Premisecitation

    A study demonstrated that AI-driven analysis of nerve conduction and sensory data could predict neuropathy before symptoms became debilitating, allowing for earlier interventions such as medication adjustments or lifestyle changes.

    “A study demonstrated how AI-driven analysis of nerve conduction and sensory data could predict neuropathy before symptoms became debilitating, allowing for earlier interventions such as medication adjustments or lifestyle changes (96).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [96] Joshi S, Rout PK, Samanta IS, et al. Advancements and Challenges in AI and ML-Based Applications for Diabetic Neuropathy: Current State and Future Directions. 2023 2nd International Conference on Ambient Intelligence in Health Care (ICAIHC). Bhubaneswar: IEEE; 2023:1-6. source

  • May combine two claims.

    Premisecitation

    Generative AI models can incorporate genetic predispositions to CVDs, which further enhances their predictive accuracy.

    “Additionally, generative AI models can incorporate genetic predispositions to CVDs, which further enhance their predictive accuracy (76).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [76] Abbasi N Nizamullah FNU Zeb S Generative AI in healthcare: revolutionizing disease diagnosis, expanding treatment options, and enhancing patient care. Journal of Knowledge Learning and Science Technology 2024; 3: 127- 38.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Premisecitation

    Traditional risk calculators often provide generalized risk assessments based on standard biomarkers but may not account for the complexity of individual risk profiles.

    “Traditional risk calculators often provide generalized risk assessments based on standard biomarkers (e.g., cholesterol levels, blood pressure) but may not account for the complexity of individual risk profiles (99).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [99] Mohd Faizal AS Thevarajah TM Khor SM A review of risk prediction models in cardiovascular disease: conventional approach vs. artificial intelligent approach. Comput Methods Programs Biomed 2021; 207: 106190. 10.1016/j.cmpb.2021.106190 34077865 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given

    Much of the current research on generative AI in healthcare, including diabetes care, is still emerging.

    “much of the current research on generative AI in healthcare, including diabetes care, is still emerging.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    Relevant studies published in other languages or not indexed in the selected databases may have been overlooked, possibly limiting the global perspective of the findings.

    “As a result, relevant studies published in other languages or not indexed in the selected databases may have been overlooked, possibly limiting the global perspective of the findings.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    Ongoing advancements in AI technology are expected to address challenges related to data availability, model transparency, and ethical considerations.

    “While challenges related to data availability, model transparency, and ethical considerations remain, ongoing advancements in AI technology are expected to address these issues.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

    Interpretationreasoning

    Generative AI is poised to shift diabetes care from a reactive to a proactive approach, with individualized treatments based on a patient's unique genetic, lifestyle, and clinical data.

    “As generative AI continues to evolve, it is poised to shift diabetes care from a reactive to a proactive approach, where individualized treatments are based on a patient’s unique genetic, lifestyle, and clinical data.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductnone given

    The review did not include a formal risk-benefit analysis or stakeholder perspectives from patients, clinicians, or developers, which are critical for real-world implementation.

    “while this review aimed to highlight the opportunities and challenges associated with generative AI, it did not include a formal risk-benefit analysis or stakeholder perspectives from patients, clinicians, or developers, which are critical for real-world implementation.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Interpretationnone given

    The field of generative AI is rapidly evolving.

    “the field of generative AI is rapidly evolving.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    The integration of AI into CDSS, personalized medicine, and patient self-management platforms will further enhance the precision and effectiveness of diabetes management.

    “The integration of AI into CDSS, personalized medicine, and patient self-management platforms will further enhance the precision and effectiveness of diabetes management.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    Generative AI, particularly LLMs, is revolutionizing the management of diabetes and its complications.

    “Generative AI, particularly LLMs, is revolutionizing the management of diabetes and its complications.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisenone given

    Generative AI can simulate the effects of potential drug compounds on diabetic patients by modeling complex interactions between drugs and biological systems, predicting efficacy, side effects, and optimal dosing regimens before clinical trials.

    “Generative AI can simulate the effects of potential drug compounds on diabetic patients by modeling complex interactions between the drugs and biological systems, predicting efficacy, side effects, and optimal dosing regimens before clinical trials even begin.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    In the years to come, generative AI will be integral to reducing the global burden of diabetes and its associated complications.

    “In the years to come, this transformative technology will be integral to reducing the global burden of diabetes and its associated complications.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductreasoning

    New developments, models, and applications may have emerged after the time of writing, potentially limiting the review's ability to capture the most recent innovations and technologies.

    “New developments, models, and applications may have emerged after the time of writing, potentially limiting the review’s ability to capture the most recent innovations and technologies.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationreasoning

    Generative AI enables earlier detection of complications such as diabetic retinopathy, neuropathy, and CVDs, allowing for more timely and personalized interventions.

    “This capability enables earlier detection of complications such as diabetic retinopathy, neuropathy, and CVDs, allowing for more timely and personalized interventions.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationreasoning

    AI-driven tools are already beginning to empower both clinicians and patients with real-time insights, improving glycemic control, reducing complications, and enhancing overall quality of life.

    “AI-driven tools are already beginning to empower both clinicians and patients with real-time insights, improving glycemic control, reducing complications, and enhancing overall quality of life.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductreasoning

    The review's methodology, as a narrative review, lacks the rigorous systematic framework typically used in systematic reviews or meta-analyses.

    “as a narrative review, the methodology lacks the rigorous systematic framework typically used in systematic reviews or meta-analyses.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationreasoning

    By analyzing vast, multi-dimensional datasets, generative AI models are giving clinicians new insights into previously undetectable patterns of disease progression, metabolic dysfunction, and risk factors.

    “By harnessing the power of these advanced models to analyze vast, multi-dimensional datasets, clinicians are gaining new insights into the hidden patterns of disease progression, metabolic dysfunction, and risk factors that were previously undetectable.”
    Quote in the paperNot a substantial claim
68 of 68

  • The paper relies on this without support.

    The included literature is sufficiently representative and reliable to support conclusions about generative AI in diabetes care despite the absence of formal quality assessment.

    “Quality assessment was not formally conducted due to the heterogeneity of the sources included (e.g., primary studies, technical papers, and reviews).”
    “No formal quality appraisal due to source heterogeneity; preference given to studies from higher-impact venues or with robust methods and real-world relevance.”

  • The paper relies on this without support.

    Conventional diagnostic tools for diabetic neuropathy, such as nerve conduction studies, are often used only after significant nerve damage has developed.

    “Conventional diagnostic tools, such as nerve conduction studies, are often used after significant nerve damage has developed.”

  • The paper relies on this without support.

    The multifactorial nature of cardiovascular complications in diabetes makes prediction and early intervention particularly challenging.

    “The multifactorial nature of cardiovascular complications in diabetes, driven by a combination of hyperglycemia, hypertension, dyslipidemia, and lifestyle factors, makes prediction and early intervention particularly challenging.”
3 of 3

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • The cited LLM study predicts diabetes onset, not neuropathy

    Minor

    The review describes reference 88 as a study predicting diabetic neuropathy and detecting early nerve damage. The cited paper's title instead identifies its outcome as new-onset type 2 diabetes.

1 of 1

Cites

The publisher hasn’t deposited this paper’s reference list.

Research works printed but not linked: 39

  1. [1]Garg P Duggal N. Type 2 diabetes mellitus, its impact on quality of life and how the disease can be managed-a review. Obesity Medicine 2022; 35: 100459.
  2. [5]Cutaș A Drugan C Roman G Inflammatory response and timeline of chronic complications in patients with type 1 and 2 diabetes mellitus. International Journal of Diabetes in Developing Countries 2020; 40: 561- 9.
  3. [8]Hussain HK Tariq A Gill AY Transforming healthcare: The rapid rise of artificial intelligence revolutionizing healthcare applications. BULLET: Jurnal Multidisiplin Ilmu 2022; 1: 592216.
  4. [9]García-Peñalvo FJ Vázquez-Ingelm A. What do we mean by GenAI? A systematic mapping of the evolution, trends, and techniques involved in Generative AI. IJIMAI 2023; 8: 7- 16.
  5. [10]Ghebrehiwet I Zaki N Damseh R Revolutionizing personalized medicine with generative AI: a systematic review. Artificial Intelligence Review 2024; 57: 128.
  6. [11]Raiaan MAK, Mukta MSH, Fatema K, et al. A Review on Large Language Models: Architectures, Applications, Taxonomies, Open Issues and Challenges. IEEE Access 2024;12:26839-74.
  7. [19]Liu Y Yang Z Yu Z Generative artificial intelligence and its applications in materials science: Current situation and future perspectives. Journal of Materiomics 2023; 9: 798- 816.
  8. [22]Salehin I Tomal Ahmed Sajib M Huda Badhon N Systematic Literature Review of LLM‐Large Language Model in Medical: Digital Health, Technology and Applications. Engineering Reports 2025; 7: e70365.
  9. [25]Alapati NK Valleru V. Leveraging AI for Predictive Modeling in Chronic Disease Management. Innovative Computer Sciences Journal 2023; 9: 5.
  10. [26]Gonçalves H Silva F Rodrigues C Navigating the Digital Landscape of Diabetes Care: Current State of the Art and Future Directions. Procedia Computer Science 2024; 237: 336- 43.
  11. [27]Adetunji O Evah PT. Developing personalized diabetes management plans using artificial intelligence and machine learning. World Journal of Advanced Research and Reviews 2022; 13: 605- 24.
  12. [30]Chaki J Ganesh ST Cidham SK Machine learning and artificial intelligence based Diabetes Mellitus detection and self-management: A systematic review. Journal of King Saud University-Computer and Information Sciences 2022; 34: 3204- 25.
  13. [33]Datta D Bhattacharya M Rajest SS Development of predictive model of diabetic using supervised machine learning classification algorithm of ensemble voting. International Journal of Bioinformatics Research and Applications 2023; 19: 151- 69.
  14. [36]Buccheri E Dell'Aquila D Russo M. Stratified analysis of the age-related waist circumference cut-off model for the screening of dysglycemia at zero-cost. Obesity Medicine 2022; 31: 100398.
  15. [44]Xie S Yu Z Lv Z. Multi-disease prediction based on deep learning: a survey. Computer Modeling in Engineering & Sciences 2021; 128: 489- 522.
  16. [48]Xu Q Wang L Sansgiry SS. A systematic literature review of predicting diabetic retinopathy, nephropathy and neuropathy in patients with type 1 diabetes using machine learning. J Med Artif Intell 2020; 3: 6.
  17. [49]Bansal V Jain A Walia NK. Diabetic retinopathy detection through generative AI techniques: A review. Results in Optics 2024; 16: 100700.
  18. [51]Khalifa M Albadawy M. Artificial intelligence for clinical prediction: exploring key domains and essential functions. Computer Methods and Programs in Biomedicine Update 2024; 5: 100148.
  19. [52]Kopitar L Fister I Jr Stiglic G. Using Generative AI to Improve the Performance and Interpretability of Rule-Based Diagnosis of Type 2 Diabetes Mellitus. Information 2024; 15: 162.
  20. [69]Asif S Wenhui Y, ur-Rehman S, et al. Advancements and prospects of machine learning in medical diagnostics: unveiling the future of diagnostic precision. Archives of Computational Methods in Engineering 2025; 32: 853- 83.
  21. [70]Bandi A Adapa PVSR Kuchi YEVPK. The Power of Generative AI: A Review of Requirements, Models, Input–Output Formats, Evaluation Metrics, and Challenges. Future Internet 2023; 15: 260.
  22. [72]Nazirun NN Wahab AA Selamat A Prediction Models for Type 2 Diabetes Progression: A Systematic Review. IEEE Access. 2024; 12: 161595- 619.
  23. [73]Zhang P Kamel Boulos MN. Generative AI in medicine and healthcare: promises, opportunities and challenges. Future Internet 2023; 15: 286.
  24. [74]Helen D, Suresh NV. Generative AI in Healthcare: Opportunities, Challenges, and Future Perspectives. In: Singla B, Shalender K, Stamer K. editors. Revolutionizing the Healthcare Sector with AI. IGI Global; 2024:79-90.
  25. [76]Abbasi N Nizamullah FNU Zeb S Generative AI in healthcare: revolutionizing disease diagnosis, expanding treatment options, and enhancing patient care. Journal of Knowledge Learning and Science Technology 2024; 3: 127- 38.
  26. [79]Nazi ZA Peng W. Large language models in healthcare and medical domain: A review. Informatics 2024; 11: 57.
  27. [82]Nassiri K Akhloufi MA. Recent Advances in Large Language Models for Healthcare. BioMedInformatics 2024; 4: 1097- 143.
  28. [83]Prabhod KJ. Integrating large language models for enhanced clinical decision support systems in modern healthcare. Journal of Machine Learning for Healthcare Decision Support 2023; 3: 18- 62.
  29. [84]De Vito G. Assessing healthcare software built using IoT and LLM technologies. Proceedings of the 28th International Conference on Evaluation and Assessment in Software Engineering. New York, NY, USA: Association for Computing Machinery; 2024:476-8.
  30. [86]Prabhod KJ. Leveraging Generative AI for Personalized Medicine: Applications in Drug Discovery and Development. Journal of AI-Assisted Scientific Discovery 2023; 3: 392- 434.
  31. [90]Zhuhadar LP Lytras MD. The Application of AutoML Techniques in Diabetes Diagnosis: Current Approaches, Performance, and Future Directions. Sustainability 2023; 15: 13484.
  32. [91]Mayya V Kamath S Kulkarni U. Automated microaneurysms detection for early diagnosis of diabetic retinopathy: A Comprehensive review. Computer Methods and Programs in Biomedicine Update 2021; 1: 100013.
  33. [97]Dhankhar S Garg N Chauhan S Role of artificial intelligence in diabetic wound screening and early detection. Current Biotechnology 2024; 13: 93- 106.
  34. [104]Singh AP Saxena R Saxena S Artificial intelligence revolution in healthcare: Transforming diagnosis, treatment, and patient care. Asian Journal of Advances in Research 2024; 7: 241- 63.
  35. [105]Zeb S Nizamullah FNU Abbasi N AI in healthcare: revolutionizing diagnosis and therapy. International Journal of Multidisciplinary Sciences and Arts 2024; 3: 118- 28.
  36. [110]Băjenaru L Tomescu M Grigorovici-Togănel I. Leveraging generative Artificial Intelligence for advanced healthcare solutions. Romanian Journal of Information Technology and Automatic Control 2024; 34: 149- 64.
  37. [115]Narayanan RR Durga N Nagalakshmi S. Impact of artificial intelligence (AI) on drug discovery and product development. Indian Journal of Pharmaceutical Education and Research 2022; 56: S387- 97.
  38. [117]Dey AK. ChatGPT in diabetes care: An overview of the evolution and potential of generative artificial intelligence model like ChatGPT in augmenting clinical and patient outcomes in the management of diabetes. International Journal of Diabetes and Technology 2023; 2: 66- 72.
  39. [119]Herrick G Frasier KM Li V Enhancing patient education and engagement through digital intelligence tools in dermatology. International Journal of Research in Dermatology 2024; 10: 391- 400.

Cited by

Sorted by citations

Authors
Osonuga, A., Olawade, D. B., Gore, M., Rotifa, O. B., Egbon, E., Chinwah, V., & Boussios, S.
Affiliations
  • York St John University (GB)
  • Symbiosis International University (IN)
  • Afe Babalola University (NG)
  • University of Applied Sciences Technikum Wien (AT)
  • Canterbury Christ Church University (GB)
Corresponding author
Olawade, David B.
ORCID
Venue
Annals of Translational Medicine
Publisher
AME Publishing Company
Published
2025-10
Volume
13
Issue
5
Pages
59-59
ISSN
  • 2305-5839 (ISSN-L)
  • 2305-5847
DOI
10.21037/atm-25-62
PMC
PMC12592000
OpenAlex
W4415682413
Semantic Scholar
e58993897a3dc321fe7628fb9babd84f94804ce1
Cited by
4
References
0
Citations by year
2026: 4
FWCI
0.33
Type
journal article
Language
en
Open access
Licence
cc-by-nc-nd
Locations
  • Annals of Translational Medicine (publishedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
  • Kent Academic Repository (University of Kent) (acceptedVersion)
  • Research at York St John (York St John University) (acceptedVersion, other-oa)
  • Europe PMC (PubMed Central) (submittedVersion, cc-by-nc-nd)
Topics
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
  • Health Sciences › Health Professions › Health Information Management › Artificial Intelligence in Healthcare
  • Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
Keywords
generative AI, large language models, type 2 diabetes, diabetes complications, diabetic retinopathy, diabetic neuropathy, cardiovascular disease, early detection, risk stratification, clinical decision support systems, precision medicine, electronic health records, multimodal data integration, data privacy

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-16.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-16.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review