Ferritin and transferrin predict common carotid intima-media thickness in females: a machine-learning informed individual participant data meta-analysis

Ferritin and transferrin predict common carotid intima-media thickness in females: a machine-learning informed individual participant data meta-analysis | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (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];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Ferritin and transferrin predict common carotid intima-media thickness in females: a machine-learning informed individual participant data meta-analysis View ORCID Profile Anand Ruban Agarvas , Richard Sparla , View ORCID Profile Janice L Atkins , View ORCID Profile Claudia Altamura , View ORCID Profile Todd Anderson , View ORCID Profile Ebru Asicioglu , View ORCID Profile Judit Bassols , View ORCID Profile Abel López-Bermejo , Hana Marie Dvořáková , View ORCID Profile José Manuel Fernández-Real , View ORCID Profile Christoph Hochmayr , Michael Knoflach , View ORCID Profile Jovana Kusic Milicevic , Silvia Lai , View ORCID Profile José María Moreno-Navarrete , View ORCID Profile Dariusz Pawlak , View ORCID Profile Krystyna Pawlak , Petr Syrovatka , View ORCID Profile Dorota Formanowicz , View ORCID Profile Pavel Kraml , View ORCID Profile Jose M Valdivielso , View ORCID Profile Luca Valenti , View ORCID Profile Martina U. Muckenthaler doi: https://doi.org/10.1101/2025.04.30.25326720 Anand Ruban Agarvas 1 Heidelberg University , Heidelberg, Germany 2 Center For Translational Biomedical Iron Research, Department of Pediatric Hematology , Oncology Immunology and Pulmonology, Heidelberg University Hospital , Heidelberg, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Anand Ruban Agarvas Richard Sparla 1 Heidelberg University , Heidelberg, Germany 2 Center For Translational Biomedical Iron Research, Department of Pediatric Hematology , Oncology Immunology and Pulmonology, Heidelberg University Hospital , Heidelberg, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Janice L Atkins 3 Department of Clinical & Biomedical Sciences, Faculty of Health & Life Sciences, University of Exeter , United Kingdom Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Janice L Atkins Claudia Altamura 4 Unit of Headache and Neurosonology, Department of Medicine and Surgery, Università Campus Bio-Medico di Roma , Rome, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Claudia Altamura Todd Anderson 5 Department of Cardiac Sciences and Libin Cardiovascular Institute of Alberta (TJA), Cumming School of Medicine, University of Calgary , Alberta, Canada Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Todd Anderson Ebru Asicioglu 6 Pendik Training and Research Hospital, Marmara University , Istanbul, Turkey Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Ebru Asicioglu Judit Bassols 7 Institut d’Investigació Biomèdica de Girona (IdIBGi), Hospital Trueta & Universitat de Girona , Girona, Spain Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Judit Bassols Abel López-Bermejo 7 Institut d’Investigació Biomèdica de Girona (IdIBGi), Hospital Trueta & Universitat de Girona , Girona, Spain Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Abel López-Bermejo Hana Marie Dvořáková 8 Department of Neonatology, University Hospital Prague-Motol , Prague, Czech Republic Find this author on Google Scholar Find this author on PubMed Search for this author on this site José Manuel Fernández-Real 9 Institut d’Investigació Biomèdica de Girona (IdIBGi), Hospital Trueta & Department of Medical Sciences, School of Medicine , Universitat de Girona, Girona, Spain 10 CIBERobn Fisiopatologia de la Obesidad y Nutrición , Madrid, Spain Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for José Manuel Fernández-Real Christoph Hochmayr 11 Department of Pediatrics II (Neonatology), Medical University of Innsbruck , Innsbruck, Austria Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Christoph Hochmayr Michael Knoflach 12 Department of Neurology, Medical University of Innsbruck , Innsbruck, Austria 13 VASCage, Centre on Clincial Stroke Research , Innsbruck, Austria Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jovana Kusic Milicevic 14 Department of Nephrology, Clinical Hospital Centre Zemun, Faculty of Medicine, University of Belgrade , Belgrade, Serbia Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Jovana Kusic Milicevic Silvia Lai 15 Department of Translational and Precision Medicine , Nephrology Unit, Sapienza University of Rome , Rome, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site José María Moreno-Navarrete 7 Institut d’Investigació Biomèdica de Girona (IdIBGi), Hospital Trueta & Universitat de Girona , Girona, Spain Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for José María Moreno-Navarrete Dariusz Pawlak 16 Department of Pharmacodynamics, Medical University of Bialystok , Poland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Dariusz Pawlak Krystyna Pawlak 17 Department of Monitored Pharmacotherapy, Medical University of Bialystok , Poland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Krystyna Pawlak Petr Syrovatka 18 Cardiocentre, Institute for Clinical and Experimental Medicine , Prague, Czech Republic Find this author on Google Scholar Find this author on PubMed Search for this author on this site Dorota Formanowicz 19 Department of Medical Chemistry and Laboratory Medicine, Poznan University of Medical Sciences , Poznan, Poland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Dorota Formanowicz Pavel Kraml 20 Charles University and University Hospital Královské Vinohrady , Prague, Czech Republic Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Pavel Kraml Jose M Valdivielso 21 Vascular and Renal Translational Research Group , IRBLleida, Lleida, Spain Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Jose M Valdivielso Luca Valenti 22 Department of Pathophysiology and Transplantation, Università degli Studi di Milano , Italy 23 Precision Medicine, SS Centro Risorse Biologiche, Fondazione IRCCS Ca’ Granda Policlinico , Pad Marangoni via F Sforza 35, 20122, Milano, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Luca Valenti Martina U. Muckenthaler 1 Heidelberg University , Heidelberg, Germany 2 Center For Translational Biomedical Iron Research, Department of Pediatric Hematology , Oncology Immunology and Pulmonology, Heidelberg University Hospital , Heidelberg, Germany 24 Molecular Medicine Partnership Unit , Heidelberg, Germany 25 Translational Lung Research Center Heidelberg (TLRC), German Center for Lung Research (DZL). 26 German Centre for Cardiovascular Research (DZHK) , Partner Site Heidelberg/Mannheim, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Martina U. Muckenthaler For correspondence: martina.muckenthaler{at}med.uni-heidelberg.de Abstract Full Text Info/History Metrics Supplementary material Preview PDF Abstract Background Iron overload promotes atherosclerosis in mice and causes vascular dysfunction in humans with Hemochromatosis. However, data are controversial on whether systemic iron availability within physiological limits affects the pathogenesis of atherosclerosis. We, therefore, performed an individual participant data (IPD) meta-analysis and studied the association between serum iron biomarkers with common carotid intima-media thickness (CC-IMT); in addition, since sex influences iron metabolism and vascular aging, we studied if there are sex-specific differences. Methods We pooled the IPD and analysed the data on adults (age≥18y) by orthogonal approaches: machine learning (ML) and a single-stage meta-analysis. For ML, we tuned a gradient-boosted tree regression model (XGBoost) and subsequently, we interpreted the features using variable importance. For the single-stage metaanalysis, we examined the association between iron biomarkers and CC-IMT using spline-based linear mixed models, accounting for sex interactions and study-specific effects. To confirm robustness, we repeated analyses on imputed data using multivariable regression adjusted for key covariates identified through machine learning. Further, subgroup analyses were performed in children and adolescents (age<18y). In addition, to evaluate causality, we used UK Biobank data to examine associations between the hemochromatosis (HFE) genotypes (C282Y/H63D) and mean CC-IMT in ∼42,500 participants with carotid ultrasound data, using sex-stratified linear regression (adjusted for age, assessment centre, and genetic principal components). Results We included IPD from 21 studies (N=10,807). The application of the ML model showed moderate predictive performance and identified iron biomarkers (transferrin, ferritin, transferrin saturation, and iron) as key features for IMT prediction. Multivariable analyses showed non-linear sex-specific relationships for ferritin and transferrin with CC-IMT: ferritin showed a significant positive association, and transferrin showed negative associations at specific ranges, both only among females. No significant associations were found between CC-IMT in those with HFE genotypes in either sex in the UK Biobank. Conclusion Our observational data show that iron biomarkers - ferritin and transferrin are non-linearly associated with CC-IMT specifically in females, while a significant causal association between the HFE genotype and CC-IMT could not be demonstrated in the UK Biobank data. We conclude that the observational associations may not only be explained by causal effects of iron on the arterial wall thickness, but also in part be driven by residual confounding factors such as inflammation. Other: No financial support was received for this meta-analysis. The protocol for this study is registered in the PROSPERO database (CRD42020155429; https://www.crd.york.ac.uk/ ). 1. Introduction Iron is a vital nutrient involved in numerous cellular functions; however, iron accumulation can be pathological. We have previously shown that excess iron accelerates the pathogenesis of atherosclerosis( 1 ). Excess iron initiates redox reactions (via Fenton chemistry) that lead to oxidative damage of membrane lipids, proteins, and DNA. In addition, iron can also lead to endothelial injury, immune cell polarization, ferroptosis, and plaque destabilization, thereby, contributing to vascular dysfunction and atherosclerosis ( 1 – 3 ). In the blood, iron biomarkers reflective of the systemic iron status [e.g., serum iron, ferritin, transferrin, transferrin saturation (TSAT), total iron binding capacity (TIBC) and hepcidin] can be quantified. Whether these iron biomarkers within physiological limits are associated with vascular disease is of clinical relevance, but available data are conflicting. Sonographic assessment of common carotid intima-media thickness (CC-IMT) is an indicator of arteriopathy and an independent predictor of a wide range of cardiovascular events ( 4 , 5 ). Therefore, by considering CC-IMT as a surrogate marker of vascular disease we aimed to investigate its association with iron biomarkers. We also wanted to investigate the influence of sex on the association between iron biomarkers and CC-IMT. Sex-specific differences in cardiovascular disease ( 6 ) and iron metabolism ( 7 , 8 ) are commonly observed. Specifically, the female sexual hormones estrogen and progesterone impair endothelial function and affect iron homeostasis by controlling hepcidin, the master regulator of iron homeostasis ( 9 , 10 ). Our recent analysis in two cohorts also found a positive association between ferritin and peripheral arterial disease preferentially in females, suggesting sex-specific effects ( 11 ). Our study also aimed to explore early vascular changes across a broader age spectrum. For example, in autopsies, atherosclerotic fibrous plaques have been detected in the aorta and coronary artery participants between 2-15 years ( 12 ). Therefore, we also included children and adolescents in subgroup analyses, to capture the onset and progression of such changes before clinical disease becomes apparent. We analysed them separately from adults since CC-IMT is strongly influenced by age and that risk factors for atherosclerosis for adults and adolescents are different. The conclusions of previous, smaller studies that have evaluated the relationship between iron biomarkers and CC-IMT are inconsistent. Therefore, we conducted an individual participant data (IPD) meta-analysis to systematically investigate this clinically relevant question. The advantages of an IPD meta-analysis cannot be overstated: the collection and analysis of raw data from multiple studies provides a large participant-pool and thus a greater statistical power to detect smaller effect sizes ( 13 ). Working with raw data allows for standardized and flexible analyses, leading to more precise and reliable estimates. We included a novel approach to the meta-analysis by using two orthogonal approaches that both offer unique advantages and drawbacks: machine learning (ML) and regression. Integration of ML approaches with regression-based analyses is an active area of research ( 14 , 15 ), which is still being methodologically optimized. Here, we aimed to combine the strengths of predictive modelling with the interpretability and inferential capacity of traditional statistical methods. ML excels at handling complex datasets, identifying nonlinear relationships, and making predictions without requiring strong assumptions about the data. However, one of the criticisms of ML is that it lacks interpretability. On the other hand, regression analysis provides interpretability and enables testing of relationships between variables under specific assumptions. However, it may struggle with high-dimensional data and complex nonlinear patterns, as is the case with clinical data. Therefore, we used a combination of the approaches to leverage their strengths and to obtain a clear understanding of the relationships. Finally, to complement the observational analyses of the iron biomarkers, we performed a Mendelian randomization (MR) to assess the potential causal relationship between hemochromatosis (HFE) genotypes (C282Y/H63D) and CC-IMT in the UK Biobank data. 2. Methods The protocol for the study is published in the PROSPERO database (CRD42020155429; https://www.crd.york.ac.uk/ ; Supplementary File 1). We report the study according to Preferred Reporting Items for Systematic Review and Meta-Analyses of individual participant data (PRISMA-IPD; See Supplementary File 2 for checklist ( 16 )). We used R version 4.3.1 ( 17 ) for the data analysis and visualization (For the specific packages used, see Supplementary Table 1 ). 2.1 Literature search We searched NLM Medline using the following string: (iron OR ferritin OR transferrin OR hepcidin) AND (atherosclerosis OR intima-media thickness). We applied filters for human studies published in English between 1 st Oct 1999 and 20 th Oct 2019 (last updated on 24 th Aug 2023). We used a three-step process for the study selection. In step 1, we screened the titles of the published studies and excluded records when the titles were specified as reviews or were performed on preclinical models. In step 2, we screened the abstract for human studies with one of the following keywords: iron, ferritin, transferrin, hepcidin, atherosclerosis or intima-media thickness. In step 3, we screened the full text and if studies contained data on iron biomarkers [iron, ferritin, transferrin, hepcidin, or transferrin saturation (TSAT)] and CC-IMT, they were included in the IPD. Screening of the retrieved records was done independently by two investigators (ARA, RS). 2.2 Data collection We contacted the investigators (first or corresponding authors) of eligible studies with the study protocol and requested formal consent for participation (for a scheme of data request, see Supplementary Figure 1 ). Through subsequent contact, we asked the investigators for anonymized data on the following variables: age, gender, CC-IMT, serum iron indices (iron, ferritin, transferrin, TSAT, hepcidin), ethnic profile, and presence of comorbidities [e.g., diabetes, hypertension, chronic kidney disease (CKD), hemochromatosis, thalassemia]. For prospective studies, only the baseline data were obtained. We sent three follow-up reminders to investigators who had not responded. From the studies for which IPD were received, we extracted the data and screened for inclusion in the meta-analysis. We piloted the data extraction by harmonizing the coding of categorical variables (e.g., gender, comorbidities) and their units of measurements, in the case of quantitative variables ( Table 2 ). At this stage, we also compared (variables available in the received datasheets and their original publications) and identified variables common to various studies. When required, we contacted investigators again to request additional variables of interest. Subsequently, we reextracted the data and checked the files for data integrity in three steps. In step 1, we compared the number of data participants, sex ratio, and the summary data of variables between the data file and its corresponding original paper. In the step 2, we verified the frequency distribution of continuous variables of interest in the individual data files. If required, we contacted the authors in a third step for clarifications. We excluded studies that did not clear the data integrity check. 2.3 Participant-level data From the received data, all participants with an available CC-IMT measurement were selected for pooling from the selected studies. Next, we selected the commonly measured serum iron biomarkers (iron, ferritin, transferrin, and TSAT) and the demographic and laboratory variables: age, sex, body mass index (BMI), smoking status, presence of comorbidities (diabetes, hypertension, CKD, thalassemia, hemochromatosis), creatinine, hemoglobin, high-density lipoprotein cholesterol (HDLc), low-density lipoprotein cholesterol (LDLc), triacylglycerols, fasting glucose, c-reactive protein (CRP), systolic blood pressure (SBP), and diastolic blood pressure (DBP). When discrete CC-IMT measurements were available on the right and left carotid arteries, the mean CC-IMT was calculated and used for downstream analyses (without further transformations). The data on age, sex, BMI, and the presence of diabetes, hypertension, CKD, thalassemia, and hemochromatosis were used, as indicated, in the original data files. Since the reporting units of the laboratory variables were not uniform, this required harmonization by conversion factors (e.g., mg/dL to mmol/L). Subsequently, we pooled the IPD and added our own data ( 11 ) from 323 individuals from the Heidelberg Study on Diabetes and Complications (HEIST-DiC study; https://clinicaltrials.gov , NCT03022721 ). At the participant level, we applied the following exclusion criteria: 1) CC-IMT value suggestive of an atherosclerotic plaque ( 18 ) (>1.5 mm) 2) diagnosed thalassemia or hemochromatosis 3) TSAT and ferritin values suggestive of possibly hemochromatosis ( 19 ) and, 4) CRP>10 mg/dL suggesting overt inflammation ( 20 ). We also analyzed the data for missing values and imputed them for downstream analyses (described below). 2.4 Implementation of the ML framework For the ML analysis, only adult participants (age≥18y) with a valid value for CC-IMT and for one of the iron biomarkers (iron, ferritin, transferrin or TSAT) were included. We converted categorical variables into numerical variables and split the data subsequently into Training and Test subsets stratified based on study (to ensure that proportion of each study in these subsets was comparable to the original dataset). In addition, to avoid information leakage between these subsets, we processed them independently with the following steps: imputation of missing values by bagged trees, removed variables with zero or near-zero variance, centering and scaling of numerical variables. We implemented a gradient-boosted tree regression model using XGBoost ( 21 ). We used the Training subset further for hyperparameter tuning (using grid search) and performed a 10-fold cross-validation. We evaluated model performance using Root Mean Squared Error (RMSE) as the primary metric and finalized the best-performing model. We subsequently predicted CC-IMT values on the held-out Test subset and quantified the variables that are of importance for the model prediction. We also used SHAP (SHapley Additive exPlanations ( 22 )) values to interpret the model and to assess feature contributions. For calculation of SHAP values, we used both the Training and Test subsets. The ML pipeline is shown in Figure 2b . 2.5 Regression analyses For the regression analyses, since iron and atherosclerosis parameters show age-specific variations, we used data from adults (age≥18y) for the main analysis. Here, we first tested the association between CC-IMT and each of the iron biomarkers (iron, ferritin, transferrin, and TSAT) by linear mixed model regression in the complete (unimputed) data. We hypothesized that the relationship could be nonlinear, and therefore, flexibly modelled the relationship using spline regression (degree of freedom=4). In addition, since iron metabolism shows strong sex-specific differences, we specified it using an interaction term (e.g. ferritin*sex) in the models. This approach is widely accepted in epidemiological and clinical research when exploring whether the association between a predictor (in this case, ferritin, as a proxy for iron metabolism) and an outcome (CC-IMT) differs by a stratifying variable (sex). The use of interaction terms allows for formal statistical testing of effect modification by sex, rather than relying on stratified analyses alone ( 23 ). Further, to account for differences between the different datasets, we included “study” as a random effect in the model. To further test the robustness of the observations from the previous step, we conducted multivariable regression analysis between CC-IMT and ferritin or transferrin. Here, we imputed the missing values based on the approach by Gibbs sampler ( 24 , 25 ) and generated multiple imputations (n=5) from a joint multivariable linear-mixed model. For continuous variables, the method considers the relationships between all variables at once, while for categorical variables, the model assumes that they are linked to underlying continuous variables that follow a normal distribution ( 26 ). We performed the regression on the independent imputation draws, using Rubin’s rule ( 27 ) and obtained the final imputed results. We included variables from the feature importance in ML analysis as additional covariates for the multivariable regression: smoking (reference=nonsmokers), the presence of diabetes (reference=absence), age, BMI, creatinine, HDLc, LDLc, triacylglycerols, CRP, hemoglobin, SBP, and DBP. All continuous variables (except CC-IMT) were mean-centered for the analyses ( 28 ). 2.5.1 Subgroup analysis We conducted a subgroup analysis for children and adolescents (age< 18y). Here, we also used spline regression (degree of freedom=4) and used sex as an interaction with iron biomarker (as above). However, since not all studies had children and adolescents in their study population, we fitted using linear regression (without including study as a random effect). 2.6 UK Biobank Further analyses were performed in the UK Biobank to determine the associations between hemochromatosis (HFE)-genotype groups and CC-IMT. UK Biobank includes ∼500,000 community volunteers aged 39-73 years at baseline assessment (2006–2010) from 22 assessment centers across England, Scotland and Wales [as described elsewhere ( 29 , 30 )]. We included participants genetically similar to the 1000 Genomes project European reference population ( 31 ) [the categorization of this population is described elsewhere ( 32 )], with HFE p.C282Y (rs1800562) and HFE p.H63D (rs1799945) genotype data from whole exome sequencing [methods developed by Regeneron ( 33 )]. We analyzed a subset of these participants with available carotid ultrasound data from an imaging visit starting in 2014 [n=42,299; ( 34 )]. Four CC-IMT variables were available in the UK Biobank imaging study [at 120, 150, 210 and 240 degrees; variable IDs 22671, 22674, 22677, 22680( 34 )]; from these, we calculated an overall mean CC-IMT value and performed linear regression analyses to test associations with C282Y/H63D genotype groups [C282Y (−/-) H63D (+/-) ; C282Y (-/-) H63D (+/+) ; C282Y (+/-) H63D (+/-) ; C282Y (+/-) H63D (-/-) ; C282Y (+/+) H63D (-/-) ], compared to those with no mutations [C282Y (-/-) H63D (-/-) ]. Models were stratified by sex, and adjusted for age, assessment centre and ten genetic principal components (to account for genetic stratification). We also performed sensitivity analyses after excluding participants diagnosed with hemochromatosis. 3. Results 3.1 Process of data collection We identified a total of 1,032 records via literature search, of which 887 were excluded by screening the title and abstract. The full text of the remaining 145 articles was investigated, and 108 publications were selected for the meta-analysis. We contacted the authors of the selected publications, from which we received IPD data from 22 studies (IPD retrieval 20.4%)( 35 – 55 ). Of these, we excluded two studies that did not clear the data integrity checks( 54 , 55 ). Additionally, we included our own data from the HEIST-DiC study ( 11 ). The flow of the literature search is shown in Figure 1 and the outline of the analysis is shown in Figure 2a . Figure 1. The outline of the study and analysis Figure 2. a) The flow of literature in the IPD. b) The machine learning pipeline 3.2 Study characteristics Eighteen studies included in the IPD were hospital-based ( 11 , 35 , 36 , 38 – 46 , 48 – 52 , 56 ) and 3 studies were population-based ( 37 , 47 , 53 ). Controls were part of the study population in 12 studies ( 35 , 36 , 38 – 40 , 44 – 46 , 49 – 51 , 53 ). The characteristics of studies and variables included in the IPD-MA are shown in Table 1 and Supplementary Table 2 . View this table: View inline View popup Table 1. Characteristics of studies included in the IPD-MA Table of the characteristics of studies that provided the IPD and included in the meta-analysis. 3.3 Participant characteristics The pooling of all available datasets yielded a total of 10,807 participants. From these, we excluded participants with CC-IMT value indicative of atherosclerotic plaque( 18 ) (>1.5 mm; N=47), CRP indicative of overt inflammation (>10 mg/dL; N=167) and all conditions suggestive of an elevated iron status [possible hemochromatosis( 19 ) (N=27), previously diagnosed thalassemia (N=131) or hemochromatosis (N=76)]. This resulted in a final pooled participant size of 10,215. Most of the participants were adults except for 4 studies( 43 , 45 , 51 , 53 ) which had included children and adolescents in their work (age< 18y). Eighteen studies( 11 , 35 – 38 , 42 – 53 ) included male and female participants, while 3 studies( 39 – 41 ) included only males in their study population. A study-wise breakdown of participant characteristics is shown in Supplementary Table 3 . 3.3.1 Adults The demographics of the adult participants (N=7,523) are shown in Table 2 . Here, the proportion of males (N=4,974; 66.12%) was higher than females (N=2,549; 33.88%). Overall, the males were younger, had a higher BMI and a greater proportion of smokers (p=0.009). The proportion of participants with hypertension (p<0.0001) was also higher among males; in line with this observation, the systolic blood pressure (SBP) and diastolic blood pressure (DBP) of males were higher ( Table 2 ). On the other hand, more female participants in the cohort had CKD (44.55%; p<0.0001; Table 2 ). Serum iron, ferritin, and TSAT were higher, while transferrin levels were lower among males than females ( Table 2 ; Supplementary Figures 2 - 5 ). View this table: View inline View popup Table 2. Characteristics of adult participants The demographics of the adult participants (age≥18y) stratified by sex are shown. For continuous variables, the summary data are shown as Median (IQR) while for categorical variables, the data are represented as N (%). P-values as calculated by Wilcoxon rank sum test (for continuous variables) or Pearson’s Chi-Squared test (for categorical variables); significant P-values are highlighted in bold. Missing data are shown as N (%). All percentages shown are based on the total number of subjects. BMI: Body Mass Index, HDLc: High-Density Lipoprotein, LDLc: Low-Density Lipoprotein, CRP: C-reactive protein, CKD: Chronic Kidney Disease, CC-IMT: Carotid Intima-Media Thickness, TSAT: Transferrin Saturation, SBP: Systolic Blood Pressure, DBP: Diastolic Blood Pressure 3.3.2 Children and Adolescents Among children and adolescents, age (p=0.131) and BMI (p=0.838) of the participants among both sexes were not different ( Table 3 ). The proportion of smokers was higher among females than males (76% vs 24%; p=0.002). Overall, females had lower hemoglobin levels (p98%; Table 3 ). With regards to the iron parameters, males had higher ferritin and transferrin levels while iron and TSAT levels were not different between the sexes ( Table 3 ; Supplementary Figures 2 - 5 ) View this table: View inline View popup Table 3. Characteristics of children and adolescents The demographics of the children and adolescent participants (age< 18y) stratified by sex are shown. For continuous variables, the summary data are shown as Median (IQR) while for categorical variables, the data are represented as N (%). P-values as calculated by the Wilcoxon rank sum test (for continuous variables) or Pearson’s Chi-Squared test (for categorical variables); significant P-values are highlighted in bold. Missing data are shown as N (%). BMI: Body Mass Index, HDLc: High-Density Lipoprotein, LDLc: Low-Density Lipoprotein, CRP: C-reactive protein, CKD: Chronic Kidney Disease, CC-IMT: Carotid Intima-Media Thickness, TSAT: Transferrin Saturation, SBP: Systolic Blood Pressure, DBP: Diastolic Blood Pressure 3.4 CC-IMT In all studies, the CC-IMT was measured using ultrasound, but there were differences in the methods used ( Supplementary Table 4 ). Among the adults, the CC-IMT [median (IQR); mm] of males [0.74 (0.6-0.85)] was higher compared to females [0.71 (IQR 0.6-0.82); p<0.0001; Supplementary Figure 6 ]. This was also the case among children and adolescents, with higher CC-IMT among males [0.41 (0.37-0.44)] compared to females [0.39 (0.35-0.43); p<0.0001; Supplementary Figure 6 ]. 3.5 Missing data We observed two types of missing values in the IPD data: 1) systematically missing variables were present across studies (since not all studies collected the variables uniformly) 2) sporadically missing values within each study. The proportion of missing values for each variable in the pooled data is shown in Supplementary Table 5 . A study-wise breakdown of missing variables is shown in Supplementary Table 6 and an age-categorized breakdown is shown in Supplementary Table 7. The pattern of co-occurrence of missing values across variables is shown in Supplementary Figure 7 . 3.6 Machine learning The dataset used for ML contained 5,740 participants and 21 covariates. The Training subset contained 4,306 (75%) participants while the Test subset contained 1,434 (25%) participants. The baseline characteristics between the two subsets were comparable ( Supplementary Table 8 ). The ML model demonstrated moderate predictive performance with an R2 of 0.447, indicating that it explained 44.7% of the variance in the outcome variable; the Root Mean Square Error (RMSE) and the Mean Average Error (MAE) of the model were 0.136 and 0.101, respectively. The feature importance plot shows the contribution of each variable for the prediction of the overall model ( Figure 3a ). On the other hand, the SHAP plot highlights how the variables and their individual values contribute to the prediction ( Figure 3b ). Together, these plots emphasize the importance of each variable in our ML model’s predictions. All the iron biomarkers (transferrin, ferritin, TSAT and iron) were ranked among important predictors of CC-IMT. In addition, age, sex, CRP, creatinine, LDLc, HDLc, hemoglobin, systolic and diastolic BP, triacylglycerols, and smoking were also identified as important features. Figure 3. a) Variables of importance identifed in the machine learning prediciton b) SHAPley plot of the variables contributing to the model performance 3.7 Regression analyses 3.7.1 Adults In the analysis of unimputed data, we found that ferritin alone showed a positive effect within specific ranges [131-233 ng/mL: β=0.08, 95% CI (0.002, 0.16), p=0.046; >233 ng/mL: β=0.16, 95% CI (0.04, 0.27), p=0.008]. A significant interactive effect was also observed between females and ferritin with CC-IMT [ferritin>233 ng/mL: β=0.04, 95% CI (0.002, 0.08), p=0.037; Supplementary Table 9 ]. The main effects of transferrin alone were all non-significant, however negative interactions were noted for transferrin within specific ranges among females [231-263 mg/dL: β=-0.21, 95% CI (−0.43, 0.003), p=0.054; >263 mg/dL: β=-0.73, 95% CI (−1.48, 0.01), p=0.055; Supplementary Table 10 ]. On the other hand, none of the terms for iron or TSAT (including interactions with sex) showed statistically significant associations with CC-IMT (Supplementary Tables 11-12). In the multivariable analyses ( Tables 4 - 5 ), we confirmed potential sex-specific effects at higher levels of ferritin and transferrin. We found that ferritin alone showed a significant effect at a specific range (131-233 ng/mL: β = 0.13, 95% CI [0.02, 0.24], p = 0.038). Further, we found a positive interaction between ferritin with CC-IMT among females with CC-IMT specifically at the higher ranges (>233 ng/mL: β = 0.54, 95% CI [0.11, 0.97], p = 0.015). Although transferrin alone showed no significant effects with CC-IMT, we found significant interactions among females at specific ranges (<199 mg/dL: β= -0.49, 95% CI [-0.93, -0.07], p = 0.05; 199-231 mg/dL: β= -0.3, 95% CI [-0.56, -0.04], p = 0.052; 231-263 mg/dL: β= -0.99, 95% CI [-1.81, -0.17], p = 0.039). View this table: View inline View popup Table 4. Multivariable regression analysis of ferritin and CC-IMT We flexibly modelled the relationship between ferritin and CC-IMT using spline regression (degree of freedom=4). In addition, we tested the associations between ferritin and CC-IMT for differences due to sex-by using an interaction term (ferritin*sex; Males as reference). All continuous variables were mean-centered for the analyses. BMI: Body Mass Index, HDLc: High-Density Lipoprotein, LDLc: Low-Density Lipoprotein, CRP: C-reactive protein, CC-IMT: Carotid Intima-Media Thickness, SBP: Systolic Blood Pressure, DBP: Diastolic Blood Pressure. View this table: View inline View popup Table 5. Multivariable regression analysis of transferrin and CC-IMT We flexibly modelled the relationship between transferrin and CC-IMT using spline regression (degree of freedom=4). In addition, we tested the associations between transferrin and CC-IMT for differences due to sex-by using an interaction term (transferrin *sex; Males as reference). All continuous variables were mean-centered for the analyses. BMI: Body Mass Index, HDLc: High-Density Lipoprotein, LDLc: Low-Density Lipoprotein, CRP: C-reactive protein, CC-IMT: Carotid Intima-Media Thickness, SBP: Systolic Blood Pressure, DBP: Diastolic Blood Pressure. The effect of sex alone was nonsignificant in both models. Other significant variables were age (β = 0.14, p < 0.001), diabetes (β = 0.02, p < 0.001), HDL-C (β = -0.006, p = 0.003), LDL-C (β = 0.008, p < 0.001), smoking (β = 0.016, p < 0.001 for former smokers; β = 0.019, p < 0.001 for current smokers), CRP (β = 0.01, p = 0.024), hemoglobin (β = 0.0098, p = 0.010), systolic blood pressure (β = 0.018, p < 0.001), and diastolic blood pressure (β = -0.0087, p = 0.0006); all of these significantly associated with CC-IMT ( Tables 4 - 5 ). 3.7.2 Children and Adolescents In the subgroup analysis, we found no significant associations between any of the iron biomarkers and CC-IMT among children and adolescents and no notable interactions with sex (Supplementary Tables 13-16). 3.8 UK Biobank Further, we analysed data from 42,299 UK Biobank participants for associations between HFE-genotypes and mean CC-IMT. Here, no statistically significant associations were detected in either males or females between CC-IMT in those with C282Y/H63D genotypes (including C282Y homozygotes) compared to those with no mutations ( Supplementary Table 17 ). The associations remained non-significant in a sensitivity analysis, restricted to participants who were undiagnosed with hemochromatosis (N=42,193). 4. Discussion To investigate the conflicting conclusions of the relationship between iron biomarkers and CC-IMT in previous studies, we conducted and report the first IPD meta-analysis. This meta-analysis includes diversely distributed studies (18 hospital-based and 3 population-based studies) and a study population comprising both adults (N=7,523) and children/adolescents (N=2,691). Although males were over-represented in the study population, the proportion of females in both age groups were sufficient to analyse sex-specific differences. The study participants showed different comorbidities e.g., diabetes, hypertension, and CKD. Therefore, this comprehensive dataset, the largest to date, allows robust exploration to test associations between iron parameters and CC-IMT across age groups and clinical settings. Our analysis approach is also novel since we used two complementary methods (ML and regression). Both, the ML and regression analyses identified iron biomarkers as important predictors of CC-IMT. Additionally, the regression analyses offered insights into sex-specific associations. Overall, the results demonstrate that elevated ferritin and reduced transferrin levels within the reference limits are significant predictors of CC-IMT, particularly in adult females. Our findings can be interpreted from an ‘iron perspective’: ferritin stores cellular iron and is a well-established marker of body iron stores ( 57 ); transferrin carries iron in the blood and supplies all cell types with the metal. The observed association between elevated ferritin levels and CC-IMT could be interpreted as supportive of the “iron hypothesis” in atherosclerosis, which suggests that elevated iron stores contribute to oxidative stress and vascular damage ( 58 , 59 ). This idea is further supported by our current finding that elevated transferrin levels (frequently observed in iron deficiency) are inversely associated with CC-IMT. A recent mendelian randomization study showing that an elevated iron status increases the risk of cardiovascular disease, specifically ischemic stroke provides additional support to this model ( 60 ). In this context, a genome-wide association study by Galesloot et al ( 61 ) showed that polymorphisms predicting higher hepcidin/ferritin ratios were associated with an increased atherosclerosis risk. Unfortunately, data on hepcidin and NTBI were not available in our study to evaluate their associations, and we suggest that future studies may consider including additional biomarkers such as NTBI and hepcidin in their analysis. As an extension of this idea, the logical argument would be that individuals with genetic iron overload could be at a higher risk of developing cardiovascular disease (CVD). We have previously shown that patients with genetic iron overload conditions (e.g., hemochromatosis or thalassemia major) show elevated markers of vascular dysfunction [Intercellular Adhesion Molecule 1 (ICAM-1), Vascular Adhesion Molecule 1 (VCAM-1)] that correlated positively with non-transferrin-bound-iron (NTBI) in these patients ( 1 , 2 ). Importantly, phlebotomy treatment of the hemochromatosis patients reverted the increased concentrations of ICAM-1 and VCAM-1 ( 1 ). To follow up on these data, we also investigated the UK Biobank for associations between mean CC-IMT and the hemochromatosis genotype (C282Y/H63D genotypes) using MR, but we did not detect significant associations. MR offers an important complement to observational analyses by reducing bias from confounding and reverse causation. The absence of an MR association contrasts with our observational findings, which suggested a potential non-linear relationship between ferritin, transferrin, and CC-IMT. These data may be explained by population studies that indicate that the HFE genotypes does not necessarily cause elevated body iron stores. For instance, the penetrance for elevated TSAT is approximately 100% for C282Y homozygotes and 37.5% for the corresponding heterozygotes ( 62 ). In addition, individuals with a homozygous C282Y HFE mutation may have received phlebotomy ( 62 ), which normalizes iron parameter and markers of vascular dysfunction. additionally, this discrepancy may indicate that the observational associations are driven, at least in part, by residual confounding, rather than a direct causal effect of iron status on arterial wall thickness. Inflammation could be one such confounder since iron metabolism and inflammation are tightly interconnected processes. Therefore, our findings can also be interpreted from an ‘inflammation perspective’. Inflammation is known to aggravate atherosclerosis ( 63 ). Ferritin is a well-known acute phase protein, which is induced in response to inflammation. Similarly, transferrin levels are reduced in inflammatory conditions ( 64 ). Although, we have excluded individuals with CRP>10 mg/dL in our analysis, we cannot completely exclude low-grade inflammation as a driver of increased CC-IMT. Thus, the opposing associations of ferritin and transferrin with CC-IMT could also be an epiphenomenon to inflammation. However, in inflammatory states, we would expect hypoferremia due to iron redistribution into reticuloendothelial macrophages ( 65 ). But, CC-IMT was not associated with reduced serum iron levels in our analysis, providing a counterargument for an inflammatory response of ferritin and transferrin. Biological sex is another factor that can influence iron metabolism and CC-IMT. It is well established that biological sex influences the risk for CVD ( 66 ). Although the overall prevalence of CVD is higher among males, several studies show a higher risk of mortality and morbidity in females due to CVD, particularly in the presence of common risk factors and comorbidities ( 67 , 68 ). Female-specific risk factors include hormones, pregnancy and reproductive health (e.g. menstruation, pregnancy-associated disorders etc.). While the influence of female sexual hormones on iron homeostasis is known ( 9 , 10 ), previous studies have not found consistent evidence linking factors such as parity, timing of menopause, duration of the reproductive period, use of hormone therapy or contraceptives with CC-IMT ( 69 , 70 ). We have accounted differences due to sex by using it as an interaction term and show that ferritin and transferrin show nonlinear associations specifically among females. Our recent work also detected a positive nonlinear association between ferritin and peripheral arterial disease in certain ferritin ranges specifically in females [48–97 ng/mL: OR 14.59, 95% CI 1.6–135.93, P = 0.019; 98–169 ng/mL: OR 171.07, 95% CI 1.27–23404, P = 0.039; ( 11 )]. Nevertheless, we were unable to adjust for effects of hormonal status or menstruation status in this study due to the unavailability of data. Further, whether these reproductive hormones also affect the production of other liver-expressed proteins, such as ferritin or transferrin is unclear. The use of medications is another confounder affecting the interpretation of our study. Medications affect CC-IMT progression and, in some cases, iron metabolism. For example, the use of statins has been associated with a significant reduction in CC-IMT progression ( 15 , 71 ). as well as lower ferritin levels ( 72 , 73 ). Additionally, other commonly used drugs such as metformin, glucagon like peptide-1 receptor agonists, dipeptidylpeptidase-4 inhibitors, phosphodiesterase III inhibitors, calcium channel blockers, and antiplatelet agents also attenuate CC-IMT progression ( 74 – 76 ), while the effects on iron metabolism remain incompletely understood. The datasets analyzed here lacked detailed information on medication, representing a limitation of our study. The strengths of our study include the comprehensive analysis of iron biomarkers with CC-IMT, thereby providing a rather complete picture of the role of systemic iron status. For the analysis, we applied both ML and regression in our approach. A key advantage of the ML approach is its ability to capture complex, nonlinear interactions between predictors. On the other hand, regression results are more interpretable due to their straightforward effect sizes and significance values and thus providing hypothesis-driven insights. Despite its advantages, the ML approach has limitations, including its reliance on large datasets for optimal performance and the potential for overfitting, particularly if not properly tuned. Additionally, ML models can be seen as “black boxes”, making it challenging to derive explicit causal inferences or explainability in clinical decision-making. Conversely, while regression models provide interpretable and statistically robust associations, to some extent, they assume linearity. Furthermore, regression models are susceptible to multicollinearity and require prespecified assumptions (such as defining covariates), which can limit their flexibility in identifying novel predictive patterns. In our current study, we identified variables of importance from ML and included them as covariates for the regression analyses and therefore, have integrated the two approaches. This complementary use of both methods strengthens the validity of our findings and provides a comprehensive perspective on cardiovascular risk assessment. We also recognize that the integration of ML and conventional statistics is still at an early stage; nevertheless, exploring this interface is meaningful, especially for hypothesis generation. An additional advantage of this meta-analysis is that it has achieved a good participant mix with studies from different geographical locations, a range of age groups (adults and children and adolescents), and common comorbidities such as CKD, diabetes, and hypertension. Thus, we believe that the findings are generalizable across healthy individuals, disease states and in different age groups. The following limitations should however be considered when interpreting our findings. First, data retrieval for the IPD was incomplete, as we received only 20.4% of the published data despite our best efforts. We also had a significant proportion of missing data, both systematically across studies and sporadically within each study. Furthermore, it is essential to recognize that the data availability for these variables was not uniform across all studies; for example, we have not adjusted the models for the different treatments (e.g., medications for diabetes, lipid-lowering, dialysis for CKD, phlebotomy for hemochromatosis etc.) which could affect the CC-IMT outcome. It is worth noting that the data on additional iron biomarkers (hepcidin, NTBI) or menstrual and hormonal status were not available. In addition, the study population also does not represent all ethnicities, so we have not estimated effects due to these. While the inclusion of diverse cohorts enhances the generalizability of our findings, the heterogeneity in study designs and measurement protocols should be considered when interpreting the results. Despite the use of study-level random effects to account for between-study variability, residual confounding from unmeasured factors—such as differences in treatment regimens, comorbidities, and subtle variations in CC-IMT imaging protocols—cannot be excluded and may partly account for the observed associations. The interpretation of iron metabolism biomarkers, particularly ferritin, is complicated by significant variability in assay standardization and traceability ( 77 ), representing a limitation of our study. This variability hinders the establishment of universal reference intervals or thresholds and complicates their clinical interpretation. Therefore, the thresholds identified in our study should be considered exploratory and are not intended to serve as cut-offs for decision making. Additionally, our study is cross-sectional and does not assess the temporal relationship between the parameters; therefore, we consider that this study is for hypothesis generation and not for implying causality. Finally, since the study relied on published data, an element of publication bias cannot be excluded, as studies with non-significant results may remain unpublished and therefore, undiscovered. 5. Conclusion Our observational results demonstrate that iron biomarkers (specifically ferritin and transferrin) are non-linearly associated with CC-IMT, specifically in females. However, a significant causal association between HFE genotypes and CC-IMT in the UK Biobank data were not detected. This discrepancy may indicate that the observational associations are driven, at least in part, by residual confounding factors (such as inflammation, medications), rather than a direct causal effect of iron status on arterial wall thickness. Future studies may want to consider these confounding factors alongside iron status indicators to better disentangle their specific contributions to atherosclerosis. We consider our findings exploratory to drive further research addressing the underlying mechanisms that could explain these associations. 6. Ethics statement All primary studies were approved by the local Ethics committees. All participants were included in the study according to the guidelines of the local ethics committees following written informed consent to participate. The North West Multi-Centre Research Ethics Committee (Research Ethics Committee reference 11/NW/0382) approved UK Biobank, and all participants provided their written informed consent at baseline. All research was conducted in accordance with both the Declarations of Helsinki and Istanbul. 7. Funding No financial support was received for this meta-analysis. MUM acknowledges funding from the Deutsche Forschungsgemeinschaft (DFG) (SFB1118; GRK 2727; FerrOs - FOR5146; Ferroptosis SPP2306: Project No.461704553) and the Federal Ministry of Education and Research (NephrESA Nr 031L0191C; Translational Lung Research Center Heidelberg (TLRC-H), German Center for Lung Research (DZL); Project No. FKZ 82DZL004A1). JLA is funded by a National Institute for Health and Care Research Advanced Fellowship (NIHR301844). 8. Author contributions Anand Ruban Agarvas: Conceptualization, Methodology, Software, Formal analysis, Investigation, Data curation, Writing-Original draft, Visualization, Project administration Richard Sparla: Investigation, Writing - Review Janice L. Atkins (for UK Biobank): Formal analysis, Investigation, Writing-Original draft José Manuel Fernández-Real, José María Moreno-Navarrete, Abel López-Bermejo, Judit Bassols, Krystyna Pawlak, Dariusz Pawlak, Claudia Altamura, H M Dvořáková, Ebru Asicioglu, Jovana Kusic Milicevic, Michael Knoflach, Christoph Hochmayr, Petr Syrovatka, Peter Riško, Silvia Lai, Todd Anderson: Data curation, Investigation, Writing - Review Dorota Formanowicz, Jose M Valdivielso, Pavel Kraml, Luca Valenti: Data curation, Investigation, Writing - Review & Editing Martina U. Muckenthaler: Supervision, Writing - Review & Editing, Project administration 9. Conflict of interest None to declare. 10. Data and Code Availability Statement Part of the data (subject to data sharing restrictions) may be available from authors upon reasonable request. Codes used for data analysis and visualization are accessible here: https://github.com/griffindoc/imt Supplementary files Supplementary File 1. Study protocol as published in PROSPERO database Supplementary File 2. PRISMA-IPD checklist Supplementary Figure 1. Scheme used for data requests in the study Supplementary Figure 2. Rain cloud plots of serum iron categorised by age group and sex Supplementary Figure 3. Rain cloud plots of serum ferritin categorised by age group and sex Supplementary Figure 4. Rain cloud plots of serum transferrin categorised by age group and sex Supplementary Figure 5. Rain cloud plots of serum TSAT categorised by age group and sex Supplementary Figure 6. Rain cloud plots of CC-IMT categorised by age group and sex Supplementary Figure 7. Figure showing the pattern of co-occurrences and instances of missing values in the dataset View this table: View inline View popup Supplementary Table 1. List of R packages used in the analysis View this table: View inline View popup Supplementary Table 2. Variables included in IPD-MA Table of the variables, included in the IPD with a breakdown of the number of data points, number of studies that contained a variable and the sources. View this table: View inline View popup Supplementary Table 3. Table of the study-wise breakdown of participant characteristics View this table: View inline View popup Supplementary Table 4. Methodological differences in CC-IMT measurement View this table: View inline View popup Supplementary Table 5. Table of a breakdown of the proportion of missing values for each variable in the pooled data View this table: View inline View popup Supplementary Table 6. Table of a studywise breakdown of the proportion of missing values for each variable View this table: View inline View popup Supplementary Table 7. Table of a breakdown of the proportion of missing values for each variable categorised by age group View this table: View inline View popup Supplementary Table 8. Comparison of baseline characteristics between the Training and Test subsets used for Machine Learning View this table: View inline View popup Supplementary Table 9. Regression output Ferritin vs CC-IMT in adults We fitted a linear mixed model in (estimated using REML and nloptwrap optimizer) to predict CC-IMT with ferritin and sex (formula: imt ∼ 1 + ns(ferritin, df = 4) * sex). The model included study as random effect (formula: ∼1 | study). Complete (unimputed) data was used. CI = Confidence Interval View this table: View inline View popup Supplementary Table 10. Regression output Transferrin vs CC-IMT in adults We fitted a linear mixed model (estimated using REML and nloptwrap optimizer) to predict CC-IMT with transferrin and sex (formula: imt ∼ 1 + ns(transferrin, df = 4) * sex). The model included study as random effect (formula: ∼1 | study). Complete (unimputed) data was used. CI = Confidence Interval View this table: View inline View popup Supplementary Table 11. Regression output TSAT vs CC-IMT in adults We fitted a linear mixed model (estimated using REML and nloptwrap optimizer) to predict CC-IMT with TSAT and sex (formula: imt ∼ 1 + ns(TSAT, df = 4) * sex). The model included study as random effect (formula: ∼1 | study). Complete (unimputed) data was used. CI = Confidence Interval View this table: View inline View popup Supplementary Table 12. Regression output Iron vs CC-IMT in adults We fitted a linear mixed model (estimated using REML and nloptwrap optimizer) to predict CC-IMT with iron and sex (formula: imt ∼ 1 + ns(iron, df = 4) * sex). The model included study as random effect (formula: ∼1 | study). Complete (unimputed) data was used. CI = Confidence Interval View this table: View inline View popup Supplementary Table 13. Regression output Ferritin vs CC-IMT in children and adolescents We fitted a linear model (estimated using maximum likelihood) to predict CC-IMT with ferritin and sex (formula: imt ∼ 1 + ns(ferritin, df = 4) * sex). Complete (unimputed) data was used. CI = Confidence Interval View this table: View inline View popup Supplementary Table 14. Regression output TSAT vs CC-IMT in children and adolescents We fitted a linear model (estimated using maximum likelihood) to predict CC-IMT with TSAT and sex (formula: imt ∼ 1 + ns(TSAT, df = 4) * sex). Complete (unimputed) data was used. CI = Confidence Interval View this table: View inline View popup Supplementary Table 15. Regression output Iron vs CC-IMT in children and adolescents We fitted a linear model (estimated using maximum likelihood) to predict CC-IMT with iron and sex (formula: imt ∼ 1 + ns(iron, df = 4) * sex). Complete (unimputed) data was used. CI = Confidence Interval View this table: View inline View popup Supplementary Table 16. Regression output Transferrin vs CC-IMT in children and adolescents We fitted a linear model (estimated using maximum likelihood) to predict CC-IMT with transferrin and sex (formula: imt ∼ 1 + ns(transferrin, df = 4) * sex). Complete (unimputed) data was used. CI = Confidence Interval View this table: View inline View popup Supplementary Table 17. Associations between hemochromatosis HFE-genotypes and mean CC-IMT in the UK Biobank data Restricted to UK Biobank participants of European genetic ancentry. Models stratified by sex and adjusted for age, assessment centre and first ten genetic principle components. Mean of CC-IMT calculated from different measurements at 120/150/210/240 degrees. Supplementary file 2. PRISMA-IPD checklist Acknowledgement We thank the participants of all included studies. We also acknowledge the contributions of investigators and their efforts without which this data collection would not have been possible (Dr. Stefan Kopf, Dr. Rashid Merchant, Dr. Amina Abdel-Salam, Dr. Peter Risko, Dr. Tessel E. Galesloot, Dr. Dorine Swinkles, Dr. George Hahalis). We acknowledge Dr. Tiago JS Lopes’s inputs on the development of the machine learning model and Svenja Elizabeth Seide’s contributions to data preprocessing and imputation. Part of this research was conducted using the UK Biobank resource, under application 14631. We thank the UK Biobank participants and coordinators. This work used data provided by patients and collected by the NHS as part of their care and support. Copyright © (2023), NHS England. Re-used with the permission of the NHS England and UK Biobank. All rights reserved. This research also used data assets made available by National Safe Haven as part of the Data and Connectivity National Core Study, led by Health Data Research UK in partnership with the Office for National Statistics and funded by UK Research and Innovation. This study was supported by the National Institute for Health and Care Research Exeter Biomedical Research Centre. The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care. Footnotes Text has been revised to clarify the rationale in the introduction. Discussion has been rewritten to account for additional confounding factors. References updated and revised. References 1. ↵ Vinchi F , et al. Atherosclerosis is aggravated by iron overload and ameliorated by dietary and pharmacological iron restriction . Eur Heart J. [published online ahead of print: March 20 , 2019 ]. doi: 10.1093/eurheartj/ehz112 . OpenUrl CrossRef 2. ↵ Vinchi F , et al. Vasculo-toxic and pro-inflammatory action of unbound haemoglobin, haem and iron in transfusion-dependent patients with haemolytic anaemias . Br J Haematol . 2021 ; 193 ( 3 ): 637 – 658 . OpenUrl CrossRef PubMed 3. ↵ Formanowicz D , et al. The role of Fenton reaction in ROS-induced toxicity underlying atherosclerosis – modeled and analyzed using a Petri net-based approach . Biosystems . 2018 ; 165 : 71 – 87 . OpenUrl PubMed 4. ↵ van den Oord SC , et al. Carotid intima-media thickness for cardiovascular risk assessment: systematic review and meta-analysis . Atherosclerosis . 2013 ; 228 ( 1 ): 1 – 11 . OpenUrl CrossRef PubMed Web of Science 5. ↵ Polak JF , O’Leary DH . Carotid Intima-Media Thickness as Surrogate for and Predictor of CVD . Glob Heart . 2016 ; 11 ( 3 ): 295 – 312 e3. OpenUrl PubMed 6. ↵ Gerdts E , Regitz-Zagrosek V . Sex differences in cardiometabolic disorders . Nat Med . 2019 ; 25 ( 11 ): 1657 – 1666 . OpenUrl CrossRef PubMed 7. ↵ Pasricha S-R , et al. Iron deficiency . Lancet . 2021 ; 397 ( 10270 ): 233 – 248 . OpenUrl CrossRef PubMed 8. ↵ Galy B , Conrad M , Muckenthaler M . Mechanisms controlling cellular and systemic iron homeostasis . Nat Rev Mol Cell Biol. [published online ahead of print: October 2 , 2023 ]. doi: 10.1038/s41580-023-00648-1 . OpenUrl CrossRef PubMed 9. ↵ Yang Q , et al. 17β-Estradiol Inhibits Iron Hormone Hepcidin Through an Estrogen Responsive Element Half-Site . Endocrinology . 2012 ; 153 ( 7 ): 3170 – 3178 . OpenUrl CrossRef PubMed Web of Science 10. ↵ Li X , et al. Progesterone receptor membrane component-1 regulates hepcidin biosynthesis . Journal of Clinical Investigation . 2015 ; 126 ( 1 ): 389 – 401 . OpenUrl CrossRef PubMed 11. ↵ Ruban Agarvas A , et al. Iron biomarkers predict peripheral artery disease in females . Atherosclerosis . 2025 ; 402 : 119111 . 12. ↵ Berenson GS , et al. Association between Multiple Cardiovascular Risk Factors and Atherosclerosis in Children and Young Adults . N Engl J Med . 1998 ; 338 ( 23 ): 1650 – 1656 . OpenUrl CrossRef PubMed Web of Science 13. ↵ Tierney JF , et al. Individual Participant Data (IPD) Meta-analyses of Randomised Controlled Trials: Guidance on Their Use . PLoS Med . 2015 ; 12 ( 7 ): e1001855 . OpenUrl CrossRef PubMed 14. ↵ Dhillon SK , et al. Theory and Practice of Integrating Machine Learning and Conventional Statistics in Medical Data Analysis . Diagnostics . 2022 ; 12 ( 10 ): 2526 . OpenUrl PubMed 15. ↵ Zhang M , et al. Integrating Machine Learning into Statistical Methods in Disease Risk Prediction Modeling: A Systematic Review . Health Data Sci . 2024 ; 4 : 0165 . OpenUrl PubMed 16. ↵ Stewart LA , et al. Preferred Reporting Items for Systematic Review and Meta-Analyses of individual participant data: the PRISMA-IPD Statement . JAMA . 2015 ; 313 ( 16 ): 1657 – 65 . OpenUrl CrossRef PubMed 17. ↵ R Core Team 2023. R Core Team ( 2023 ). 2023;R: A language and environment for statistical computing . R Foundation for Statistical Computing , Vienna, Austria . 18. ↵ Stein JH , et al. Use of Carotid Ultrasound to Identify Subclinical Vascular Disease and Evaluate Cardiovascular Disease Risk: A Consensus Statement from the American Society of Echocardiography Carotid Intima-Media Thickness Task Force Endorsed by the Society for Vascular Medicine . Journal of the American Society of Echocardiography . 2008 ; 21 ( 2 ): 93 – 111 . OpenUrl CrossRef PubMed Web of Science 19. ↵ Zoller H , et al. EASL Clinical Practice Guidelines on haemochromatosis . Journal of Hepatology . 2022 ; 77 ( 2 ): 479 – 502 . OpenUrl CrossRef PubMed 20. ↵ Yeh ETH , Willerson JT . Coming of Age of C-Reactive Protein: Using Inflammation Markers in Cardiology . Circulation . 2003 ; 107 ( 3 ): 370 – 371 . OpenUrl FREE Full Text 21. ↵ Chen T , Guestrin C . XGBoost: A Scalable Tree Boosting System . Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining . San Francisco California USA: ACM; 2016 : 785 – 794 . 22. ↵ Lundberg S , Lee S-I . A Unified Approach to Interpreting Model Predictions . [ published online ahead of print : 2017 ]. doi: 10.48550/ARXIV.1705.07874 . OpenUrl CrossRef 23. ↵ Cotter J , et al. How to Interact With Interactions: What Clinicians Should Know About Statistical Interactions . Hospital Pediatrics . 2023 ; 13 ( 10 ): e319 – e323 . OpenUrl PubMed 24. ↵ Yucel RM . Random-covariances and mixed-effects models for imputing multivariate multilevel continuous data . Stat Modelling . 2011 ; 11 ( 4 ): 351 – 370 . OpenUrl CrossRef PubMed Web of Science 25. ↵ Schafer JL , Yucel RM . Computational Strategies for Multivariate Linear Mixed-Effects Models With Missing Values . Journal of Computational and Graphical Statistics . 2002 ; 11 ( 2 ): 437 – 457 . OpenUrl 26. ↵ Carpenter RJ , Kenward MG . Mutlilevel imputation . Multiple Imputation and its Application . 2013 ; 203 – 228 . 27. ↵ Rubin DB . Multiple Imputation for Nonresponse in Surveys . 1987 . https://www.wiley.com/en-gb/Multiple+Imputation+for+Nonresponse+in+Surveys-p-9780471655749 . 28. ↵ Enders CK , Tofighi D . Centering predictor variables in cross-sectional multilevel models: A new look at an old issue . Psychological Methods . 2007 ; 12 ( 2 ): 121 – 138 . OpenUrl CrossRef PubMed Web of Science 29. ↵ Sudlow C , et al. UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age . PLoS Med . 2015 ; 12 ( 3 ): e1001779 . OpenUrl CrossRef PubMed 30. ↵ Fry A , et al. Comparison of Sociodemographic and Health-Related Characteristics of UK Biobank Participants With Those of the General Population . American Journal of Epidemiology . 2017 ; 186 ( 9 ): 1026 – 1034 . OpenUrl CrossRef PubMed 31. ↵ Committee on the Use of Race, Ethnicity, and Ancestry as Population Descriptors in Genomics Research , et al. Using Population Descriptors in Genetics and Genomics Research: A New Framework for an Evolving Field . Washington, D.C. : National Academies Press ; 2023 . 32. ↵ Casanova F , et al. Iron and risk of dementia: Mendelian randomisation analysis in UK Biobank . J Med Genet . 2024 ;jmg-2023-109295. 33. ↵ Van Hout CV , et al. Exome sequencing and characterization of 49,960 individuals in the UK Biobank . Nature . 2020 ; 586 ( 7831 ): 749 – 756 . OpenUrl CrossRef PubMed 34. ↵ UK Biobank . Carotid ultrasound measures [Internet] . https://biobank.ndph.ox.ac.uk/ukb/label.cgi?id=101 . 35. ↵ Pawlak K , Pawlak D , Mysliwiec M . Long-term erythropoietin therapy decreases CC-chemokine levels and intima-media thickness in hemodialyzed patients . Am J Nephrol . 2006 ; 26 ( 5 ): 497 – 502 . OpenUrl CrossRef PubMed 36. ↵ Hahalis G , et al. Global vasomotor dysfunction and accelerated vascular aging in beta-thalassemia major . Atherosclerosis . 2008 ; 198 ( 2 ): 448 – 57 . OpenUrl CrossRef PubMed 37. ↵ Anderson TJ , et al. Microvascular function predicts cardiovascular events in primary prevention: long-term results from the Firefighters and Their Endothelium (FATE) study . Circulation . 2011 ; 123 ( 2 ): 163 – 9 . OpenUrl Abstract / FREE Full Text 38. ↵ Altamura C , et al. Ceruloplasmin/Transferrin system is related to clinical status in acute stroke . Stroke . 2009 ; 40 ( 4 ): 1282 – 8 . OpenUrl Abstract / FREE Full Text 39. ↵ Risko P , et al. The labile iron pool in monocytes reflects the activity of the atherosclerotic process in men with chronic cardiovascular disease . Physiol Res . 2017 ; 66 ( 1 ): 49 – 61 . OpenUrl PubMed 40. ↵ Risko P , et al. Long-term donors versus non-donor men: Iron metabolism and the atherosclerotic process . Atherosclerosis . 2018 ; 272 : 14 – 20 . OpenUrl CrossRef PubMed 41. ↵ Syrovatka P , et al. Iron stores are associated with asymptomatic atherosclerosis in healthy men of primary prevention . Eur J Clin Invest . 2011 ; 41 ( 8 ): 846 – 53 . OpenUrl CrossRef PubMed 42. ↵ Valenti L , et al. Serum ferritin levels are associated with vascular damage in patients with nonalcoholic fatty liver disease . Nutr Metab Cardiovasc Dis . 2011 ; 21 ( 8 ): 568 – 75 . OpenUrl CrossRef PubMed 43. ↵ Prats-Puig A , et al. Serum Ferritin Relates to Carotid Intima-Media Thickness in Offspring of Fathers With Higher Serum Ferritin Levels . Arterioscler Thromb Vasc Biol . 2016 ; 36 ( 1 ): 174 – 80 . OpenUrl Abstract / FREE Full Text 44. ↵ Merchant RH , et al. Evaluation of carotid artery dynamics & correlation with cardiac & hepatic iron in beta-thalassaemia patients . Indian J Med Res . 2016 ; 143 ( 4 ): 443 – 8 . OpenUrl PubMed 45. ↵ Dvorakova HM , et al. Determinants of premature atherosclerosis in children with end-stage renal disease . Physiol Res . 2012 ; 61 ( 1 ): 53 – 61 . OpenUrl PubMed 46. ↵ Asicioglu E , et al. Fibroblast growth factor-23 levels are associated with uric acid but not carotid intima media thickness in renal transplant recipients . Transplant Proc . 2014 ; 46 ( 1 ): 180 – 3 . OpenUrl PubMed 47. ↵ Galesloot TE , et al. Serum hepcidin is associated with presence of plaque in postmenopausal women of a general population . Arterioscler Thromb Vasc Biol . 2014 ; 34 ( 2 ): 446 – 56 . OpenUrl Abstract / FREE Full Text 48. ↵ Arroyo D , et al. Observational multicenter study to evaluate the prevalence and prognosis of subclinical atheromatosis in a Spanish chronic kidney disease cohort: baseline data from the NEFRONA study . BMC Nephrol . 2014 ; 15 ( 1 ): 168 . OpenUrl CrossRef PubMed 49. ↵ Formanowicz D , et al. Usefulness of serum interleukin-18 in predicting cardiovascular mortality in patients with chronic kidney disease–systems and clinical approach . Sci Rep . 2015 ; 5 : 18332 . 50. Lai S , et al. Early markers of cardiovascular risk in chronic kidney disease . Ren Fail . 2015 ; 37 ( 2 ): 254 – 61 . OpenUrl PubMed 51. ↵ Abaza SE , et al. Carotid Doppler ultrasonography as a screening tool of early atherosclerotic changes in children and young adults with beta-thalassemia major . J Ultrasound . 2017 ; 20 ( 4 ): 301 – 308 . OpenUrl PubMed 52. ↵ Kusic Milicevic J , et al. Cardiovascular risk assessment and coronary artery calcification burden in asymptomatic patients in the initial years of hemodialysis . Ther Apher Dial . 2022 ; 26 ( 1 ): 64 – 70 . OpenUrl 53. ↵ Bernar B , et al. The Tyrolean early vascular ageing-study (EVA-Tyrol): study protocol for a non-randomized controlled trialL: Effect of a cardiovascular health promotion program in youth, a prospective cohort study . BMC Cardiovasc Disord . 2020 ; 20 ( 1 ): 59 . OpenUrl PubMed 54. ↵ Hanafy AS , et al. Efficacy of a non-invasive model in predicting the cardiovascular morbidity and histological severity in non-alcoholic fatty liver disease . Diabetes Metab Syndr . 2019 ; 13 ( 3 ): 2272 – 2278 . OpenUrl PubMed 55. ↵ Nassef S , et al. Assessment of Atherosclerosis in Peripheral and Central Circulation in Adult beta Thalassemia Intermedia Patients by Color Doppler Ultrasound: Egyptian Experience . J Vasc Res . 2020 ; 57 ( 4 ): 206 – 212 . OpenUrl PubMed 56. ↵ Fernandez-Real JM , et al. Iron and obesity status-associated insulin resistance influence circulating fibroblast-growth factor-23 concentrations . PLoS One . 2013 ; 8 ( 3 ): e58961 . OpenUrl CrossRef PubMed 57. ↵ DePalma RG , et al. Cytokine signatures in atherosclerotic claudicants . Journal of Surgical Research . 2003 ; 111 ( 2 ): 215 – 221 . OpenUrl CrossRef PubMed Web of Science 58. ↵ Sullivan Jerome L . IRON AND THE SEX DIFFERENCE IN HEART DISEASE RISK . The Lancet . 1981 ; 317 ( 8233 ): 1293 – 1294 . OpenUrl 59. ↵ Sullivan JL . Iron in arterial plaque: modifiable risk factor for atherosclerosis . Biochim Biophys Acta . 2009 ; 1790 ( 7 ): 718 – 723 . OpenUrl CrossRef PubMed 60. ↵ Barad A , et al. Associations Between Genetically Predicted Iron Status and Cardiovascular Disease Risk: A Mendelian Randomization Study . JAHA . 2024 ; 13 ( 11 ): e034991 . OpenUrl CrossRef PubMed 61. ↵ Galesloot TE , et al. Iron and hepcidin as risk factors in atherosclerosis: what do the genes say? BMC Genet . 2015 ; 16 ( 1 ): 79 . OpenUrl PubMed 62. ↵ Gallego CJ , et al. Penetrance of Hemochromatosis in HFE Genotypes Resulting in p.Cys282Tyr and p.[Cys282Tyr];[His63Asp] in the eMERGE Network . The American Journal of Human Genetics . 2015 ; 97 ( 4 ): 512 – 520 . OpenUrl CrossRef PubMed 63. ↵ Baek JH , et al. Iron accelerates hemoglobin oxidation increasing mortality in vascular diseased guinea pigs following transfusion of stored blood . JCI Insight . 2017 ; 2 ( 9 ). doi: 10.1172/jci.insight.93577 . OpenUrl CrossRef 64. ↵ Fleck A . Clinical and nutritional aspects of changes in acute-phase proteins during inflammation . Proc Nutr Soc . 1989 ; 48 ( 3 ): 347 – 354 . OpenUrl CrossRef PubMed Web of Science 65. ↵ Guida C , et al. A novel inflammatory pathway mediating rapid hepcidin-independent hypoferremia . Blood . 2015 ; 125 ( 14 ): 2265 – 2275 . OpenUrl Abstract / FREE Full Text 66. ↵ Rajendran A , et al. Sex-specific differences in cardiovascular risk factors and implications for cardiovascular disease prevention in women . Atherosclerosis . 2023 ; 384 : 117269 . 67. ↵ Kannel WB , McGee DL. Diabetes and cardiovascular disease. The Framingham study . JAMA . 1979 ; 241 ( 19 ): 2035 – 2038 . OpenUrl CrossRef PubMed Web of Science 68. ↵ Huxley RR , Woodward M . Cigarette smoking as a risk factor for coronary heart disease in women compared with men: a systematic review and meta-analysis of prospective cohort studies . Lancet . 2011 ; 378 ( 9799 ): 1297 – 1305 . OpenUrl CrossRef PubMed Web of Science 69. ↵ Stöckl D , et al. Reproductive factors, intima media thickness and carotid plaques in a cross-sectional study of postmenopausal women enrolled in the population-based KORA F4 study . BMC Womens Health . 2014 ; 14 : 17 . 70. ↵ Miller VM , et al. Changes in carotid artery intima-media thickness 3 years after cessation of menopausal hormone therapy: follow-up from the Kronos Early Estrogen Prevention Study . Menopause . 2019 ; 26 ( 1 ): 24 – 31 . OpenUrl PubMed 71. ↵ Crouse JR , et al. Effect of rosuvastatin on progression of carotid intima-media thickness in low-risk individuals with subclinical atherosclerosis: the METEOR Trial . JAMA . 2007 ; 297 ( 12 ): 1344 – 1353 . OpenUrl CrossRef PubMed Web of Science 72. ↵ Zacharski LR , et al. The statin-iron nexus: anti-inflammatory intervention for arterial disease prevention . Am J Public Health . 2013 ; 103 ( 4 ): e105 – 112 . OpenUrl CrossRef PubMed Web of Science 73. ↵ Jamialahmadi T , et al. The Effects of Statin Treatment on Serum Ferritin Levels: A Systematic Review and Meta-Analysis . J Clin Med . 2022 ; 11 ( 17 ): 5251 . OpenUrl PubMed 74. ↵ Lv C , et al. Multi-faced neuroprotective effects of geniposide depending on the RAGE-mediated signaling in an Alzheimer mouse model . Neuropharmacology . 2015 ; 89 : 175 – 184 . OpenUrl PubMed 75. Kodama M , et al. Antiplatelet drugs attenuate progression of carotid intima-media thickness in subjects with type 2 diabetes . Thromb Res . 2000 ; 97 ( 4 ): 239 – 245 . OpenUrl CrossRef PubMed Web of Science 76. ↵ Chen Y , et al. The effect of metformin on carotid intima-media thickness (CIMT): A systematic review and meta-analysis of randomized clinical trials . Eur J Pharmacol . 2020 ; 886 : 173458 . 77. ↵ Swinkels DW , et al. Equivalence in clinical assessment of iron status requires ferritin assay standardisation before harmonisation of ferritin reference intervals . The Lancet Haematology . 2024 ; 11 ( 10 ): e721 . OpenUrl View the discussion thread. Back to top Previous Next Posted August 18, 2025. Download PDF Supplementary Material Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Ferritin and transferrin predict common carotid intima-media thickness in females: a machine-learning informed individual participant data meta-analysis Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Ferritin and transferrin predict common carotid intima-media thickness in females: a machine-learning informed individual participant data meta-analysis Anand Ruban Agarvas , Richard Sparla , Janice L Atkins , Claudia Altamura , Todd Anderson , Ebru Asicioglu , Judit Bassols , Abel López-Bermejo , Hana Marie Dvořáková , José Manuel Fernández-Real , Christoph Hochmayr , Michael Knoflach , Jovana Kusic Milicevic , Silvia Lai , José María Moreno-Navarrete , Dariusz Pawlak , Krystyna Pawlak , Petr Syrovatka , Dorota Formanowicz , Pavel Kraml , Jose M Valdivielso , Luca Valenti , Martina U. Muckenthaler medRxiv 2025.04.30.25326720; doi: https://doi.org/10.1101/2025.04.30.25326720 Share This Article: Copy Citation Tools Ferritin and transferrin predict common carotid intima-media thickness in females: a machine-learning informed individual participant data meta-analysis Anand Ruban Agarvas , Richard Sparla , Janice L Atkins , Claudia Altamura , Todd Anderson , Ebru Asicioglu , Judit Bassols , Abel López-Bermejo , Hana Marie Dvořáková , José Manuel Fernández-Real , Christoph Hochmayr , Michael Knoflach , Jovana Kusic Milicevic , Silvia Lai , José María Moreno-Navarrete , Dariusz Pawlak , Krystyna Pawlak , Petr Syrovatka , Dorota Formanowicz , Pavel Kraml , Jose M Valdivielso , Luca Valenti , Martina U. Muckenthaler medRxiv 2025.04.30.25326720; doi: https://doi.org/10.1101/2025.04.30.25326720 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Cardiovascular Medicine Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (300) Cardiovascular Medicine (4440) Dentistry and Oral Medicine (444) Dermatology (383) Emergency Medicine (608) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1510) Epidemiology (15229) Forensic Medicine (30) Gastroenterology (1126) Genetic and Genomic Medicine (6605) Geriatric Medicine (668) Health Economics (998) Health Informatics (4541) Health Policy (1369) Health Systems and Quality Improvement (1613) Hematology (543) HIV/AIDS (1265) Infectious Diseases (except HIV/AIDS) (15921) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (147) Nephrology (668) Neurology (6604) Nursing (346) Nutrition (998) Obstetrics and Gynecology (1145) Occupational and Environmental Health (957) Oncology (3334) Ophthalmology (974) Orthopedics (369) Otolaryngology (420) Pain Medicine (436) Palliative Medicine (130) Pathology (663) Pediatrics (1693) Pharmacology and Therapeutics (692) Primary Care Research (711) Psychiatry and Clinical Psychology (5448) Public and Global Health (9234) Radiology and Imaging (2199) Rehabilitation Medicine and Physical Therapy (1370) Respiratory Medicine (1196) Rheumatology (594) Sexual and Reproductive Health (712) Sports Medicine (530) Surgery (712) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a015bee5799cc13d',t:'MTc3OTcyNDc0OQ=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();