Annals of Epidemiology最新文献

{"title":"Multilevel modeling in egocentric network analysis: A practical guide with SAS and R","authors":"Hongjie Liu PhD., MS","doi":"10.1016/j.annepidem.2025.12.009","DOIUrl":"10.1016/j.annepidem.2025.12.009","url":null,"abstract":"<div><h3>Purpose</h3><div>This paper illustrates the application of multilevel modeling to egocentric network data, where network alters are nested within their respective egos. The nested structure and intra-ego dependencies in such data violate the independence assumptions of traditional regression models.</div></div><div><h3>Methods</h3><div>Multilevel modeling addresses this dependency by accommodating hierarchical data structures, allowing for more accurate estimation of ego–alter associations. It also distinguishes the effects of variables measured at the alter, ego, and dyadic (ego–alter) levels on outcome variable. We describe model specifications involving random intercepts and slopes, cross-level interactions, and assumptions related to residuals and variance structures. An illustrative example is provided to demonstrate how to estimate fixed and random effects for both continuous and binary outcome variables, assess intraclass correlation, test cross-level interactions, and interpret model results.</div></div><div><h3>Results</h3><div>This paper serves as a practical guide for applying multilevel models to egocentric network data, outlining key conceptual foundations, methodological considerations, and step-by-step implementation using SAS and R.</div></div><div><h3>Conclusions</h3><div>The guide aims to support researchers in the social and health sciences in rigorously applying multilevel modeling to egocentric network data, fostering deeper insights into how individual, relational, and structural factors influence health-related outcomes.</div></div>","PeriodicalId":50767,"journal":{"name":"Annals of Epidemiology","volume":"113 ","pages":"Pages 64-70"},"PeriodicalIF":3.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145806257","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Impact of integrase strand transfer inhibitors on cardiovascular disease in people with HIV","authors":"Buwei He ,&nbsp;Bankole Olatosi ,&nbsp;Jiajia Zhang ,&nbsp;Sharon Weissman ,&nbsp;Xiaoming Li ,&nbsp;Xueying Yang","doi":"10.1016/j.annepidem.2025.11.006","DOIUrl":"10.1016/j.annepidem.2025.11.006","url":null,"abstract":"<div><h3>Purpose</h3><div>Integrase Strand Transfer Inhibitors (INSTIs) are effective and well-tolerated in HIV treatment, but their cardiovascular impact remains uncertain. This study evaluated the association between INSTI-based antiretroviral therapy (ART) and cardiovascular disease (CVD) risk in people with HIV (PWH).</div></div><div><h3>Methods</h3><div>We conducted a retrospective cohort study using electronic health records and survey data from the All of Us research program. Adults with HIV on continuous ART for ≥ 1 year, free of CVD at baseline and within one year of ART initiation, were included. INSTI exposure was categorized by regimen status. Cox proportional hazards models assessed the association between INSTI use and CVD. A subgroup analysis examined INSTI-naïve individuals who later switched to INSTI-based ART.</div></div><div><h3>Results</h3><div>Among 2175 PWH, 437 (20.09 %) experienced CVD events; 1715 (78.85 %) used INSTIs. INSTI-only (aHR = 0.65, 95 % CI: 0.50–0.83) and partial-INSTI use (aHR = 0.34, 95 % CI: 0.27–0.42) were linked to lower CVD risk. In the INSTI-naïve cohort (N = 1300), switching to INSTIs reduced CVD risk (aHR = 0.32, 95 % CI: 0.25–0.40). Older age (≥ 70) increased CVD risk.</div></div><div><h3>Conclusions</h3><div>INSTI-based ART may lower CVD risk among PWH. Further research is needed to validate these findings.</div></div><div><h3>Summary</h3><div>INSTI-based ART was associated with reduced cardiovascular disease risk among people with HIV, suggesting potential cardioprotective effects that warrant further investigation.</div></div>","PeriodicalId":50767,"journal":{"name":"Annals of Epidemiology","volume":"113 ","pages":"Pages 13-22"},"PeriodicalIF":3.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145610296","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multiple causes of death data to track trends of mortality from chronic diseases: Insights from the COVID-19 pandemic in Switzerland","authors":"D. Durán N MSc. ,&nbsp;J.S. Kaufman PhD ,&nbsp;A. Chiolero MD. PhD ,&nbsp;C. Carmeli PhD","doi":"10.1016/j.annepidem.2025.12.002","DOIUrl":"10.1016/j.annepidem.2025.12.002","url":null,"abstract":"<div><h3>Purpose</h3><div>With the rising prevalence of multi-morbidity among aging populations and disruptive events as new infectious disease threats, the traditional focus on the underlying cause of death (UCOD) can obscure the contribution of chronic diseases to mortality trends. The multiple cause of death (MCOD) approach addresses this limitation by considering all causes recorded on a death certificate. We aimed to quantify the effect of the COVID-19 pandemic on trends of chronic disease mortality by comparing the UCOD and MCOD approaches.</div></div><div><h3>Methods</h3><div>We conducted a population-based time series analysis using all deaths occurred 2011–2022 in Switzerland. We modelled pre-pandemic (2011–2019) trends to predict the expected chronic disease deaths during 2020–2022 had the pandemic not occurred. We quantified the monthly effect of the pandemic as the observed minus expected chronic disease deaths and estimated these effects by sex and age using (1) the UCOD and (2) the MCOD.</div></div><div><h3>Results</h3><div>Both approaches revealed similar overall trends of effects. However, a marked discrepancy occurred at the end of 2020, when Switzerland experienced the highest COVID-19 mortality, with MCOD quantifying a substantially higher excess of chronic disease deaths compared to UCOD.</div></div><div><h3>Conclusion</h3><div>MCOD complements UCOD in quantifying cause-specific mortality trends and may improve population health monitoring.</div></div>","PeriodicalId":50767,"journal":{"name":"Annals of Epidemiology","volume":"113 ","pages":"Pages 23-29"},"PeriodicalIF":3.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145696482","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Unravelling impact of comorbidities on mortality risks in CKD patients during the COVID-19 pandemic: An explainable AI-driven study","authors":"Zeinab Abdollahi ,&nbsp;Lin Huo ,&nbsp;Donald Fraser ,&nbsp;Shang-Ming Zhou","doi":"10.1016/j.annepidem.2025.11.005","DOIUrl":"10.1016/j.annepidem.2025.11.005","url":null,"abstract":"<div><h3>Objectives</h3><div>The chronic kidney disease (CKD) patients were at high risk for severe clinical complications during the COVID-19 pandemic. Our objectives were to evaluate comorbidity prevalence; predict mortality risks for CKD patients during the pandemic; assess how various health factors interact to influence mortality; and provide insights for targeted prevention strategies.</div></div><div><h3>Method</h3><div>We analysed data from 186,396 CKD patients in Mexico during the entire pandemic (Jan 2020- May 2023). Explainable artificial intelligence (XAI) methods with extreme gradient boosting (XGBoost) models and Shapley Additive Explanations (SHAP) were developed to predict mortality for CKD patients with model interpretations. Different metrics were used to comprehensively evaluate model’s generalisation performances.</div></div><div><h3>Results</h3><div>The most prevalent comorbidities were hypertension (64.39 %), diabetes (49.79 %), and obesity (16.46 %). Male patients and older individuals showed higher risk for adverse outcomes. The overall mortality rate was 19.33 %, with significantly higher mortality in COVID-19 positive patients (33.9 %) compared to COVID-19 negative patients (10.1 %). Comorbidities with the most significant impact on the mortality included diabetes, hypertension, and obesity, which were more frequent in the COVID-19 positive group and associated with higher rates of intubation, and ICU admission. Pneumonia was identified as a major predictor of negative outcomes in CKD patients with COVID-19. CVD was more common in the COVID-19 negative group. Our machine learning models achieved performances of AUC= 0.76 and F₁-score= 0.75 for predicting mortality during the pandemic.</div></div><div><h3>Conclusion</h3><div>Targeted management of comorbid conditions, especially respiratory infections, is crucial in CKD patients during pandemics.</div></div>","PeriodicalId":50767,"journal":{"name":"Annals of Epidemiology","volume":"113 ","pages":"Pages 1-12"},"PeriodicalIF":3.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145610297","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Injury preceding the classical recognition of multiple sclerosis: A population-based study","authors":"Fardowsa L.A. Yusuf PhD ,&nbsp;Mohammad Ehsanul Karim PhD ,&nbsp;Jason M. Sutherland PhD ,&nbsp;Feng Zhu MSc ,&nbsp;Yinshan Zhao PhD ,&nbsp;Ruth Ann Marrie MD, PhD ,&nbsp;Helen Tremlett PhD","doi":"10.1016/j.annepidem.2025.11.008","DOIUrl":"10.1016/j.annepidem.2025.11.008","url":null,"abstract":"<div><h3>Background</h3><div>We investigated the association between multiple sclerosis (MS) and fractures, dislocations/sprains/strains, and burns preceding MS recognition.</div></div><div><h3>Methods</h3><div>We conducted a cohort study using clinical and population-based health administrative data in British Columbia, Canada (1991–2020). We compared the risk of a fracture, dislocation/sprain/strain, and burn in the six years preceding an MS cases’ first demyelinating claim (administrative cohort=9197) or MS symptom onset (clinical cohort=1446) to that of matched general population controls using modified Poisson regression. As sensitivity analyses, we used high-dimensional propensity scores (hdPS) to address residual confounding and targeted maximum likelihood estimation (TMLE) for mis-specification.</div></div><div><h3>Results</h3><div>In the six years before the first demyelinating claim (administrative cohort), the risk of a fracture (adjusted relative risks [adjRR]=1.28;95 %CI:1.20–1.36), dislocation/sprain/strain (adjRR=1.20;95 %CI:1.15–1.23), and burn (adjRR=1.40;95 %CI:1.22–1.62) was higher among MS cases. After hdPS adjustment and TMLE, the adjusted relative risks decreased slightly: fracture (hdPS=1.20; TMLE=1.20), dislocation/sprain/strain (hdPS=1.15; TMLE=1.15), and burn (hdPS=1.25; TMLE=1.26). Pre-MS symptom onset (clinical cohort), the associations were weaker but in the same direction.</div></div><div><h3>Conclusion</h3><div>Fractures, dislocations/sprains/strains, and burns were more common among people with MS before its classical recognition, suggesting that MS could be detected earlier.</div></div>","PeriodicalId":50767,"journal":{"name":"Annals of Epidemiology","volume":"113 ","pages":"Pages 30-37"},"PeriodicalIF":3.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145702404","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Mediation learning module: Pedersen et al (2025), Associations of early life body size and pubertal timing with breast density and postmenopausal breast cancer risk: A mediation analysis","authors":"Jeb Jones PhD, MPH, MS","doi":"10.1016/j.annepidem.2025.10.021","DOIUrl":"10.1016/j.annepidem.2025.10.021","url":null,"abstract":"<div><div>Educational Engagement Modules (EEMs) are teaching materials for educators and students that facilitate a deeper understanding of key epidemiological methods and concepts. Each EEM poses a series of questions using a recently published paper in Annals to further understanding of a specific study design and to encourage critical thinking and careful evaluation. This EEM focuses on the use of mediation in a study exploring whether breast density mediates the relationship between early life characteristics and postmenopausal breast cancer: Pedersen DC, Hameiri-Bowen D, Aarestrup J, Jensen BW, Tjønneland A, Mellemkjær L, von Euler-Chelpin M, Vejborg I, Andersen ZJ, Baker JL. Associations of early life body size and pubertal timing with breast density and postmenopausal breast cancer risk: A mediation analysis. Ann Epidemiol. 2025 Feb;102:68–74. doi: 10.1016/j.annepidem.2025.01.004. Epub 2025 Jan 10. PMID: 39798680 <span><span>[1]</span></span>.</div></div>","PeriodicalId":50767,"journal":{"name":"Annals of Epidemiology","volume":"113 ","pages":"Pages 86-88"},"PeriodicalIF":3.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145938917","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Vaginal microbiome structure in pregnancy and host factors predict preterm birth: Results from the ECHO Cohort","authors":"Kimberly S. McKee PhD, MPH ,&nbsp;Christine M. Bassis PhD ,&nbsp;Jonathan Golob MD ,&nbsp;Beatrice Palazzolo MSc ,&nbsp;Sarah S. Comstock PhD ,&nbsp;Christian Rosas-Salazar MD, MPH ,&nbsp;Joseph B. Stanford MD ,&nbsp;Sen Ananda PhD ,&nbsp;Thomas O’Connor PhD ,&nbsp;James E. Gern MD ,&nbsp;Nigel Paneth MD, MPH ,&nbsp;Anne L. Dunlop MD, MPH ,&nbsp;for the ECHO Cohort Consortium","doi":"10.1016/j.annepidem.2025.11.003","DOIUrl":"10.1016/j.annepidem.2025.11.003","url":null,"abstract":"<div><h3>Purpose</h3><div>The vaginal microbiome is dynamic, typically shifting during pregnancy toward enrichment of Lactobacillus. However, proliferation of Lactobacillus may be absent among women with preterm births (PTBs). We sought to identify robust vaginal microbiota signatures along with host factors that predicted PTB across diverse U.S. cohorts.</div></div><div><h3>Methods</h3><div>We meta-analyzed 16S rRNA gene amplicon sequence data from the Environmental influences on Child Health Outcomes Cohort. We classified community state types (CSTs) and employed penalized logistic regression models to assess the association between vaginal CST and PTB. We generated supervised random forest models and validated them using a train-and-test approach to identify the most predictive vaginal taxa and host factors.</div></div><div><h3>Results</h3><div>Of 683 births, 12 % were preterm. Overall, 26 % had a non- L. iners Lactobacillus-dominant CST (I, II, V), 43 % had a L. iners-dominant CST (III), and 30 % had a diverse, non-Lactobacillus-dominant (IV-B, IV-C) CST. Vaginal CST was strongly associated with PTB (adjusted odds ratio [aOR], 3.86, 95 % confidence interval [CI], 1.57–11.3 for diverse, non-Lactobacillus-dominant communities and aOR, 3.03, 95 % CI, 1.25–8.78 for L. iners-dominant compared to L. crispatus-dominant communities). The model with the highest area under the curve (AUC=.77) included Gardnerella vaginalis, age, Prevotella timonensis, and L. crispatus.</div></div><div><h3>Conclusions</h3><div>Along with host factors, vaginal microbiota could be used for predictive risk scoring for PTB across different U.S. cohorts.</div></div>","PeriodicalId":50767,"journal":{"name":"Annals of Epidemiology","volume":"112 ","pages":"Pages 110-118"},"PeriodicalIF":3.0,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145566163","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Impact of school-based infection prevention strategies on household COVID-19 and respiratory disease outcomes: A cross-sectional study","authors":"Sanjana Pampati ,&nbsp;Elizabeth A. Stuart ,&nbsp;Justin Lessler ,&nbsp;Kirsten E. Wiens ,&nbsp;Lance A. Waller ,&nbsp;Benjamin Lopman ,&nbsp;Jodie L. Guest ,&nbsp;M. Kate Grabowski ,&nbsp;Jeb Jones","doi":"10.1016/j.annepidem.2025.10.019","DOIUrl":"10.1016/j.annepidem.2025.10.019","url":null,"abstract":"<div><h3>Purpose</h3><div>Schools used a wide range of infection prevention strategies during 2022 when there were surges of the Omicron variant of SARS-CoV-2, respiratory syncytial virus, and influenza. We examined data from the Spring semester of the 2021/2022 school year in the United States to describe use of school-based infection prevention strategies, factors associated with implementation, and associations with COVID-19 and respiratory disease related outcomes.</div></div><div><h3>Methods</h3><div>We analyzed data from January-June 2022 from the COVID-19 Trends and Impact Survey, a daily, cross-sectional survey on Facebook to examine associations between 11 school-based infection prevention strategies and 5 household COVID-19 and respiratory disease related outcomes (e.g., positive COVID-19 test, COVID-19 like illness). Analyses were restricted to parents with a child &lt; 18 years old who was attending school in-person (n = 228,624). Multivariable, quasibinomial regression models were fit, adjusting for child-, household-, county-level covariates and state.</div></div><div><h3>Results</h3><div>Universal mask requirements were associated with decreased odds and having no school-based extracurriculars with increased odds of all examined outcomes. Additional prevention strategies were associated with reduced odds of specific outcomes: modified cafeteria use, regular testing of teachers and staff, ventilation improvements, and restricted entry.</div></div><div><h3>Conclusions</h3><div>Findings affirm the protective effect of universal mask requirements in school settings. No school based extracurriculars being associated with increased odds of examined outcomes may be a result of reverse causation bias. State and local health departments and education agencies can provide technical assistance to schools to ensure appropriate adoption of strategies based on each school’s unique circumstances.</div></div>","PeriodicalId":50767,"journal":{"name":"Annals of Epidemiology","volume":"112 ","pages":"Pages 76-83"},"PeriodicalIF":3.0,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145460227","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Sexual orientation identity and attraction changes among youth: prospective versus retrospective measurement","authors":"Lynsie R. Ranker PhD, MPH ,&nbsp;Sabra L. Katz-Wise PhD ,&nbsp;Allegra R. Gordon ScD, MPH ,&nbsp;R. Korkodilos ,&nbsp;Ziming Xuan ScD, SM, MA ,&nbsp;Kimberly M. Nelson PhD, MPH","doi":"10.1016/j.annepidem.2025.10.006","DOIUrl":"10.1016/j.annepidem.2025.10.006","url":null,"abstract":"<div><h3>Purpose</h3><div>Prior work has documented sexual orientation change, or fluidity, among youth over time. Some studies measure change via retrospective recall while others assess change prospectively. It is unclear whether retrospective recall and prospective assessment measure change experiences similarly. The current study examines whether prospectively assessed change in sexual orientation identity (SOI) and attractions aligns with change as assessed via retrospective recall of change.</div></div><div><h3>Methods</h3><div>US youth (N = 1235), aged 14–25 years, participated in a longitudinal online cohort. Prospective change in SOI and attraction were determined by comparing self-reported SOI and attraction at baseline and 4-month follow-up. Retrospective recall of SOI and attraction change was assessed at 4-month follow-up. The sensitivity and specificity of retrospective recall of change was assessed in relation to change detected prospectively.</div></div><div><h3>Results</h3><div>Prospectively, 12.8 % of youth changed their reported SOI between baseline and follow-up. Retrospectively, only 5.0 % recalled change for the same period. Prospectively, 21.9 % reported attraction changes between baseline and follow-up. Retrospectively, only 12.4 % recalled attraction changes. Only 17.3 % (95 % CI 11.3, 23.4) who prospectively reported SOI change and 28.7 % (95 % CI 23.3, 34.2) who prospectively reported attraction change also reported change retrospectively.</div></div><div><h3>Conclusions</h3><div>The prevalence of change in SOI and attractions varied substantially by whether change was assessed prospectively or retrospectively. Researchers and clinicians should consider how the method and timing of assessment may influence their ability to detect changes in SOI and attractions within youth populations. Identification of youth experiencing change is critical for allocation of relevant resources and support.</div></div>","PeriodicalId":50767,"journal":{"name":"Annals of Epidemiology","volume":"112 ","pages":"Pages 8-14"},"PeriodicalIF":3.0,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145253604","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A machine learning approach for a 15-year prediction model of liver cancer incidence: Results from two large Chinese population cohorts","authors":"Yu-Xuan Xiao ,&nbsp;Yi-Xin Zou ,&nbsp;Zhuo-Ying Li ,&nbsp;Qiu-Ming Shen ,&nbsp;Da-Ke Liu ,&nbsp;Yu-Ting Tan ,&nbsp;Hong-Lan Li ,&nbsp;Yong-Bing Xiang","doi":"10.1016/j.annepidem.2025.10.015","DOIUrl":"10.1016/j.annepidem.2025.10.015","url":null,"abstract":"<div><h3>Background</h3><div>Primary liver cancer (PLC) remains a major public health concern, particularly in China where the incidence is high. Existing prediction models often focus on high-risk populations and depend heavily on laboratory data, which limits their utility in general population screening.</div></div><div><h3>Methods</h3><div>We developed and validated a 15-year PLC risk prediction model using data from two large prospective cohort studies in Shanghai (n = 132,360), including 618 incident PLC cases. Candidate variables encompassed sociodemographic characteristics, lifestyle behaviors, medical history, and dietary factors. Predictor selection was performed using LASSO regression and the Boruta algorithm. Five machine learning models and logistic regression were compared. Model performance was evaluated using AUC, calibration plots and net reclassification improvement (NRI). SHapley Additive exPlanations (SHAP) were used to interpret model predictions. Web-based tools, including a simplified risk calculator, were developed to facilitate practical application.</div></div><div><h3>Results</h3><div>LightGBM achieved the best discrimination (AUC = 0.766) and excellent calibration. Net reclassification analysis indicated an improved ability to correctly classify low-risk individuals. The model effectively stratified the population: the high-risk group had a 15-year PLC risk that was 39.56 times that of the low-risk group. SHAP analysis revealed biologically meaningful associations. A simplified logistic model with fewer variables also performed well (AUC = 0.762), supporting effective risk stratification.</div></div><div><h3>Conclusion</h3><div>We developed a questionnaire-based 15-year PLC risk prediction model applicable to the general Chinese population. Both the full and simplified models demonstrated strong performance and interpretability, making them valuable tools for large-scale screening and targeted prevention, especially in resource-limited settings.</div></div>","PeriodicalId":50767,"journal":{"name":"Annals of Epidemiology","volume":"112 ","pages":"Pages 28-37"},"PeriodicalIF":3.0,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145318874","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}