Triangulated causal inference with deep counterfactual learningfor individualized statin-associated type 2 diabetes risk

Triangulated causal inference with deep counterfactual learningfor individualized statin-associated type 2 diabetes risk | 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 Article Triangulated causal inference with deep counterfactual learningfor individualized statin-associated type 2 diabetes risk Hao Zhou, Jorge Passamani Zubelli, Haralampos Hatzikirou, Andreas Henschel, and 9 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8659216/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 Determining the causal architecture of drug induced metabolic side effects is a fundamental challenge in digital medicine. While statins are the cornerstone of cardiovascular prevention, their association with type 2 diabetes incidence, particularly in normoglycemic individuals, remains a critical clinical concern. We addressed this problem using a triangulated causal inference framework that integrated two stage residual inclusion Mendelian randomization with CausalT2DNet, a deep counterfactual neural network that uses Maximum Mean Discrepancy and Standardized Mean Difference penalties to improve covariate balance. Analyses of 12 year longitudinal data from the UK Biobank, including more than 100,000 normoglycemic individuals, validation in the diverse All of Us cohort, and confirmatory analysis in a newly assembled Middle Eastern health system cohort in the United Arab Emirates showed that statin initiation was associated with a marked elevation in type 2 diabetes risk. Adjusted 10 year risk ratios were 2.30 in the UK cohort and 2.54 in the US cohort, and the adjusted 5 year risk ratio was 2.01 in the UAE cohort. All associations were highly statistically significant. Causal decomposition on the log odds scale indicated that this excess risk is predominantly driven by an LDL independent pathway, with a direct effect log odds of 0.4198 and a P value of 4.88 × 10⁻⁷⁷, whereas the LDL mediated component was modest and not statistically significant, with a Sobel P value of 0.1088. CausalT2DNet estimated that statin exposure increases 12 year type 2 diabetes risk from 2.24 percent to 4.05 percent, with 1.29 percentage points of this excess risk attributable to non lipid mechanisms on the risk scale. The resulting counterprediction framework yields individualized type 2 diabetes risk projections under alternative statin treatment scenarios, providing a scalable digital health tool to support shared clinical decision making. Biological sciences/Computational biology and bioinformatics Health sciences/Diseases Health sciences/Endocrinology Health sciences/Medical research Health sciences/Risk factors Statin LDL Cardiovascular Diabetes Machine Learning Causal inference Full Text Additional Declarations No competing interests reported. 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-8659216","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":580124572,"identity":"69f6b388-956d-4f58-8af8-6cfa005bcfe7","order_by":0,"name":"Hao Zhou","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAyklEQVRIiWNgGAWjYBACAxDB2MDAwM9DshbJHpK1GJwhVos5++EDjD932CRuPnPGgOFHDYM8PyEtlj1pCQySZ9ISt53tMWDsOcZgOLOBkMMO5BgwGLYdTtx2nseAgbeBgXHDAUJazr8xYEgEatncz2PA+LeBwX4/QS03gLYcBGrZwNtjwAy0JXEDIb8Y3HiWcLCxLc14xpljBYdljkkkzyDssOSDD3+22cj29yRvfPimxsa2v4GQNUBwAIkhQYT6UTAKRsEoGAUEAQBU0UJ0BOzVpAAAAABJRU5ErkJggg==","orcid":"","institution":"Khalifa University of Science and Technology","correspondingAuthor":true,"prefix":"","firstName":"Hao","middleName":"","lastName":"Zhou","suffix":""},{"id":580124573,"identity":"83e6e99d-15ac-4186-a1ae-d0a6a2d21f09","order_by":1,"name":"Jorge Passamani Zubelli","email":"","orcid":"","institution":"Khalifa University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Jorge","middleName":"Passamani","lastName":"Zubelli","suffix":""},{"id":580124574,"identity":"7d45d277-5104-4436-9d32-af088b287bbc","order_by":2,"name":"Haralampos Hatzikirou","email":"","orcid":"","institution":"Khalifa University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Haralampos","middleName":"","lastName":"Hatzikirou","suffix":""},{"id":580124575,"identity":"c9e4b574-41b7-45e3-81c4-5b954046214c","order_by":3,"name":"Andreas Henschel","email":"","orcid":"","institution":"Khalifa University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Andreas","middleName":"","lastName":"Henschel","suffix":""},{"id":580124576,"identity":"5fe20e9b-06bb-42b8-93af-c0967558a822","order_by":4,"name":"Laurent Alain Najman","email":"","orcid":"","institution":"Khalifa University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Laurent","middleName":"Alain","lastName":"Najman","suffix":""},{"id":580124577,"identity":"439b8b14-7a6b-445b-b667-47e1d056c12a","order_by":5,"name":"Daniel E. Platt","email":"","orcid":"","institution":"IBM","correspondingAuthor":false,"prefix":"","firstName":"Daniel","middleName":"E.","lastName":"Platt","suffix":""},{"id":580124578,"identity":"ee594d28-d2a7-4bb7-8671-0bb628523730","order_by":6,"name":"Antonello Maruotti","email":"","orcid":"","institution":"Khalifa University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Antonello","middleName":"","lastName":"Maruotti","suffix":""},{"id":580124579,"identity":"ccea5961-21ac-4c6a-abc1-41b16ca0dc3e","order_by":7,"name":"Siobhan O’Sullivan","email":"","orcid":"","institution":"Khalifa University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Siobhan","middleName":"","lastName":"O’Sullivan","suffix":""},{"id":580124580,"identity":"1f9dd2c6-a382-428a-829a-884d89c56b7b","order_by":8,"name":"Lithe Basbous","email":"","orcid":"","institution":"Khalifa University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Lithe","middleName":"","lastName":"Basbous","suffix":""},{"id":580124581,"identity":"909f6159-e751-4ff1-a2d3-fa39d71da64f","order_by":9,"name":"Cynthia Al Hageh","email":"","orcid":"","institution":"Khalifa University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Cynthia","middleName":"Al","lastName":"Hageh","suffix":""},{"id":580124582,"identity":"b125433a-7abc-4890-b597-a735da16241f","order_by":10,"name":"Mariam AlHarbi","email":"","orcid":"","institution":"Abu Dhabi Public Health Center","correspondingAuthor":false,"prefix":"","firstName":"Mariam","middleName":"","lastName":"AlHarbi","suffix":""},{"id":580124583,"identity":"e28b9995-5f6e-4b67-b959-8e9e3458d268","order_by":11,"name":"Antoine Abchee","email":"","orcid":"","institution":"University of Balamand","correspondingAuthor":false,"prefix":"","firstName":"Antoine","middleName":"","lastName":"Abchee","suffix":""},{"id":580124584,"identity":"3e573f0a-7825-47e5-b40d-9cbf123d23f1","order_by":12,"name":"Pierre Zalloua","email":"","orcid":"","institution":"Khalifa University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Pierre","middleName":"","lastName":"Zalloua","suffix":""}],"badges":[],"createdAt":"2026-01-21 11:43:39","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-8659216/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8659216/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":101179627,"identity":"34f4c3f0-bfaf-4ef8-9955-231eebd83492","added_by":"auto","created_at":"2026-01-27 03:33:19","extension":"json","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14479,"visible":true,"origin":"","legend":"","description":"","filename":"925894a81b7e41bfb7db2c3c013a6a51.json","url":"https://assets-eu.researchsquare.com/files/rs-8659216/v1/81d7a420f6409ccc8339f50d.json"},{"id":101209819,"identity":"aa754f76-d13c-4e8b-a3ec-cfdb9e4b9222","added_by":"auto","created_at":"2026-01-27 10:18:33","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2519045,"visible":true,"origin":"","legend":"","description":"","filename":"statinmanuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8659216/v1_covered_947ebf0b-6af7-4907-95ff-6a2a2e06c627.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Triangulated causal inference with deep counterfactual learningfor individualized statin-associated type 2 diabetes risk","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"isPdf":true,"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":"Statin, LDL, Cardiovascular, Diabetes, Machine Learning, Causal inference","lastPublishedDoi":"10.21203/rs.3.rs-8659216/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8659216/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Determining the causal architecture of drug induced metabolic side effects is a fundamental challenge in digital medicine. While statins are the cornerstone of cardiovascular prevention, their association with type 2 diabetes incidence, particularly in normoglycemic individuals, remains a critical clinical concern. We addressed this problem using a triangulated causal inference framework that integrated two stage residual inclusion Mendelian randomization with CausalT2DNet, a deep counterfactual neural network that uses Maximum Mean Discrepancy and Standardized Mean Difference penalties to improve covariate balance. Analyses of 12 year longitudinal data from the UK Biobank, including more than 100,000 normoglycemic individuals, validation in the diverse All of Us cohort, and confirmatory analysis in a newly assembled Middle Eastern health system cohort in the United Arab Emirates showed that statin initiation was associated with a marked elevation in type 2 diabetes risk. Adjusted 10 year risk ratios were 2.30 in the UK cohort and 2.54 in the US cohort, and the adjusted 5 year risk ratio was 2.01 in the UAE cohort. All associations were highly statistically significant. Causal decomposition on the log odds scale indicated that this excess risk is predominantly driven by an LDL independent pathway, with a direct effect log odds of 0.4198 and a P value of 4.88 × 10⁻⁷⁷, whereas the LDL mediated component was modest and not statistically significant, with a Sobel P value of 0.1088. CausalT2DNet estimated that statin exposure increases 12 year type 2 diabetes risk from 2.24 percent to 4.05 percent, with 1.29 percentage points of this excess risk attributable to non lipid mechanisms on the risk scale. The resulting counterprediction framework yields individualized type 2 diabetes risk projections under alternative statin treatment scenarios, providing a scalable digital health tool to support shared clinical decision making.","manuscriptTitle":"Triangulated causal inference with deep counterfactual learningfor individualized statin-associated type 2 diabetes risk","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-27 03:33:10","doi":"10.21203/rs.3.rs-8659216/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":"84a21bfb-ddb3-4227-b22c-caa34b8a7b20","owner":[],"postedDate":"January 27th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":61719304,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":61719305,"name":"Health sciences/Diseases"},{"id":61719306,"name":"Health sciences/Endocrinology"},{"id":61719307,"name":"Health sciences/Medical research"},{"id":61719308,"name":"Health sciences/Risk factors"}],"tags":[],"updatedAt":"2026-01-27T09:57:01+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-27 03:33:10","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8659216","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8659216","identity":"rs-8659216","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}