Real‑World Safety Signals of Tirzepatide Versus GLP‑1RAs: A Machine Learning Analysis of FDA FAERS Data

Real‑World Safety Signals of Tirzepatide Versus GLP‑1RAs: A Machine Learning Analysis of FDA FAERS Data | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Short Report Real‑World Safety Signals of Tirzepatide Versus GLP‑1RAs: A Machine Learning Analysis of FDA FAERS Data Sherry Yun Wang, Zhouzhou Chu, Ang Li, Daniel Umoru, Ariel Tran, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8546087/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background: Tirzepatide, a dual glucose-dependent insulinotropic polypeptide (GIP) and glucagon-like peptide-1 (GLP-1) receptor agonist, has rapidly expanded in clinical use. However, its real-world safety profile relative to established GLP-1 receptor agonists (GLP-1RAs) remains incompletely characterized. Methods: We conducted a retrospective pharmacovigilance analysis of U.S. Food and Drug Administration Adverse Event Reporting System (FAERS) data from 2015 to 2025. Adverse events associated with tirzepatide and other GLP-1RAs were categorized into clinically meaningful groups. We applied interpretable machine-learning models, including logistic regression and random forest classifiers, to identify factors associated with serious adverse events and to compare agent-specific safety patterns. Results: Among 479,921 GLP-1RA-related adverse event reports, gastrointestinal disorders, dosing and administration issues, and injection-site reactions were most common. Tirzepatide accounted for a high volume of reports but showed comparatively lower proportions of pancreatic, thyroid, and gallbladder events than several legacy GLP-1RAs. Across both models, seriousness classification was driven primarily by adverse-event type and reporting context rather than by drug identity. Tirzepatide demonstrated mid-range feature importance and did not independently drive serious adverse-event classification. Conclusions: In this large real-world pharmacovigilance analysis, tirzepatide did not exhibit disproportionate serious safety signals compared with other GLP-1RAs. These findings highlight that absolute reporting volume in FAERS may reflect uptake and reporting behavior rather than intrinsic drug risk, underscoring the importance of interpretable analytic approaches when evaluating post-marketing safety. Health Economics & Outcomes Research GIP/GLP-1 Glucagon-like Peptide-1 Receptor Agonist (GLP-1 RA) Adverse Drug Reactions Adverse Outcomes Machine Learning Figures Figure 1 Figure 2 Background Since its approval, tirzepatide has reshaped the global pharmaceutical landscape. The dual Glucagon-Like Peptide-1 (GLP-1) and Glucose-Dependent Insulinotropic Polypeptide (GIP) receptor agonist has now surpassed long-dominant cancer immunotherapy to become the world’s best-selling drug 1 . While its clinical efficacy and commercial momentum are undeniable, the real-world safety profile of tirzepatide remains insufficiently characterized. As millions of patients use this new class of incretin-based agents, comparative post-marketing data are urgently needed to evaluate whether tirzepatide’s dual-agonist mechanism alters adverse-event (AE) patterns relative to legacy GLP-1 receptor agonists such as semaglutide or liraglutide 2 . Existing safety evidence synthesized in Diabetes, Obesity and Metabolism 3 – 6 demonstrated that GLP-1 receptor agonists are predominantly associated with gastrointestinal adverse effects, with less frequent but clinically significant events including pancreatitis and gallbladder disease. However, the current literature largely characterizes safety at the class level, providing limited insight into agent-specific real-world AE profiles. As a result, whether newly approved dual agonists such as tirzepatide exhibit distinct post-marketing safety signals relative to legacy GLP-1 receptor agonists remains an open question. To address this gap, we conducted a retrospective pharmacovigilance analysis of U.S. Food and Drug Administration Adverse Event Reporting System (FAERS) data from 2015 to 2025 to compare adverse-event patterns across GLP-1 receptor agonists. Leveraging interpretable machine-learning models, our analysis aims to identify post-marketing safety signals and predictors of serious adverse events, enabling systematic comparison of agent-specific risk profiles and providing real-world evidence to inform clinical decision-making in the rapidly evolving incretin therapeutic landscape. Methods We conducted a descriptive pharmacovigilance analysis using cleaned data from the FAERS in Python, leveraging DuckDB to enable efficient SQL-based querying and manipulation of large FAERS datasets. To ensure consistent identification of GLP-1RA exposure, we implemented a comprehensive drug normalization and mapping strategy. Drug names were standardized using case normalization and string-cleaning procedures, and a curated GLP-1RA dictionary was constructed to include active ingredients, brand names, and combination products. Normalized drug identifiers were mapped across heterogeneous FAERS drug fields, including free-text drug names and standardized identifiers such as RxNorm concept unique identifiers (RxCUIs), allowing robust capture of GLP-1RA exposure across reports. FAERS reports listing a GLP-1RA as a primary or secondary suspect were deduplicated at the case level, with outcomes and indications harmonized to a one-case–one-indication structure. Reported adverse events were further grouped into clinically meaningful categories, including gastrointestinal, psychiatric, pancreatic, musculoskeletal, injection-site, metabolic, renal, and effectiveness-related events. This categorization enabled higher-level interpretation of safety patterns beyond individual MedDRA preferred terms and facilitated clinically interpretable summaries of adverse event profiles. Serious adverse events were identified using standard FAERS seriousness indicators. In addition to descriptive analyses, we implemented a supervised machine learning framework to generate interpretable hypotheses regarding factors associated with serious adverse events among GLP-1RA–associated FAERS reports. Two complementary classifiers were evaluated: (1) logistic regression, selected for transparency and interpretability, and (2) random forest, selected to capture potential nonlinear relationships and higher-order interactions among predictors. The primary prediction target was case-level seriousness at the time of report submission. To avoid post-outcome information leakage, only predictors available at the time of submission were included, including age, sex, reporter type, receipt month and year, primary indication category (e.g., type 2 diabetes, obesity), molecule identity, and binary adverse event category flags derived from MedDRA groupings. The analytic dataset was constructed at the case or case–molecule level, enabling molecule-specific signal assessment where applicable. Standard preprocessing, including imputation and one-hot encoding, was applied using scikit-learn pipelines. Model interpretability was assessed using Shapley Additive Explanations. This study was deemed exempt from institutional review board oversight by Chapman University because it used de-identified, publicly available data. Results Across the 479,921 adverse event reports about GLP-1 receptor agonists, dosing and administration–related events (n = 70,161; 37.8%), gastrointestinal disorders (n = 63,525; 34.3%), and injection-site reactions (n = 51,732; 27.9%) accounted for the largest absolute numbers of FAERS reports agent (Fig. 1 ). Tirzepatide contributed the second largest number of reports overall, with particularly high counts of dosing/administration issues (n = 33,147), injection-site reactions (n = 20,786), and gastrointestinal disorders (n = 16,605). Despite its high overall reporting volume, tirzepatide exhibited comparatively lower counts of pancreatic (n = 1,065), thyroid (n = 314), and gallbladder events (n = 557) relative to several legacy GLP-1 receptor agonists. In contrast, liraglutide and semaglutide (injectable) showed substantially higher counts of pancreatic effects and thyroid/gallbladder-related. Dulaglutide demonstrated high reporting across multiple categories, including metabolic effects (n = 17,908) and renal/urinary events (n = 2,968), consistent with its long-standing market presence. Exenatide exhibited the highest number of injection-site reactions (n = 10,368) among all agents, while psychiatric adverse events were most frequently reported for dulaglutide (n = 6,618), semaglutide injectable (n = 5,656), and tirzepatide (n = 4,972). Overall, although tirzepatide accounted for a large absolute number of FAERS reports across most adverse-event categories, given its short market window, pancreatic, thyroid, and gallbladder events constituted a smaller share of its total reports compared with several established GLP-1 RAs (Fig. 1 ). Two complementary models were examined: a logistic regression model, which captures linear and additive effects, and a random forest model, which captures nonlinear relationships and interactions. In the logistic regression model (Fig. 2 a, panel A), reporter characteristics dominated seriousness classification. Reports submitted by consumers showed the largest average contribution to model predictions, followed by sex (female and male) and reports from healthcare professionals. Indication-related variables (e.g., type 2 diabetes vs other indications) and selected adverse-event categories (i.e., injection-site reactions, pancreatic effects, and dosing/administration issues) had a moderate influence. Drug identity variables were comparatively less influential. Tirzepatide ranked in the middle of the top 20 predictors, indicating that while molecule identity contributed to seriousness prediction, it did not drive the model. Other GLP-1RA molecules (e.g., dulaglutide, albiglutide) ranked similarly or lower. The beeswarm plot (Fig. 2 a Panel B), shows that consumer-reported cases and female sex are predominantly associated with negative SHAP values, indicating a lower predicted probability of being classified as serious. In contrast, reports from healthcare professionals tend to shift predictions toward seriousness. The presence of pancreatic effects shifts predictions toward the serious class, while its absence shifts predictions toward non-serious classification. In contrast to the logistic regression model, the random forest model (Fig. 2 b Panel A) identified adverse-event categories as the most influential predictors of seriousness. Injection-site reactions and dosing/administration issues ranked highest, followed by molecule identity (semaglutide and tirzepatide), pancreatic effects, and temporal variables (receipt year and month). Demographic and reporter variables (e.g., consumer vs healthcare professional) played a secondary role. The random forest beeswarm plot (Fig. 2 b Panel B) demonstrates more asymmetric and spread-out SHAP distributions, consistent with nonlinear decision boundaries. The presence of injection-site reactions and dosing/administration issues strongly pushes predictions towards non-seriousness while the presence of pancreatic effects does the opposite. Temporal variables (receipt year and month) show bidirectional effects, suggesting evolving reporting patterns over time. Molecule-specific features, including tirzepatide, show modest but discernible contributions, indicating that drug identity may matter in specific clinical contexts, though it remains secondary to AE category signals. Discussion Although tirzepatide accounted for a large absolute number of FAERS reports across most adverse-event categories, pancreatic, thyroid, and gallbladder events constituted a smaller proportion of its total reports compared with several established GLP-1 receptor agonists. Our findings illustrate that absolute reporting frequency is a poor proxy for clinical severity, particularly for newly approved, high-profile therapies. While tirzepatide’s overall report volume remained high despite its more recent market entry, this pattern likely reflects rapid uptake, stimulated reporting, and heightened media attention rather than increased intrinsic risk. Preclinical studies have suggested that dual GLP-1/GIP receptor agonism may be associated with fewer adverse effects than equipotent doses of semaglutide in animal models 7 . However, FAERS data require cautious interpretation, as differences in patient populations, indications, and reporting behaviors can substantially influence observed patterns. A key innovation of this study lies in the application of interpretable ML models to disentangle reporting context from clinical signal. The logistic regression model revealed that the seriousness classification in FAERS was driven predominantly by who reports the event and how it is reported, with reporter type, sex, and indication exerting a stronger influence than the specific GLP-1RA identity, while the random forest model, capable of capturing nonlinear effects, emphasized clinical event characteristics, particularly injection-site reactions, dosing/administration issues, and pancreatic effects, as primary drivers of seriousness. The convergence of these two modeling approaches demonstrates that molecule identity, including tirzepatide, plays a secondary role once reporting behavior and AE type are taken into account. Across both models, tirzepatide exhibited mid-range feature importance and balanced SHAP distributions, indicating no disproportionate contribution to serious adverse-event classification relative to other GLP-1 receptor agonists. Moreover, tirzepatide demonstrated relatively lower proportions of pancreatic, thyroid/gallbladder, psychiatric, and renal/urinary events compared with semaglutide, dulaglutide, and liraglutide, suggesting a potentially favorable safety signal in pancreatic and neuroendocrine domains despite higher overall AE counts. The elevated number of injection-site reactions associated with tirzepatide likely reflects pharmacologic, formulation, and user-related factors rather than systemic toxicity. Tirzepatide is administered at higher doses and injection volumes than other GLP-1RAs and uses a single-use auto-injector distinct from pen-based systems, which may increase early user error, perceived trauma, or local irritation. These real-world observations are consistent with the findings of the SURPASS trialfindings 8 , in which injection-site reactions were more frequent with tirzepatide than with semaglutide, but were generally mild, transient, and rarely led to discontinuation. High adverse-event reporting does not necessarily equate to high clinical severity: our analysis reveals that seriousness in FAERS is primarily driven by event type and reporting behavior, rather than the identity of the GLP-1RA. Definitive conclusions regarding comparative safety, however, will require controlled, head-to-head studies that incorporate exposure denominators, dosage, and indication-specific risk. Declarations Author Contributions Sherry Yun Wang conceived the study concept and drafted the manuscript. Zhouzhou Chu performed the analyses and ran all code. Ang Li provided high-level methodological guidance. Tannaz Moin reviewed and critically edited the manuscript. All authors reviewed and approved the final version of the manuscript. Conflicts of Interest None Guarantor and Funding statements Drs. Sherry Yun Wang served as the guarantor and had full access to all study data, taking responsibility for the integrity of the data and the accuracy of the analyses. No external funding was received for this work. References Chen E. Eli Lilly’s weight loss and diabetes drug tops Keytruda as world’s best-selling medicine. 2025; https://www.statnews.com/2025/10/30/eli-lilly-zepbound-mounjaro-bestselling-obesity-drug/. Xie X, Yang S, Deng S, Liu Y, Xu Z, He B. Comparative Gastrointestinal Adverse Effects of GLP-1 Receptor Agonists and Multi-Target Analogs in Type 2 Diabetes: A Bayesian Network Meta-Analysis in Type 2 Diabetes: A Bayesian Network Meta-Analysis. Frontiers in Pharmacology. 2025;16:1613610. Patel H, Khunti K, Rodbard HW, et al. Gastrointestinal adverse events and weight reduction in people with type 2 diabetes treated with tirzepatide in the SURPASS clinical trials. Diabetes, Obesity and Metabolism. 2024;26(2):473-481. Yeo D, Jo Y, Jeong J, et al. Efficacy and safety of glucagon‐like peptide 1 receptor agonists across all health outcomes in type 2 diabetes: An umbrella review and evidence map of randomised controlled trials. Diabetes, Obesity and Metabolism. 2025. Kim TH, Lee K, Park S, et al. Adverse drug reaction patterns of GLP‐1 receptor agonists approved for obesity treatment: Disproportionality analysis from global pharmacovigilance database. Diabetes, Obesity and Metabolism. 2025;27(6):3490-3502. Thomsen RW, Mailhac A, Løhde JB, Pottegård A. Real‐world evidence on the utilization, clinical and comparative effectiveness, and adverse effects of newer GLP‐1RA‐based weight‐loss therapies. Diabetes, Obesity and Metabolism. 2025;27:66-88. Borner T, Pataro AM, Doebley SA, et al. Hypophagia and body weight loss by tirzepatide are accompanied by fewer GI adverse events compared to semaglutide in preclinical models. Science Advances. 2025;11(25):eadu1589. Frías JP, Davies MJ, Rosenstock J, et al. Tirzepatide versus semaglutide once weekly in patients with type 2 diabetes. New England Journal of Medicine. 2021;385(6):503-515. Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8546087","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Short Report","associatedPublications":[],"authors":[{"id":571163292,"identity":"c5d5ef96-3c99-4652-bf21-3f79940edd15","order_by":0,"name":"Sherry Yun Wang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA10lEQVRIiWNgGAWjYBADOQYJODuBOC3GpGtJbCBai8GN3GMSP3fUpvfPbn726EbNHQZ+9hwDAlry0iR7zxzPnXHnmLlxzrFnDJI9bwhpyTG7wdt2LHeDRIKZdA7bYZAIYS03/7YdSzeQSP8mnfPvMIM9MVpu87bVJBhI5JhJ57YBbZEgoEXyzBvz37JtBwxn3MgpN87tO8wjceZZAV4tfMdzjA3fttXJ889I3/Y459thOf725A14tSgcAFOHQQQbiODBqxwE5BvAVB1cyygYBaNgFIwCDAAAOvBMGhoGzgAAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0001-9438-5586","institution":"Chapman University","correspondingAuthor":true,"prefix":"","firstName":"Sherry","middleName":"Yun","lastName":"Wang","suffix":""},{"id":571163756,"identity":"63cd66bc-7eff-4435-8874-e223bfb3741d","order_by":1,"name":"Zhouzhou Chu","email":"","orcid":"","institution":"Chapman University","correspondingAuthor":false,"prefix":"","firstName":"Zhouzhou","middleName":"","lastName":"Chu","suffix":""},{"id":571163757,"identity":"fe591fa8-8a15-4cc1-bc18-c2efc54c8b05","order_by":2,"name":"Ang Li","email":"","orcid":"","institution":"University of Maryland, College Park","correspondingAuthor":false,"prefix":"","firstName":"Ang","middleName":"","lastName":"Li","suffix":""},{"id":571163758,"identity":"83987a0e-1424-46be-b00d-d6e103ee6764","order_by":3,"name":"Daniel Umoru","email":"","orcid":"","institution":"Chapman University","correspondingAuthor":false,"prefix":"","firstName":"Daniel","middleName":"","lastName":"Umoru","suffix":""},{"id":571163759,"identity":"82cc1f43-9cdf-4bc1-9874-15b5dece2ee2","order_by":4,"name":"Ariel Tran","email":"","orcid":"","institution":"Chapman University","correspondingAuthor":false,"prefix":"","firstName":"Ariel","middleName":"","lastName":"Tran","suffix":""},{"id":571163760,"identity":"b6f3166a-c6ad-45e4-8bb0-3919238c56de","order_by":5,"name":"Bowei Tian","email":"","orcid":"","institution":"University of Maryland, College Park","correspondingAuthor":false,"prefix":"","firstName":"Bowei","middleName":"","lastName":"Tian","suffix":""},{"id":571163761,"identity":"59acf5b2-1e23-492c-8acb-53af518a5fb7","order_by":6,"name":"Peihao Dou","email":"","orcid":"","institution":"Chapman University","correspondingAuthor":false,"prefix":"","firstName":"Peihao","middleName":"","lastName":"Dou","suffix":""},{"id":571163762,"identity":"658b13c8-f3ec-4982-a17a-cab0cde395a6","order_by":7,"name":"Tannaz Moin","email":"","orcid":"","institution":"University of California Los Angeles","correspondingAuthor":false,"prefix":"","firstName":"Tannaz","middleName":"","lastName":"Moin","suffix":""}],"badges":[],"createdAt":"2026-01-08 02:29:38","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-8546087/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8546087/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":99848755,"identity":"eb781052-b645-41f6-b9c7-89d875dcf6f6","added_by":"auto","created_at":"2026-01-09 03:01:38","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":812644,"visible":true,"origin":"","legend":"","description":"","filename":"DiabetesObesityandMetabolism.docx","url":"https://assets-eu.researchsquare.com/files/rs-8546087/v1/0cefceda18d90e2de56ce114.docx"},{"id":100357345,"identity":"42dbea40-7016-4b73-8ec5-bf89d4b0ab24","added_by":"auto","created_at":"2026-01-16 07:19:43","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":342,"visible":true,"origin":"","legend":"","description":"","filename":"rs8546087.json","url":"https://assets-eu.researchsquare.com/files/rs-8546087/v1/b085a71cdd5efb716d13d25b.json"},{"id":100357300,"identity":"fb1ae6f6-fcb6-465f-82d3-40c3d7920dcd","added_by":"auto","created_at":"2026-01-16 07:19:37","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":31546,"visible":true,"origin":"","legend":"","description":"","filename":"rs85460870enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8546087/v1/3bdaeb97b7c8c7dfea85dcff.xml"},{"id":99848746,"identity":"884ff135-f409-4015-b535-aa4ae24d1c37","added_by":"auto","created_at":"2026-01-09 03:01:38","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":273497,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8546087/v1/203407595eaebade98414022.png"},{"id":99848749,"identity":"5673ac14-cdaa-4fb2-8807-65b842a86a70","added_by":"auto","created_at":"2026-01-09 03:01:38","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":335745,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8546087/v1/d9880133e5efd5c8c23303b4.png"},{"id":100357362,"identity":"c65c3ca0-1bfe-4532-83e3-f017b89d4b13","added_by":"auto","created_at":"2026-01-16 07:19:45","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":99872,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8546087/v1/d4a59ec8b54560454f4f95f2.png"},{"id":99848752,"identity":"1fa9360c-aa6a-424d-9c49-f4133c706118","added_by":"auto","created_at":"2026-01-09 03:01:38","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":67769,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8546087/v1/51635ed009627871faf4f696.png"},{"id":100356860,"identity":"e7001a04-cc9c-4737-843d-6a688cee54b0","added_by":"auto","created_at":"2026-01-16 07:17:49","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":65368,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8546087/v1/5c4b6fac63e51c9a8e5791f5.png"},{"id":99848756,"identity":"0b4b35d5-e508-47de-a2d5-971e04d540b3","added_by":"auto","created_at":"2026-01-09 03:01:38","extension":"xml","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":30686,"visible":true,"origin":"","legend":"","description":"","filename":"rs85460870structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8546087/v1/eb179f693cb647d78150e66f.xml"},{"id":100357656,"identity":"cbd67fb1-a101-4851-8381-16f38ff71222","added_by":"auto","created_at":"2026-01-16 07:20:09","extension":"html","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":34509,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8546087/v1/0e13133bec0d5a99d213581b.html"},{"id":100357372,"identity":"017876d9-f70d-4d08-8acf-e14623a3920a","added_by":"auto","created_at":"2026-01-16 07:19:46","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":514318,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDistribution of adverse event categories across GLP-1 receptor agonists\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8546087/v1/0376accf42b7d85c701b6dde.jpeg"},{"id":100357710,"identity":"5de0df96-31f1-4853-9736-826672ccbe14","added_by":"auto","created_at":"2026-01-16 07:20:14","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":516428,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSHAP-Based Interpretation of Logistic Regression and Random Forest Models Predicting Serious Adverse Events in FAERS GLP-1RA Reports\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8546087/v1/e2ca615af7956344f08cb4bb.png"},{"id":100377064,"identity":"3da09db8-68a4-40d7-b653-9807eb59106d","added_by":"auto","created_at":"2026-01-16 08:46:52","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1341694,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8546087/v1/bf892f2e-54f5-4458-94a6-b3ae6940dd90.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eReal‑World Safety Signals of Tirzepatide Versus GLP‑1RAs: A Machine Learning Analysis of FDA FAERS Data\u003c/p\u003e","fulltext":[{"header":"Background","content":"\u003cp\u003eSince its approval, tirzepatide has reshaped the global pharmaceutical landscape. The dual Glucagon-Like Peptide-1 (GLP-1) and Glucose-Dependent Insulinotropic Polypeptide (GIP) receptor agonist has now surpassed long-dominant cancer immunotherapy to become the world\u0026rsquo;s best-selling drug\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. While its clinical efficacy and commercial momentum are undeniable, the real-world safety profile of tirzepatide remains insufficiently characterized. As millions of patients use this new class of incretin-based agents, comparative post-marketing data are urgently needed to evaluate whether tirzepatide\u0026rsquo;s dual-agonist mechanism alters adverse-event (AE) patterns relative to legacy GLP-1 receptor agonists such as semaglutide or liraglutide\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eExisting safety evidence synthesized in Diabetes, Obesity and Metabolism\u003csup\u003e\u003cspan additionalcitationids=\"CR4 CR5\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e demonstrated that GLP-1 receptor agonists are predominantly associated with gastrointestinal adverse effects, with less frequent but clinically significant events including pancreatitis and gallbladder disease. However, the current literature largely characterizes safety at the class level, providing limited insight into agent-specific real-world AE profiles. As a result, whether newly approved dual agonists such as tirzepatide exhibit distinct post-marketing safety signals relative to legacy GLP-1 receptor agonists remains an open question. To address this gap, we conducted a retrospective pharmacovigilance analysis of U.S. Food and Drug Administration Adverse Event Reporting System (FAERS) data from 2015 to 2025 to compare adverse-event patterns across GLP-1 receptor agonists. Leveraging interpretable machine-learning models, our analysis aims to identify post-marketing safety signals and predictors of serious adverse events, enabling systematic comparison of agent-specific risk profiles and providing real-world evidence to inform clinical decision-making in the rapidly evolving incretin therapeutic landscape.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003eWe conducted a descriptive pharmacovigilance analysis using cleaned data from the FAERS in Python, leveraging DuckDB to enable efficient SQL-based querying and manipulation of large FAERS datasets. To ensure consistent identification of GLP-1RA exposure, we implemented a comprehensive drug normalization and mapping strategy. Drug names were standardized using case normalization and string-cleaning procedures, and a curated GLP-1RA dictionary was constructed to include active ingredients, brand names, and combination products. Normalized drug identifiers were mapped across heterogeneous FAERS drug fields, including free-text drug names and standardized identifiers such as RxNorm concept unique identifiers (RxCUIs), allowing robust capture of GLP-1RA exposure across reports. FAERS reports listing a GLP-1RA as a primary or secondary suspect were deduplicated at the case level, with outcomes and indications harmonized to a one-case\u0026ndash;one-indication structure. Reported adverse events were further grouped into clinically meaningful categories, including gastrointestinal, psychiatric, pancreatic, musculoskeletal, injection-site, metabolic, renal, and effectiveness-related events. This categorization enabled higher-level interpretation of safety patterns beyond individual MedDRA preferred terms and facilitated clinically interpretable summaries of adverse event profiles. Serious adverse events were identified using standard FAERS seriousness indicators. In addition to descriptive analyses, we implemented a supervised machine learning framework to generate interpretable hypotheses regarding factors associated with serious adverse events among GLP-1RA\u0026ndash;associated FAERS reports. Two complementary classifiers were evaluated: (1) logistic regression, selected for transparency and interpretability, and (2) random forest, selected to capture potential nonlinear relationships and higher-order interactions among predictors. The primary prediction target was case-level seriousness at the time of report submission. To avoid post-outcome information leakage, only predictors available at the time of submission were included, including age, sex, reporter type, receipt month and year, primary indication category (e.g., type 2 diabetes, obesity), molecule identity, and binary adverse event category flags derived from MedDRA groupings. The analytic dataset was constructed at the case or case\u0026ndash;molecule level, enabling molecule-specific signal assessment where applicable. Standard preprocessing, including imputation and one-hot encoding, was applied using scikit-learn pipelines. Model interpretability was assessed using Shapley Additive Explanations. This study was deemed exempt from institutional review board oversight by Chapman University because it used de-identified, publicly available data.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eAcross the 479,921 adverse event reports about GLP-1 receptor agonists, dosing and administration\u0026ndash;related events (n\u0026thinsp;=\u0026thinsp;70,161; 37.8%), gastrointestinal disorders (n\u0026thinsp;=\u0026thinsp;63,525; 34.3%), and injection-site reactions (n\u0026thinsp;=\u0026thinsp;51,732; 27.9%) accounted for the largest absolute numbers of FAERS reports agent (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Tirzepatide contributed the second largest number of reports overall, with particularly high counts of dosing/administration issues (n\u0026thinsp;=\u0026thinsp;33,147), injection-site reactions (n\u0026thinsp;=\u0026thinsp;20,786), and gastrointestinal disorders (n\u0026thinsp;=\u0026thinsp;16,605). Despite its high overall reporting volume, tirzepatide exhibited comparatively lower counts of pancreatic (n\u0026thinsp;=\u0026thinsp;1,065), thyroid (n\u0026thinsp;=\u0026thinsp;314), and gallbladder events (n\u0026thinsp;=\u0026thinsp;557) relative to several legacy GLP-1 receptor agonists. In contrast, liraglutide and semaglutide (injectable) showed substantially higher counts of pancreatic effects and thyroid/gallbladder-related. Dulaglutide demonstrated high reporting across multiple categories, including metabolic effects (n\u0026thinsp;=\u0026thinsp;17,908) and renal/urinary events (n\u0026thinsp;=\u0026thinsp;2,968), consistent with its long-standing market presence. Exenatide exhibited the highest number of injection-site reactions (n\u0026thinsp;=\u0026thinsp;10,368) among all agents, while psychiatric adverse events were most frequently reported for dulaglutide (n\u0026thinsp;=\u0026thinsp;6,618), semaglutide injectable (n\u0026thinsp;=\u0026thinsp;5,656), and tirzepatide (n\u0026thinsp;=\u0026thinsp;4,972). Overall, although tirzepatide accounted for a large absolute number of FAERS reports across most adverse-event categories, given its short market window, pancreatic, thyroid, and gallbladder events constituted a smaller share of its total reports compared with several established GLP-1 RAs (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTwo complementary models were examined: a logistic regression model, which captures linear and additive effects, and a random forest model, which captures nonlinear relationships and interactions. In the logistic regression model (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea, panel A), reporter characteristics dominated seriousness classification. Reports submitted by consumers showed the largest average contribution to model predictions, followed by sex (female and male) and reports from healthcare professionals. Indication-related variables (e.g., type 2 diabetes vs other indications) and selected adverse-event categories (i.e., injection-site reactions, pancreatic effects, and dosing/administration issues) had a moderate influence. Drug identity variables were comparatively less influential. Tirzepatide ranked in the middle of the top 20 predictors, indicating that while molecule identity contributed to seriousness prediction, it did not drive the model. Other GLP-1RA molecules (e.g., dulaglutide, albiglutide) ranked similarly or lower. The beeswarm plot (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea Panel B), shows that consumer-reported cases and female sex are predominantly associated with negative SHAP values, indicating a lower predicted probability of being classified as serious. In contrast, reports from healthcare professionals tend to shift predictions toward seriousness. The presence of pancreatic effects shifts predictions toward the serious class, while its absence shifts predictions toward non-serious classification.\u003c/p\u003e \u003cp\u003eIn contrast to the logistic regression model, the random forest model (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb Panel A) identified adverse-event categories as the most influential predictors of seriousness. Injection-site reactions and dosing/administration issues ranked highest, followed by molecule identity (semaglutide and tirzepatide), pancreatic effects, and temporal variables (receipt year and month). Demographic and reporter variables (e.g., consumer vs healthcare professional) played a secondary role. The random forest beeswarm plot (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb Panel B) demonstrates more asymmetric and spread-out SHAP distributions, consistent with nonlinear decision boundaries. The presence of injection-site reactions and dosing/administration issues strongly pushes predictions towards non-seriousness while the presence of pancreatic effects does the opposite. Temporal variables (receipt year and month) show bidirectional effects, suggesting evolving reporting patterns over time. Molecule-specific features, including tirzepatide, show modest but discernible contributions, indicating that drug identity may matter in specific clinical contexts, though it remains secondary to AE category signals.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eAlthough tirzepatide accounted for a large absolute number of FAERS reports across most adverse-event categories, pancreatic, thyroid, and gallbladder events constituted a smaller proportion of its total reports compared with several established GLP-1 receptor agonists. Our findings illustrate that absolute reporting frequency is a poor proxy for clinical severity, particularly for newly approved, high-profile therapies. While tirzepatide\u0026rsquo;s overall report volume remained high despite its more recent market entry, this pattern likely reflects rapid uptake, stimulated reporting, and heightened media attention rather than increased intrinsic risk. Preclinical studies have suggested that dual GLP-1/GIP receptor agonism may be associated with fewer adverse effects than equipotent doses of semaglutide in animal models\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. However, FAERS data require cautious interpretation, as differences in patient populations, indications, and reporting behaviors can substantially influence observed patterns.\u003c/p\u003e \u003cp\u003eA key innovation of this study lies in the application of interpretable ML models to disentangle reporting context from clinical signal. The logistic regression model revealed that the seriousness classification in FAERS was driven predominantly by who reports the event and how it is reported, with reporter type, sex, and indication exerting a stronger influence than the specific GLP-1RA identity, while the random forest model, capable of capturing nonlinear effects, emphasized clinical event characteristics, particularly injection-site reactions, dosing/administration issues, and pancreatic effects, as primary drivers of seriousness. The convergence of these two modeling approaches demonstrates that molecule identity, including tirzepatide, plays a secondary role once reporting behavior and AE type are taken into account. Across both models, tirzepatide exhibited mid-range feature importance and balanced SHAP distributions, indicating no disproportionate contribution to serious adverse-event classification relative to other GLP-1 receptor agonists. Moreover, tirzepatide demonstrated relatively lower proportions of pancreatic, thyroid/gallbladder, psychiatric, and renal/urinary events compared with semaglutide, dulaglutide, and liraglutide, suggesting a potentially favorable safety signal in pancreatic and neuroendocrine domains despite higher overall AE counts. The elevated number of injection-site reactions associated with tirzepatide likely reflects pharmacologic, formulation, and user-related factors rather than systemic toxicity. Tirzepatide is administered at higher doses and injection volumes than other GLP-1RAs and uses a single-use auto-injector distinct from pen-based systems, which may increase early user error, perceived trauma, or local irritation. These real-world observations are consistent with the findings of the SURPASS trialfindings\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e, in which injection-site reactions were more frequent with tirzepatide than with semaglutide, but were generally mild, transient, and rarely led to discontinuation. High adverse-event reporting does not necessarily equate to high clinical severity: our analysis reveals that seriousness in FAERS is primarily driven by event type and reporting behavior, rather than the identity of the GLP-1RA. Definitive conclusions regarding comparative safety, however, will require controlled, head-to-head studies that incorporate exposure denominators, dosage, and indication-specific risk.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e Sherry Yun Wang conceived the study concept and drafted the manuscript. Zhouzhou Chu performed the analyses and ran all code. Ang Li provided high-level methodological guidance. Tannaz Moin reviewed and critically edited the manuscript. All authors reviewed and approved the final version of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflicts of Interest\u0026nbsp;\u003c/strong\u003eNone\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGuarantor and Funding statements\u0026nbsp;\u003c/strong\u003eDrs. Sherry Yun Wang served as the guarantor and had full access to all study data, taking responsibility for the integrity of the data and the accuracy of the analyses. No external funding was received for this work.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eChen E. Eli Lilly\u0026rsquo;s weight loss and diabetes drug tops Keytruda as world\u0026rsquo;s best-selling medicine. 2025; https://www.statnews.com/2025/10/30/eli-lilly-zepbound-mounjaro-bestselling-obesity-drug/.\u003c/li\u003e\n \u003cli\u003eXie X, Yang S, Deng S, Liu Y, Xu Z, He B. Comparative Gastrointestinal Adverse Effects of GLP-1 Receptor Agonists and Multi-Target Analogs in Type 2 Diabetes: A Bayesian Network Meta-Analysis in Type 2 Diabetes: A Bayesian Network Meta-Analysis. \u003cem\u003eFrontiers in Pharmacology.\u0026nbsp;\u003c/em\u003e2025;16:1613610.\u003c/li\u003e\n \u003cli\u003ePatel H, Khunti K, Rodbard HW, et al. Gastrointestinal adverse events and weight reduction in people with type 2 diabetes treated with tirzepatide in the SURPASS clinical trials. \u003cem\u003eDiabetes, Obesity and Metabolism.\u0026nbsp;\u003c/em\u003e2024;26(2):473-481.\u003c/li\u003e\n \u003cli\u003eYeo D, Jo Y, Jeong J, et al. Efficacy and safety of glucagon‐like peptide 1 receptor agonists across all health outcomes in type 2 diabetes: An umbrella review and evidence map of randomised controlled trials. \u003cem\u003eDiabetes, Obesity and Metabolism.\u0026nbsp;\u003c/em\u003e2025.\u003c/li\u003e\n \u003cli\u003eKim TH, Lee K, Park S, et al. Adverse drug reaction patterns of GLP‐1 receptor agonists approved for obesity treatment: Disproportionality analysis from global pharmacovigilance database. \u003cem\u003eDiabetes, Obesity and Metabolism.\u0026nbsp;\u003c/em\u003e2025;27(6):3490-3502.\u003c/li\u003e\n \u003cli\u003eThomsen RW, Mailhac A, L\u0026oslash;hde JB, Potteg\u0026aring;rd A. Real‐world evidence on the utilization, clinical and comparative effectiveness, and adverse effects of newer GLP‐1RA‐based weight‐loss therapies. \u003cem\u003eDiabetes, Obesity and Metabolism.\u0026nbsp;\u003c/em\u003e2025;27:66-88.\u003c/li\u003e\n \u003cli\u003eBorner T, Pataro AM, Doebley SA, et al. Hypophagia and body weight loss by tirzepatide are accompanied by fewer GI adverse events compared to semaglutide in preclinical models. \u003cem\u003eScience Advances.\u0026nbsp;\u003c/em\u003e2025;11(25):eadu1589.\u003c/li\u003e\n \u003cli\u003eFr\u0026iacute;as JP, Davies MJ, Rosenstock J, et al. Tirzepatide versus semaglutide once weekly in patients with type 2 diabetes. \u003cem\u003eNew England Journal of Medicine.\u0026nbsp;\u003c/em\u003e2021;385(6):503-515.\u003cstrong\u003e\u003cbr\u003e\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Chapman University","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"GIP/GLP-1, Glucagon-like Peptide-1 Receptor Agonist (GLP-1 RA), Adverse Drug Reactions, Adverse Outcomes, Machine Learning","lastPublishedDoi":"10.21203/rs.3.rs-8546087/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8546087/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eBackground: Tirzepatide, a dual glucose-dependent insulinotropic polypeptide (GIP) and glucagon-like peptide-1 (GLP-1) receptor agonist, has rapidly expanded in clinical use. However, its real-world safety profile relative to established GLP-1 receptor agonists (GLP-1RAs) remains incompletely characterized.\u003c/p\u003e\n\u003cp\u003eMethods: We conducted a retrospective pharmacovigilance analysis of U.S. Food and Drug Administration Adverse Event Reporting System (FAERS) data from 2015 to 2025. Adverse events associated with tirzepatide and other GLP-1RAs were categorized into clinically meaningful groups. We applied interpretable machine-learning models, including logistic regression and random forest classifiers, to identify factors associated with serious adverse events and to compare agent-specific safety patterns.\u003c/p\u003e\n\u003cp\u003eResults: Among 479,921 GLP-1RA-related adverse event reports, gastrointestinal disorders, dosing and administration issues, and injection-site reactions were most common. Tirzepatide accounted for a high volume of reports but showed comparatively lower proportions of pancreatic, thyroid, and gallbladder events than several legacy GLP-1RAs. Across both models, seriousness classification was driven primarily by adverse-event type and reporting context rather than by drug identity. Tirzepatide demonstrated mid-range feature importance and did not independently drive serious adverse-event classification.\u003c/p\u003e\n\u003cp\u003eConclusions: In this large real-world pharmacovigilance analysis, tirzepatide did not exhibit disproportionate serious safety signals compared with other GLP-1RAs. These findings highlight that absolute reporting volume in FAERS may reflect uptake and reporting behavior rather than intrinsic drug risk, underscoring the importance of interpretable analytic approaches when evaluating post-marketing safety.\u003c/p\u003e","manuscriptTitle":"Real‑World Safety Signals of Tirzepatide Versus GLP‑1RAs: A Machine Learning Analysis of FDA FAERS Data","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-09 03:01:33","doi":"10.21203/rs.3.rs-8546087/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"c0e52615-8477-44eb-a035-95339b0347b2","owner":[],"postedDate":"January 9th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":60779601,"name":"Health Economics \u0026 Outcomes Research"}],"tags":[],"updatedAt":"2026-01-09T03:01:34+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-09 03:01:33","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8546087","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8546087","identity":"rs-8546087","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}