Expanded Studies on the SCAPIS Stockholm Reexamination Cohort

Area under the receiver operating characteristic curve (AUC-ROC) for prediction of CAD-RADS ≥2 using machine learning analysis of finger photoplethysmography. CAD-RADS 2.0 classification is based on coronary computed tomography angiography (CCTA), where stenosis severity is graded 0-5, atherosclerotic burden is quantified using segment involvement score (SIS) and calcification using Agatston Units (CAC-score).

Incidence of major adverse renal and cardiovascular events during follow-up. Associations will be evaluated with baseline immune mediators, lipid mediators, oral microbiota composition, vascular stiffness, steatohepatitis, and fibrosis measures.

Correlation between circulating biomarkers of immune activation and the degree of calcification and/or atherosclerosis at examined cardiovascular sites. Atherosclerosis and calcification are assessed by CCTA and/or ultrasound imaging and/or prior diagnosis of atherosclerotic cardiovascular disease. Immune profiling distinguishes resolving and non-resolving inflammatory phenotypes.

Correlation between circulating lipid mediator profiles and degree of calcification and/or atherosclerosis at examined cardiovascular sites. Lipid profiling characterizes inflammatory phenotypes and evaluates associations with prior diagnosis of atherosclerotic cardiovascular disease, calcification, and/or imaging-defined atherosclerosis assessed by CCTA and/or ultrasound.

Correlation between oral viral, fungal, and bacterial microbiota composition and degree of coronary artery disease as assessed by CCTA (CAD-RADS, SIS and CAC-score).

Prevalence of increased estimated hepatic lipid content defined as controlled attenuation parameter (CAP) >285 dB/m measured by transient elastography (FibroScan).

Prevalence of abdominal aortic aneurysm ≥ 30 mm as determined by abdominal ultrasound imaging.

Machine learning methods will be applied to identify the most informative predictors of CAD. Multimodal data, including ECG-derived features, lipid mediator measurements, inflammatory biomarkers, ultrasound variables, and oral microbiota profiles, will be used to develop and evaluate a prediction model for CAD, defined as CAD-RADS ≥2. Cross-validation will be used to reduce the risk of overfitting. Model performance will be evaluated using measures of discrimination and calibration. Clinical decision-curve analysis will be used to assess the clinical utility of the model by estimating net benefit across relevant decision thresholds and comparing model-guided decisions with current clinical practice (the pooled cohort equation).

Improved area under receiver operating curve (ROC) to predict CAD-RADS ≥2 and SIS and CAC-score when adding PPG features to age and sex.

Improvement in AUC-ROC for predicting CAD-RADS category, segment involvement score (SIS), and coronary artery calcium (CAC) score when adding machine learning-derived PPG features to traditional cardiovascular risk factors including the Pooled Cohort Equation.

Association between circulating lipid mediator profiles including plasma remnant cholesterol and coronary artery disease severity assessed by CCTA using CAD-RADS, SIS and CAC-score.

Associations between immune mediators, lipid mediators, oral microbiota, and pulse wave velocity by Arteriograph and PPG features of arterial stiffness.

Prevalende of mild, -moderate and severe aortic and mitral stenosis and insufficiency as determined by cardiac echo in relation to valvular and aortic calcification.

Machine learning to predict and to identify top predictors for aortic stenosis, insufficiency and mitral stenosis of insufficiency

Agreement between valvular characterisation on CCTA and degree of aortic and mitral stenosis and insufficiency by cardiac ultrasond

Finger PPG features will be will be used to develop and evaluate a machine learning prediction model for aortic valvular disease (stenosis or insufficiency) detected on cardiac ultraound. Cross-validation will be used to reduce the risk of overfitting. Model performance will be evaluated using measures of discrimination (Receiver Operating Characteristic) and calibration.

The prevalence of thoracic aortic dilatation (>39 mm) will be assessed in the study population using coronary CT angiography (CCTA). Available multimodal data, including clinical characteristics, ECG-derived features, physiological measurements, laboratory biomarkers, and imaging variables, will be used to identify independent predictors of thoracic aortic dilatation and to develop prediction models. Model development will include cross-validation to reduce the risk of overfitting. Model performance will be evaluated using measures of discrimination (Receiver Operating Characteristics) and calibration.

Prevalence and associations between blood pressure, hypertension diagnosis and left ventricular mass index

Prevalence and associations between left ventricular and left atrial strain and baseline characteristics

Carotid femoral pulse wave velocity (PWV) in relation to left ventricular mass index (LVMI), as assessed by echocardiography.

Photoplethysmography (PPG)-derived features will be used to develop and evaluate prediction models for cardiac function, including left ventricular ejection fraction (LVEF) and degree of diastolic dysfunction (graded by echocardiography). Machine learning methods will be applied to identify the most informative predictors. Cross-validation will be used to reduce the risk of overfitting. Model performance will be evaluated using measures of (Receiver Operation Caracteristics) and calibration.

Prevalence of increased estimated hepatic stiffness defined as ≥8 kPa by liver elastography (Fibroscan)

Associations between baseline characteristics immune mediators, lipid mediators, oral microbiota, vascular stiffness, and prevalent hepatic disease

Prevalence of abdominal aortic subaneurysm ≥ 25 mm as determined by abdominal ultrasound

Prediction of Infrarenal aortic diameter as measured by ultrasound using all available subject data.

Association between changes in indexes in microbiome dysbiosis, increase in selected pathogen bacterial counts and relative abundances, and increase in specific inflammatory protein concentrations and lung health, as presented by symptoms captured by the questionnaires, lung function and lung imaging.

Association between changes in indexes in microbiome dysbiosis, increase in selected pathogen bacterial counts and relative abundances, and increase in specific inflammatory protein concentrations and prevalence and presentation of airway and lung parenchyma diseases

Association between changes in indexes in microbiome dysbiosis, increase in selected pathogen bacterial counts and relative abundances, and increase in specific inflammatory protein concentrations and prevalence of airway and lung parenchyma diseases and the presence of systemic inflammation, as measured by blood inflammatory proteins and urine mediators.

Association between changes in indexes in microbiome dysbiosis, increase in selected pathogen bacterial counts and relative abundances, and increase in specific inflammatory protein concentrations and prevalence of airway diseases and comorbidities, such as cardiovascular, inflammatory and immune-mediated systemic diseases.

Targeted plasma inflammatory protein panel quantified using Olink Target 48 Cytokine Panels (PEA technology) together with lipid profiling to predict CADRADS, SIS and CAC-score.

Gene expression profiling in peripheral blood mononuclear cells (PBMC) to characterize immune/inflammatory signatures associated with remnant cholesterol levels.

Additional valie of plasma levels of ApoB, ApoAI, ApoE, and ApoCIII quantified by commercially available ELISA kits to predict prevalent CAD as CADRADS, SIS- and CAC-score.

Prevalence of lipoprotein lipid composition (total cholesterol, free cholesterol, cholesteryl esters, triglycerides, phospholipids) and surface/core ratio metrics analyzed by size-exclusion chromatography.

Association between measures of erythrocyte function and coronary artery disease severity assessed by CCTA using CAD-RADS, SIS and CAC-score

Finger PPG features will be will be used to develop and evaluate a machine learning prediction model for mitral valv disease (stenosis or insufficiency) detected on cardiac ultraound. Cross-validation will be used to reduce the risk of overfitting. Model performance will be evaluated using measures of discrimination (Receiver Operating Characteristic) and calibration.