medRxiv : the preprint server for health sciences最新文献

{"title":"Can Large Language Models Recognize Primary Immunodeficiency from Patient-Described Symptoms?","authors":"Leon C Reteig, Steven Woloshin, Paul J Maglione, Jocelyn R Farmer, Mei-Sing Ong","doi":"10.64898/2026.05.26.26353818","DOIUrl":"10.64898/2026.05.26.26353818","url":null,"abstract":"<p><p>Patients with primary immunodeficiency (PI) face prolonged diagnostic delays and may increasingly turn to large language models (LLMs) to interpret their symptoms during this period. We evaluated whether an LLM could recognize PI from symptom descriptions derived from interviews with 21 PI patients. In prior work, we showed that GPT-4o identified PI in 96% of cases when prompted with physician-written patient histories (Rider <i>et al</i>., 2025). Here, when prompted with symptom descriptions in patients' own words, GPT-5.2 identified PI in only 7 cases (33%), although it suggested general immune system concerns in 17 cases (81%). This expert-versus-patient gap may indicate that LLMs are sensitive to the language and framing of symptom descriptions, performing substantially worse when patients describe their own symptoms in everyday language than when clinicians summarize patient histories in structured medical terms. This study underscores the need to carefully evaluate how LLMs are used in patient-facing applications.</p>","PeriodicalId":94281,"journal":{"name":"medRxiv : the preprint server for health sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13648762/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149167300","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Clinical and mechanistic responses to a music-based intervention for pain in adults with irritable bowel syndrome: Protocol for a single-arm pilot study.","authors":"Weizi Wu, Eunhea You, Aolan Li, Jie Chen, Luana Colloca, Aza Allsop, Shannon Kiley, Boluwatife Faremi, Camila Jimenez Wong, John M Toribio, Kyle J Mahoney, Hugo Fernando Posada-Quintero, Ming-Hui Chen, Debra S Burns, Xiaomei Cong","doi":"10.64898/2026.09.11.26362754","DOIUrl":"10.64898/2026.09.11.26362754","url":null,"abstract":"<p><p>Irritable bowel syndrome (IBS) is a common disorder of brain-gut interaction characterized by recurrent abdominal pain and altered bowel habits. Although music-based interventions (MBIs) have shown benefits on pain and stress, the mechanisms associated with MBI-related changes in IBS pain remain poorly understood, and the feasibility of capturing these responses using multimodal wearable monitoring in everyday settings has not been established. This protocol describes a single-arm pilot mechanistic trial examining clinical and multimodal mechanistic responses to an MBI in adults with IBS and the feasibility of extending physiological monitoring from the laboratory to home settings. Thirty-six adults aged 18-50 years with provider-confirmed IBS will be enrolled, with approximately 30 expected to complete the post-intervention laboratory visit. Participants will receive board-certified music therapist-guided laboratory sessions and complete four weeks of self-administered home practice using a standardized 20-minute MBI protocol at least five days per week. Guided by a brain-gut mechanistic framework, assessments will include patient-reported pain, stress, and IBS symptoms; quantitative sensory testing; gut microbiome profiling using 16S rRNA gene sequencing; and multimodal physiological recordings of neural, autonomic, and muscular activity. Analyses will characterize within-session and longitudinal changes in clinical and mechanistic measures and assess recruitment, retention, MBI adherence, wearable data completeness, and participant acceptability. Findings will inform the selection of candidate mechanisms and wearable monitoring approaches for evaluation in a future adequately powered randomized controlled trial.</p>","PeriodicalId":94281,"journal":{"name":"medRxiv : the preprint server for health sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13596518/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149031508","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Sequencing for a Lifetime: Professional Perspectives on the Value and Feasibility of Lifelong Genomic Medicine.","authors":"Anna C F Lewis, Ingrid A Holm, Adam H Buchanan, Aaron J Goldenberg, Bartha M Knoppers, Amy L McGuire, Robert C Green","doi":"10.64898/2026.04.29.26352046","DOIUrl":"10.64898/2026.04.29.26352046","url":null,"abstract":"<p><strong>Introduction: </strong>Lifelong genomic medicine, in which stored genomic data informs a lifetime of care, is moving from vision toward reality, with dozens of newborn sequencing projects underway worldwide and reanalysis of data already in clinical practice. Professional perspectives on its value and feasibility have not been examined.</p><p><strong>Methods: </strong>We conducted and thematically analyzed semi-structured interviews with 52 US-based professionals, including clinical geneticists, genetic counselors, primary care clinicians, laboratory personnel, and genomic screening implementers. Interviews explored delivery setting, longitudinal clinical relationships, data reanalysis and recontact, consent, data stewardship, privacy, and the pediatric-to-adult transition. Analysis combined deductive domains with inductive coding, with attention to similarities and differences across stakeholder groups.</p><p><strong>Results: </strong>Participants in all five groups saw clinical value in lifelong genomic medicine, citing the relevance of genomic information at multiple life stages, the ability to requery the genome as circumstances change, and the capacity to act on evolving knowledge. A minority contested this value, favoring more targeted testing. Many participants viewed the model as inevitable and described operational precedent for many component activities, from reanalysis of decade-old data and reclassification pipelines to \"sequence once, query often\" workflows and EHR-integrated alerts. Participants described substantial gaps in governance, spanning delivery setting, consent, data stewardship, recontact, and the pediatric-to-adult transition, alongside continuing operational challenges such as interoperability. No delivery setting was clearly favored: participants saw newborn screening as offering the strongest equity case but as conflicting with the explicit consent it would require, health-system delivery as excluding those without access, and commercial models as potentially short-lived. Participants raised equity concerns at enrollment, in maintaining follow-up, and at the pediatric-to-adult handoff, which we argue are cumulative and mutually reinforcing for those least connected to health systems.</p><p><strong>Conclusion: </strong>Many participants expected lifelong genomic medicine to be implemented, but no comprehensive or clearly favored model for delivering it responsibly emerged. Coordinated investment in governance is needed, alongside evidence on which forms of lifelong genomic medicine offer enough clinical value to justify their costs.</p>","PeriodicalId":94281,"journal":{"name":"medRxiv : the preprint server for health sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13648683/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149167419","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"<i>TMEM106B</i> analysis in diverse cohorts suggests no association with Parkinson's disease risk.","authors":"Henry Mauricio Chaparro-Solano, Sajanth Kanagasingam, Ziv Gan-Or, Konstantin Senkevich","doi":"10.64898/2026.09.28.26363833","DOIUrl":"https://doi.org/10.64898/2026.09.28.26363833","url":null,"abstract":"<p><p>Recent evidence suggests that TMEM106B expression is reduced in the brains of patients with Parkinson's disease (PD). However, an association between rare genetic variants in TMEM106B and PD risk has not been previously established. In this work, we investigated the association between rare TMEM106B variants and PD risk. We leveraged whole-genome sequencing data from 22,010 cases and 40,550 controls across diverse ancestries from the Global Parkinson's Genetics Program (GP2), the Accelerating Medicines Partnership in Parkinson's Disease (AMP-PD), and the UK Biobank (UKBB). A total of 4,581 unique rare variants were identified across all groups. We found no evidence that rare coding variants in TMEM106B were associated with PD risk, either individually or through cumulative burden. These findings suggest that rare variation in TMEM106B does not play a major role in PD susceptibility across the populations studied.</p>","PeriodicalId":94281,"journal":{"name":"medRxiv : the preprint server for health sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13648797/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149171725","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Real-life Psychological Experiences and Subsequent Cardiovascular Events among Survivors of a Myocardial Infarction in Midlife: Prospective Cohort Study.","authors":"Viola Vaccarino, Hua She, Lisa Elon, Tené T Lewis, Yi-An Ko, Nancy Murrah, Lucy H Shallenberger, Tatum Roberts, Lewam Stefanos, Jose Medina Inojosa, Kristen Harris, Taha Ahmed, Alexander Razavi, J Douglas Bremner, Paolo Raggi, Arshed Quyyumi, Amit J Shah","doi":"10.64898/2026.09.30.26364457","DOIUrl":"https://doi.org/10.64898/2026.09.30.26364457","url":null,"abstract":"<p><strong>Background: </strong>Information is lacking about everyday psychological states of individuals with a recent myocardial infarction (MI) and whether they influence outcomes. We investigated real world psychological experiences and their association with subsequent cardiovascular events in survivors of an MI.</p><p><strong>Methods: </strong>We studied 287 individuals ≤ 60 years of age within 8 months from an MI. At the baseline visit, we assessed sociodemographic and clinical factors, and administered a battery of psychological scales of depression, anxiety, anger, perceived stress and social support. Participants underwent a 7-day home monitoring of mood and stress using ecological momentary assessments, and a momentary negative affect (MNA) score, integrating daily mean ratings of sadness, anxiety, anger, distress and loneliness, was calculated. Participants were followed for a median time of 28 months. The study endpoint was a composite index of major adverse cardiovascular events (MACE), including cardiovascular death, nonfatal MI, stroke or hospitalizations for heart failure. Cox proportional hazards models adjusted for sociodemographic and clinical factors.</p><p><strong>Results: </strong>The mean age of the sample was 51 years, and 43% of participants were women. Each standard deviation increase in MNA was significantly associated with a 60% increased risk of MACE (95% confidence interval (CI), 20% to 100%), irrespective of adjustment for sociodemographic factors and clinical characteristics. The association was similar in women and men. Momentary positive mood and momentary perceived stress were not associated with MACE. The MNA score showed a larger independent contribution to MACE compared to the combined battery of baseline psychological assessments and the association was independent of baseline history of depression.</p><p><strong>Conclusion: </strong>Among individuals after MI, moment-to-moment negative affect during one week of monitoring in daily life is a powerful predictor of subsequent cardiovascular events. A prospective evaluation of negative mood during everyday living could represent a novel strategy to uncover residual risk after MI.</p>","PeriodicalId":94281,"journal":{"name":"medRxiv : the preprint server for health sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13648790/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149170694","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Prefrontal EEG Delta-Beta Coupling Marks Depressed State in Adult Siblings.","authors":"Mina Elhamiasl, Lilian Y Li, Felix A Maldonado Osorio, Brent I Rappaport, Stewart A Shankman, Ivan Alekseichuk","doi":"10.64898/2026.10.01.26364438","DOIUrl":"https://doi.org/10.64898/2026.10.01.26364438","url":null,"abstract":"<p><p>Major depressive disorder (MDD) is marked by disorganization in electrophysiological brain networks, yet it remains unclear whether such disorganization reflects the state-dependent pathological effects or a trait-like vulnerability. We analyzed resting-state electroencephalographic (EEG) delta-beta phase-amplitude coupling (PAC) in two sibling cohorts. Current-MDD families included individuals with current MDD and their currently unaffected siblings (n=36), and remitted-MDD families comprised individuals with remitted MDD and their lifetime unaffected siblings (n=110). Delta-beta _high PAC was significantly elevated (p _corr ≤0.05) over the left-medial frontal region in participants with current MDD relative to their unaffected siblings. In the remitted-MDD families, no significant differences were observed. Regrouping all individuals irrespective of family ties into current MDD, remitted MDD, and never-depressed controls confirmed significantly elevated frontal delta-beta _high PAC at rest in the current-MDD subgroup compared to other subgroups. Overall, delta-beta _high coupling reflects the depressive state rather than a trait-like or familial vulnerability.</p>","PeriodicalId":94281,"journal":{"name":"medRxiv : the preprint server for health sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13648825/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149170562","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Comparative immunogenicity and relative vaccine efficacy of recombinant and cell culture-based influenza vaccines: Results from a randomized trial of adults in the United States, 2024-2025.","authors":"Kelsey M Sumner, Emma K Noble, Lauren B Grant, Jessica R Meeker, Vel Murugan, Lora Nordstrom, Stacey L House, Rachel M Presti, Emmanuel B Walter, Olivia L Williams, Elie A Saade, David H Canaday, Katherine V Williams, Richard K Zimmerman, F Liaini Gross, Sara Valencia, Brian Gurbaxani, Nathaniel M Lewis, Zhu-Nan Li, Vasiliy Mishin, Larisa Gubareva, Bin Zhou, Min Z Levine, Brendan Flannery, Sascha Ellington","doi":"10.64898/2026.10.01.26364413","DOIUrl":"10.64898/2026.10.01.26364413","url":null,"abstract":"<p><strong>Background: </strong>Recombinant (RIV) and cell culture-based inactivated (ccIIV) influenza vaccines could improve seasonal influenza prevention; however, evidence is limited in how protection may differ among non-egg-based vaccines.</p><p><strong>Methods: </strong>We conducted a randomized trial of RIV versus ccIIV to compare vaccine immunogenicity in adults aged 18 to 64 years during the 2024-25 influenza season (ClinicalTrials.gov, NCT06518577). Active surveillance for influenza-like illness was conducted for six months post-vaccination and cumulative incidence of influenza by vaccine type was compared to estimate relative vaccine efficacy. Participants were tested for influenza by reverse-transcription polymerase chain reaction or rapid antigen tests. Immunogenicity was evaluated by hemagglutination inhibition (HAI), virus microneutralization (MN), and neuraminidase inhibition (NAI) assays in a subset of blood specimens collected pre-vaccination and one-month post-vaccination.</p><p><strong>Results: </strong>Among the 591 participants, 25 cases of influenza A were detected; 2.4% cumulative incidence among RIV recipients versus 6.1% among ccIIV recipients. The probability of influenza was 61% lower with RIV than ccIIV (95% confidence interval: 7 to 84). Among 151 participants, post-vaccination HAI and MN titers vaccine components were higher among RIV recipients while NAI titers were higher among ccIIV recipients.</p><p><strong>Conclusions: </strong>RIV provided better protection than ccIIV against influenza A illness in adults during the U.S. 2024-2025 season.</p>","PeriodicalId":94281,"journal":{"name":"medRxiv : the preprint server for health sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13648820/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149169420","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Bayesian Anchored Latent Learning for Estimating State Trajectories.","authors":"Jason B Gibbons, Nicolas Leenaerts, Amar D Mandavia, Jacob C Jameson, Annie Liang, David Hathaway, Jie Yang, Bowen Gu, Brian P Marx, Shan Siddiqi, Rachel Wood, Matt Ruble, Stanley Lyndon, Philip S Wang","doi":"10.64898/2026.09.30.26363325","DOIUrl":"https://doi.org/10.64898/2026.09.30.26363325","url":null,"abstract":"<p><p>Dynamic systems often evolve faster than reliable measurements can be collected. More frequent indirect observations may fill these gaps, but their noise and differing scales complicate estimation of the underlying state. We developed Bayesian Anchored Latent Learning (BALL) to combine sparse reference measurements with denser records on an interpretable scale. A bidirectional transformer teacher reconstructs trajectories from complete training histories, and a forward-only transformer student learns from those reconstructions and observed measurements. Across 30 simulations, BALL reduced composite trajectory error by 7% relative to the same transformer trained directly from measurements, achieving lower error across all datasets. In an applied example using psychiatric records from 1,609 patients, BALL improved the prediction of withheld questionnaire scores compared with direct transformer training at 21- and 28-day measurement schedules. Requests guided by estimated uncertainty further improved prediction. BALL connects retrospective learning with prospective state estimation and provides a framework for directing additional measurement.</p>","PeriodicalId":94281,"journal":{"name":"medRxiv : the preprint server for health sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13648800/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149168850","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Improving MASLD Identification in Patients with T2D: A Computable Phenotype Framework Integrating Structured Data and Clinical Notes.","authors":"Yuanhan Chen, Anu Sharma, Qiang Yang, Yuyao Chen, Rui Yin","doi":"10.64898/2026.09.26.26364078","DOIUrl":"https://doi.org/10.64898/2026.09.26.26364078","url":null,"abstract":"<p><p>Metabolic dysfunction-associated steatotic liver disease (MASLD) is highly prevalent among individuals with type 2 diabetes (T2D), yet accurate identification from electronic health records (EHRs) remains challenging because clinically relevant information is fragmented across structured data and unstructured clinical documentation. We developed a rule-based computable phenotype (CP) framework that integrates structured EHRs with clinical notes for scalable MASLD identification in patients with T2D. The framework comprises five complementary CP algorithms derived from validated fibrosis indices and liver injury biomarkers, together with a lightweight natural language processing module that identifies MASLD-related evidence using keyword matching, contextual filtering, and negation detection. The proposed framework was systematically evaluated across multiple threshold settings, with additional validation through age- and sex-stratified analyses and manual chart review. The findings demonstrate that integrating structured and unstructured EHR data provides an accurate and scalable framework for real-world MASLD phenotyping and offers a practical strategy for developing CPs for other chronic diseases.</p>","PeriodicalId":94281,"journal":{"name":"medRxiv : the preprint server for health sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13648749/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149169456","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"HPV-Stratified Causal Digital Twins for Personalized Survival Benefit Estimation in Head and Neck Cancer.","authors":"Keyur D Shah, Chih-Wei Chang, Jun Zhou, Jie Ding, Anees Dhabaan, Hania A Al-Hallaq, James Bates, William Stokes, David S Yu, Xiaofeng Yang","doi":"10.64898/2026.09.30.26363910","DOIUrl":"https://doi.org/10.64898/2026.09.30.26363910","url":null,"abstract":"<p><strong>Purpose: </strong>Digital-twin frameworks can potentially offer a pathway toward individualized treatment but remain largely untested in radiation oncology. We present a causal digital-twin model for head-and-neck (H&N) cancer that estimates patient-specific survival benefit from chemoradiotherapy (CRT) versus radiotherapy (RT) alone, stratified by HPV status.</p><p><strong>Methods: </strong>Using the public RADCURE cohort (n = 3,346), we developed a causal digital-twin framework in which patients were divided into HPV-positive and HPV-negative groups and assigned to training (RT start year ≤ 2007) and test sets (RT start year > 2007). Bayesian-network structure learning identified each cohort's Markov Blanket around the 2-year survival endpoint (SVy2), yielding the minimal causally relevant feature set. A causal-forest model (double machine learning, DML framework) estimated the individualized treatment effect <i>τ</i> = <i>P</i> ( <i>Svy2</i> | <i>CRT,X</i> ) - <i>P</i> ( <i>Svy2</i> | <i>RT,X</i> ), with denoting the model-estimated effect. Overlap restrictions were applied to mitigate extrapolation. Predictive discrimination (AUC, C-index), average treatment effects (ATE), individualized effects (ITE), and decision-curve analyses were computed.</p><p><strong>Results: </strong>For HPV-negative disease, internal evaluation demonstrated strong discriminative performance (test AUC = 0.72; C-index = 0.81, p < 0.01). The estimated test-set ATE favored CRT with a modest but positive improvement in SVy2 (mean ). Patterns in were clinically coherent, including increased benefit in node-positive and good-performance patients and reduced benefit in high-burden or poor-performance presentations. Decision-curve analysis showed that a -guided selective-CRT strategy provided higher net benefit than uniform treatment policies. HPV-positive disease exhibited substantially greater but less clinically structured treatment-effect heterogeneity, accompanied by reduced overlap and greater uncertainty.</p><p><strong>Conclusion: </strong>This study demonstrates a proof-of-concept causal digital-twin framework for interpretable, patient-specific survival simulation in H&N cancer. The model captures clinically plausible treatment-effect heterogeneity and establishes a foundation for multi-cohort validation and future integration into individualized treatment decision support.</p>","PeriodicalId":94281,"journal":{"name":"medRxiv : the preprint server for health sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13648796/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149171676","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}