Area Under The Curve Research Articles

Background: Phenotypic Age (PhenoAge), which is calculated based on chronological age and nine biomarkers, has been shown to better predict in-hospital outcomes in patients with acute myocardial infarction (AMI). Research Question: This study aimed to determine whether PhenoAge predicts long-term cardiac outcomes in AMI patients more accurately than chronological age. Methods: In this retrospective study, we included 5,440 patients who underwent percutaneous coronary intervention (PCI) for AMI at a tertiary hospital between December 2009 and November 2018. Clinical outcomes included cardiac death (defined as death due to heart failure, myocardial infarction, or arrhythmia) and all-cause death. Results: The mean PhenoAge and chronological age of the entire patient cohort were 66.24 ± 11.77 years and 78.55 ± 19.07 years, respectively. During a follow-up period of 51.03 ± 36.97 months, there were 331 cases of cardiac death and 570 cases of all-cause death. The area under the curve (AUC) of the receiver operating characteristic (ROC) curve for predicting cardiac death was higher for PhenoAge (AUC: 0.746, p < 0.01) than for chronological age (AUC: 0.638, p < 0.01). Similarly, for all-cause death, PhenoAge (AUC: 0.719, p < 0.01) outperformed chronological age (AUC: 0.649, p < 0.01). However, both PhenoAge (AUC: 0.436, p < 0.01) and chronological age (AUC: 0.455, p < 0.01) showed low predictive power for revascularization. Patients with PhenoAge ≥80 had a higher risk of cardiac death than those with PhenoAge <80 [hazard ratio (HR) 5.12; 95% confidence interval (CI) 3.97–6.59; p < 0.01], and this risk was greater than that observed in patients with chronological age ≥80 compared to those <80 (HR 3.39; CI 2.66–4.34; p < 0.01). Conclusion: PhenoAge more accurately predicts long-term outcomes, including cardiac death, than chronological age in AMI patients who underwent PCI. The risk of cardiac death in patients with PhenoAge ≥80 is higher than in those with PhenoAge <80 and exceeds the corresponding risk difference observed using chronological age.

Introduction/Background: Genetic cardiomyopathy (GEN-CM) has overlapping features of inflammation and arrhythmias with cardiac sarcoidosis (CS) and non-genetic myocarditis (MYOC); however, the treatments are very different, making it critical to distinguish them accurately. Radiomics is an AI-based approach to pixel-based analysis for CMR images with the potential for broad implementation for this diagnostic problem using publicly available software. Research Questions/Hypothesis: We tested the hypothesis that CMR radiomics would have an area under the curve (AUC) of ≧ 0.8 for diagnosing GEN-CM relative to CS and MYOC. Methods/Approach: The study design was a secondary analysis of an observational cohort from an academic medical center. We evaluated contrast-enhanced CMR images from 145 patients in the University of Minnesota CMR Registry with either GEN-CM, CS, or MYOC. Radiomics was applied to SSFP cine end-diastolic images and late gadolinium enhancement (LGE) images. Principal component analysis (PCA) generated combinations of radiomics features ordered based on the percent of variance explained, and the most predictive PCA predictors for the two models (Model 1: GEN-CM v. CS; Model 2: GEN-CM v. MYOC) were chosen. Bootstrapping (resampling with replacement) was used to generate 100 different samples from the original dataset, and receiver operating characteristic (ROC) curves with 95% confidence bands were used to report the AUC and 95% CI for each model. Results/Data: Among 145 patients (age 43.2 ± 16.0 years, 31% female) who had GEN-CM (n=50), CS (n=47) or MYOC (n=48), the most prominent features influencing the diagnosis were the maximum mean absolute deviation in pixel intensity, the maximum run length non-uniformity (normalized), and the large area emphasis feature, which were all higher in GEN-CM v. CS and GEN-CM v. MYOC ( Figure 1 ). Four PCA predictors (three LGE predictors and one cine predictor) were identified for Model 1, and four PCA predictors (two LGE predictors and 2 cine predictors) were identified for Model 2. The AUC for Model 1 was 0.90 (95% CI 0.83 to 0.95), and the AUC for Model 2 was 0.88 (AUC 0.81 to 0.94), indicating a high level of performance ( Figure 2 ). Conclusions: Radiomics applied to CMR cine and LGE images provides a high level of performance for the diagnosis of genetic cardiomyopathy, accurately distinguishing it from cardiac sarcoidosis and non-genetic myocarditis with contributions from both CMR imaging sequences.

Background: Contrast-induced nephropathy (CIN) is an important cause of morbidity in patients with myocardial infarction (MI) undergoing percutaneous coronary intervention (PCI). Although the current guidelines support a multivariable risk assessment, the specific role of systemic immune-inflammation index (SII) remains unclear. This study aims to assess the diagnostic accuracy of the SII in predicting CIN among MI patients undergoing PCI. Methods: Embase, Scopus, and PubMed were systematically searched from inception to April 2025 to identify studies assessing the accuracy of SII in predicting CIN in MI patients (STEMI and NSTEMI) undergoing PCI. Mean differences (MD) were pooled using the inverse variance method under a random-effects model and presented with MD and 95% confidence intervals (CIs) using R Studio. Area under the curve (AUC) was pooled using the same approach and reported with 95% CIs using Review Manager v5.4.1. Diagnostic test accuracy meta-analysis was performed using Meta-DiSc v2 to present pooled sensitivity and specificity. Results: Six studies including 2,659 NSTEMI patients (CIN=393; no CIN=2,266) and 2,714 STEMI patients (CIN=251; no CIN=2,443) were included. Patients with CIN showed significantly higher SII in NSTEMI (MD: 593.38; 95% CI: 321.55, 865.20; p<0.01) and STEMI (MD: 768.12; 95% CI: 452.94, 1083.31; p<0.01). The AUC of SII was significant for NSTEMI (AUC = 0.80; 95% CI: 0.77, 0.84; p<0.00001) and STEMI (AUC = 0.73; 95% CI: 0.60, 0.85; p<0.00001). Diagnostic test accuracy meta-analysis identified SII as a reliable predictor for NSTEMI (sensitivity 77%, 95% CI: 69, 82; specificity 72%, 95% CI: 70, 74) and STEMI (sensitivity 78%, 95% CI: 72, 83; specificity 83%, 95% CI: 76, 88). Conclusion: The SII is a reliable biomarker for predicting CIN in NSTEMI and STEMI patients, showing significant diagnostic accuracy. Integrating SII into existing risk models can improve early risk stratification and guide preventive measures. Future research should validate optimal SII thresholds, explore its dynamic changes around contrast exposure, and evaluate combined use with other biomarkers to enhance personalized CIN risk prediction and management in line with current cardiology guidelines.

Background: Cardiovascular disease (CVD) is a global health concern. Traditional models often miss nonlinear dependencies among physiological and behavioral factors. Transformer-based deep learning can capture complex patterns in structured health data. We hypothesized that such a model, trained on large-scale check-up records, would improve long-term CVD risk prediction. Methods: Using annual health check-up data from Toyama Prefecture, Japan (n = 100,056; 2010–2024), we excluded individuals with baseline CVD. The outcome was time to incident CVD over 10 years, modeled as right-censored survival data. For external validation, we used data from Kanazawa City (n = 79,756). The Transformer model was trained using anthropometric, laboratory, and self-reported lifestyle data. Benchmark models included Cox regression, XGBoost survival embeddings, multilayer perceptron (MLP), the Framingham Risk Score (FRS), and the Hisayama Risk Score (HRS). Model performance was evaluated using C-index, time-dependent area under the curve (AUC), and precision-recall AUC (PR-AUC). Interpretability was assessed using SHapley Additive exPlanations (SHAP) and a Feature Attention Network (FAN), which visualizes directional relationships via Transformer attention weights. Attention was computed across all features, but only the top 12 ranked by SHAP were visualized to highlight key interactions. Results: There were 4,113 CVD events in the Toyama cohort. The Transformer achieved the best internal performance: C-index 0.796 (95% confidence interval [CI]: 0.790–0.802), 10-year AUC 0.821 (CI: 0.817–0.828), and PR-AUC 0.465 (CI: 0.456–0.475). In the Kanazawa cohort, performance remained strong (C-index 0.743; AUC 0.775; PR-AUC 0.504). SHAP identified age, electrocardiogram (ECG), antihypertensive medication, and sex as key predictors. FAN highlighted interpretable relationships—for example, weight gain shaped the model’s interpretation of age-related risk. Age was the most connected node in the attention network, linking behavioral and physiological features. Conclusion: The Transformer-based model outperformed conventional methods in both discrimination and calibration for long-term CVD risk prediction. Its consistent performance across distinct populations supports its utility in community-level risk stratification. By combining SHAP and FAN, the model reveals how modifiable behaviors influence physiological risk, supporting personalized prevention and public health strategies.

Introduction: Traditional risk models predicting mortality in cardiogenic shock (CS) rely on static data snapshots, failing to capture the temporal dynamics represented in electronic health records. Conventional machine learning (ML) models can incorporate some temporal information but often struggle to model the complex, non-linear, and long-range relationships present in clinical time series data. Foundation models, pretrained on large, diverse time series datasets, offer a powerful alternative by capturing richer temporal patterns and making predictions in dynamic clinical settings. Hypothesis: We hypothesized that a ML approach using a pretrained multi-task time series model (UniTS) could be fine-tuned to predict mortality in patients with CS admitted to the cardiac intensive care unit (CICU) using high-resolution, multivariable clinical time series data. Methods: We performed a retrospective analysis of patients with CS admitted to the Brigham and Women’s Hospital CICU (2015-2024). Patients were split into training (80%) and validation (20%) cohorts. For each observation, 24h of clinical data were used to fine-tune a pretrained UniTS model to develop: (1) a dynamic model generating rolling 24h mortality predictions every 6h, and (2) a static model predicting overall in-hospital mortality after the first 24h of CICU admission. Validation performance was evaluated using area under the curve (AUC). The static model was compared to the IABP-Shock II risk score and SCAI shock stages for the same task using the DeLong test. Results: Among 2,109 admissions with CS (median age 68 years, 38% women), 25% were AMI-related, and 39% received temporary mechanical circulatory support. In-hospital mortality was 37%, with a median time to death of 3.0 days following CICU admission. The final model included 31 clinical variables ( Fig A ). The dynamic model achieved a per-prediction AUC of 0.88 for 24-hour mortality. The static model achieved an AUC of 0.83 for in-hospital mortality, significantly outperforming the IABP-Shock II risk score (AUC 0.74; p = 0.03) and SCAI Shock stage (AUC 0.61; p < 0.001) ( Fig B ). Conclusion: Leveraging the full temporal and multivariable complexity of CICU time series clinical data, fine-tuned ML foundation models accurately predict both very early and in-hospital mortality. By substantially outperforming traditional risk stratification methods, this time series modeling approach offers a promising tool for dynamic risk assessment in CS.

Red cell distribution width (RDW) is recognised as a prognostic biomarker for predicting mortality in neonatal sepsis. However, its utility in forecasting 28-day mortality amongst very low birth weight (VLBW) neonates remains uncertain. We conducted a retrospective, observational, single-centre study involving neonates with a birth weight of less than 1500 g. The primary endpoint was 28-day mortality. RDW and the Score for Neonatal Acute Physiology-Perinatal Extension II (SNAPPE-II) were measured on Days 1 and 3 following admission to evaluate their predictive value for mortality. Receiver operating characteristic (ROC) curves and corresponding areas under the curve (AUC) were used to assess the association between RDW and 28-day mortality, and to compare the predictive performance of RDW with SNAPPE-II. RDW levels were significantly higher in the non-survivor group (n = 35) compared to the survivor group (n = 70) on both Day 1 and 3 post-admission (p < 0.01 for both). Additionally, RDW was positively correlated with SNAPPE-II scores (r = 0.63 and 0.61, p < 0.01). ROC analysis revealed that RDW measured on Day 3 had a strong predictive value for 28-day mortality (AUC = 0.89, p < 0.01), which was significantly greater than the AUC for RDW on Day 1 (AUC = 0.73). These findings suggest that dynamic RDW changes (on post-admission Day 1-3) are a reliable prognostic biomarker for 28-day mortality in VLBW neonates, with predictive performance comparable to SNAPPE-II. It can serve as a low-cost, easily accessible tool for risk stratification in resource-limited settings. China Clinical Trial Registration Center Registration No.: ChiCTR2100043217.

Clinical models for predicting massive intraoperative blood loss (IBL) in spinal metastasis surgery exhibit a systematic, vascularity-dependent bias, underestimating risk in non-hypervascular tumors while overestimating it in hypervascular ones. We aimed to develop and validate an AI model integrating MRI radiomics to reduce this bias and improve risk stratification. This retrospective study included 601 patients who underwent surgery for spinal metastases between January 2016 and December 2022. They were randomized to a development cohort (n=479) and a test cohort (n=122). Clinical characteristics and radiomic features from T1c MRI were used to develop predictive models. Based on internal validation across nine machine learning algorithms, the best-performing model was selected. External testing was performed using an independent cohort of 101 patients to assess generalizability. The primary outcome was defined as massive IBL, with an estimated blood loss of 2,500ml or more. Model performance was evaluated using the area under the curve (AUC), calibration curves, and decision curve analysis. An AI tool was developed to facilitate clinical use. Among the 702 patients included, the combined model integrating MRI radiomics and clinical variables outperformed the clinical model in both internal (AUC: 0.901 [95%CI: 0.8330-0.9690] vs. 0.735 [95%CI: 0.6238-0.8458]) and external validation cohorts (AUC: 0.885 [95%CI: 0.8052-0.9639] vs. 0.604 [95%CI: 0.4355-0.7720]). Subgroup analysis revealed that in non-hypervascular tumors, the combined model significantly increased the sensitivity for identifying massive bleeding (0.85 vs. 0.30, p<0.001). In hypervascular tumors, the specificity was notably enhanced (0.81 vs. 0.55, p<0.001), and meanwhile the false-positive rate was reduced. The use of AI tools also improved the prediction performance of spine surgeons. The model is freely accessible for download at https://github.com/banluqihao/A-predict-tool-for-spinal-metastases-surgery. By integrating MRI radiomics features, our model reduces the systemic biases of clinical-only models that depend on unreliable histological surrogates. This enables more accurate and individualized risk stratification, providing a reliable tool to guide preoperative planning and support more accurate risk stratification for patients with spinal metastases.

Background: The prognosis of hypertrophic cardiomyopathy (HCM) varies by several factors, including imaging characteristics and genotype, reflecting underlying pathology. Traditional machine learning models typically require large datasets, which are challenging to obtain in rare disease research. Bayesian Networks (BN) excel with limited data by incorporating domain knowledge and handling imaging uncertainty. Research Question: Can integrated radiogenomics improve the classification of genetic variants in HCM compared to models using clinical or radiomic features alone? Aim: To establish a novel BN that evaluates the ability of radiogenomics to improve disease classification by integrating imaging derived radiomic features with clinical genotyping data. Methods: We used a BN to analyze 41 HCM patients who had both clinical cardiovascular magnetic resonance (CMR, 3T, Siemens Health Systems, Germany) and genotyping (Ambry Genetics, Aliso Viejo, CA; Lapcorp Invitae, San Francisco, CA), identifying 20 Pathogenic/likely pathogenic (P/LP) and 21 VUS between 2018-2024. CMR included pre- and post-contrast T1 mapping and late gadolinium enhancement (LGE) sequences. Clinical and CMR variables included cardiac structure and function, body size, blood pressure, heart rate, age, and sex. Clinical variables and radiomic features from T1 and LGE were extracted, normalized, and reduced using mutual information and random forest feature importance. A BN with domain-informed priors trained using 5-fold cross-validation (CV) (30 repeats) classified P/LP vs. VUS, reporting sensitivity, specificity, accuracy and area under the curve (AUC) (Figure 1). Results: The radiomics model for classifying P/LP vs. VUS in HCM achieved a mean CV accuracy of 0.84 and an AUC of 0.82, outperforming the clinical model (0.82 accuracy, 0.74 AUC). The combined clinical+radiomics model improved further (0.88 accuracy, 0.88 AUC) (Figure 2), with a net reclassification index of 0.68, demonstrating the strength of integrated modeling (Figure 3). Conclusions: Integrated radiogenomics using BNs shows promise in classifying HCM variants with limited data, improving performance above clinical models. External validation will further refine these findings.

Introduction: Right ventricular–pulmonary arterial (RV–PA) coupling refers to the relationship between right ventricular contractility and its afterload. This relationship can be noninvasively estimated using tricuspid annular plane systolic excursion (TAPSE) to pulmonary arterial systolic pressure (PASP) ratio. A study by Yurditsy et al. demonstrated that in patients with acute pulmonary embolism (APE) undergoing mechanical thrombectomy, the TAPSE/PASP ratio was a strong predictor of normotensive shock. By using noninvasive echocardiographic measurements to assess hemodynamics, this approach may offer valuable insights for enhanced PE risk stratification. Research Question/Hypothesis: We predict that the TAPSE/PASP ratio will strongly predict shock in patients with APE. Methods: This retrospective analysis examined 83 patients with low-risk, intermediate-risk, or high-risk APE who had complete invasive hemodynamic results and echocardiograms at a large quaternary care academic medical center. The patients were stratified into shock (CI &lt;2.2) versus no-shock (CI &gt;2.2) groups based on right heart catheterization measurements. Univariate testing was performed between the two shock groups with Student’s t-test for continuous variables and chi-squared testing for categorical variables. Receiver operating characteristic curves were constructed for both the TAPSE/PASP ratio and TAPSE alone to predict shock—with AUCs and 95% confidence intervals estimated by DeLong’s method, optimal thresholds identified via the Youden index, and a DeLong test applied to compare the difference in AUCs. Results/Data: Invasive hemodynamics identified 43 (52%) of our patients with cardiogenic shock. The age, gender, and race were evenly distributed among the shock and no-shock groups (Table 1). Patients classified in the shock group had a significantly lower TAPSE/PASP ratio compared to the no-shock group (Table 2). The TAPSE/PASP ratio yielded an area under the curve (AUC) of 0.76, with a sensitivity of 0.79, specificity of 0.63, positive predictive value of 0.69, and negative predictive value of 0.74. Notably, the TAPSE/PASP ratio had a significantly higher AUC for predicting shock compared to TAPSE alone (p=0.043) (Figure 1). Conclusion: In conclusion, the TAPSE/PASP ratio can be a useful non-invasive measurement in predicting shock in patients presenting with APE. The measurement can be used to quickly assess hemodynamics to further risk-stratify patients with an APE.

Background: The revised 2018 AHA/ACC guidelines for adult congenital heart disease (ACHD) introduced the anatomic physiological (AP) classification to better categorize disease severity and prognosis in the ACHD population. Validation of the ACHD AP classification score (ACAP) to predict short-term and long-term postoperative morbidity and mortality in a US cohort has not been previously performed. Objective: We aimed to assess the accuracy of the ACHD AP classification in predicting short-term and long-term postoperative morbidity and mortality. Methods: This retrospective cohort study included 310 ACHD patients at a single academic institution between 2018 to 2022. Patients were identified by the STS congenital surgery registry and had undergone a congenital surgical procedure. The primary outcome was overall mortality. Secondary outcomes included short-term postoperative morbidity (stroke, arrythmia, bleeding, readmission, length of stay), long-term mortality, and comparison of the ACAP score to other existing surgical mortality risk scores. Kaplan-Meier and area under the curve (AUC) of Receiver Operating Characteristic curves were used to evaluate mortality. Logistic regression was used to compare short-term morbidity. Results: A total of 305 patients were included with a median age of 30 years (interquartile range 21-41 years) and 52% were female. There was a total of 16 deaths with 7 (2.3%) early postoperative deaths and 9 (3%) long-term deaths. By increasing anatomy complexity, overall mortality was 0%, 5% (n=11), and 4% (n=2), respectively. By increasing physiologic severity, overall mortality was 0%, 3% (n=2), 3% (n=6), and 15% (n=6). Moderate and complex anatomy trended towards increased mortality but were not statistically significant (p = 0.66, Figure 1). More severe physiology scores predicted increased mortality (p = 0.014, Figure 2). Higher physiologic or anatomic complexity scores were associated with longer postoperative length of stay. The ACAP AUC was 0.711 for mortality (Figure 3), which was comparable to other scores (PEACH AUC 0.575, ACHS AUC 0.798). Conclusions: The ACAP score revealed comparable predictive power to existing risk models. Worsening physiologic and anatomy scores were associated with worse postoperative outcomes. Further prospective studies are needed to validate the ACAP score as a prognostic factor for patients undergoing ACHD surgery.

Heart transplantation remains the ultimate treatment option for patients with end stage heart disease, but donor supply remains insufficient, thereby complicating its allocation. Traditional scoring systems have demonstrated limited accuracy in predicting transplant outcomes, implying the need for more advanced approaches. Machine learning (ML) and artificial intelligence (AI) models show promising potential in predicting post-transplant prognosis, rejection risk and mortality by analyzing complex multidimensional variables beyond the capacity of conventional models. This meta analysis evaluates the application and performance of ML algorithms in image analysis and outcome prediction in patients undergoing cardiac transplantation. This meta analysis was conducted in accordance with the “Preferred Reporting Items for Systematic Reviews and Meta-analysis” (PRISMA) guidelines. An extensive search was conducted in all the major medical databases for relevant articles concerning machine learning algorithms and its application in cardiac transplantation. This was followed by an in-depth review of the included papers for relevant characteristics and outcomes. The statistical analysis was performed in R-Studio. Pooled area under the curve (AUC) was assessed using Ruttergatsonis model and the heterogeneity was assessed using the I^2 test. This review included a total of 17 papers with 512504 patients (prognostic elements) and 10 AI algorithms. Statistical analysis indicated pooled area under the curve(AUC) of 0.77[0.68;0.87,95%CI,p=0.9999]. A maximum AUC of 0.89 was observed with the RF algorithm by Miller et al and a minimum of 0.64 with the ANN algorithm by Lisboa et al and Nilsson et al. The papers were reviewed for relevant qualitativeas well as quantitative data pertaining to the performance of AI models. The capacity of AI algorithms in the domain of predicting cardiac transplant outcomes has been statistically established. The machine learning algorithms show promising clinical applications and utility in further enhancing the effectiveness of cardiac transplantation.

Background: Coronary computed tomography angiography (CCTA) is vital for diagnosing ischemic heart disease, yet its accuracy is affected by varying reader expertise. Artificial Intelligence (AI)-driven automated stenosis assessment offers promise for enhancing diagnostic consistency. Aim: We aim to evaluate an AI-CCTA assessment against invasive coronary angiography, invasive FFR, and expert readings. Methods: We performed a comprehensive search in Web of Science, Scopus, PubMed, Cochrane Library, and EMBASE from inception until March 2025. Two independent reviewers screened articles and extracted data on study design, patient demographics, AI methodology, stenosis thresholds, and outcomes. For statistical analysis, we constructed summary receiver operating characteristic (SROC) curves and used a bivariate random-effects model to derive pooled sensitivity, specificity, diagnostic odds ratios (DOR), and area under the curve (AUC). Forest plots were generated to visualize these metrics. Results: Our meta-analysis included 34 studies with 10,067 patients. AI-based CCTA demonstrated excellent diagnostic performance with an AUC of 0.932 for per-patient analysis. The pooled per-patient sensitivity was 0.89 (95% CI: 0.87–0.91) and specificity was 0.80 (95% CI: 0.74–0.86), with a DOR of 37.07 (95% CI: 24.57–55.92). AI validated against expert readers achieved the highest accuracy (0.94, 95% CI: 0.87-0.98). The &gt;70% stenosis threshold demonstrated superior performance (accuracy: 0.89, specificity: 0.96) compared to the &gt;50% threshold (accuracy: 0.86, specificity: 0.87). Per-vessel analysis showed comparable results with an AUC of 0.905. Conclusion: Our meta-analysis confirms that AI-assisted coronary CT angiography delivers high diagnostic performance for coronary stenosis detection, with strong AUC values, high sensitivity and specificity, and robust diagnostic odds ratios across both per-patient and per-vessel assessments.

Background: Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia and a major cause of stroke and heart failure. Early identification of individuals at risk of AF is critical for preventive interventions. We aimed to develop and validate a novel metabolomics-based model to predict incident AF in a general population cohort. Methods: Our original cohort comprised 37,627 participants who underwent annual health checkups at St. Luke’s International Hospital, Tokyo, between 2015 and 2016. A case-cohort study was subsequently conducted with a selected sub-cohort. The participants were followed-up for three years. At health check-ups, anthropometric measurements and electrocardiography were performed and laboratory tests were ordered. Serum metabolomic profiling targeting 40 metabolites was performed using a GCMS-TQ8040 (Shimadzu Corporation) in multiple reaction monitoring mode. Least absolute shrinkage and selection operator (LASSO)-Cox regression with cross-validation was used for model development. Risk categories were defined based on hazard ratios (HRs) relative to the median hazard. Metabolite Set Enrichment Analysis (MSEA) was also performed. Results: The sub-cohort was 6,463 participants and 109 had incident AF. Those with AF were significantly older (66.8 vs. 52.8 years), had a higher body mass index (BMI) (24.8 vs. 22.5 kg/m2), and had elevated levels of other indicators of metabolic syndrome. The three-year incidence of AF ranged from 0.03% (lowest-risk, HR&lt;0.5) to 1.86% (highest-risk, HR&gt;4). Time-dependent receiver operating characteristic curve analysis of the model showed a high area under the curve (AUC) of 0.803 for incident AF within three years. The predictive performance of the model improved by incorporating BMI and age (AUC, 0.877). Additionally, the LASSO-based score provided good predictive performance in any age-group (HR for highest- vs. lowest-risk: 4.2 for &lt;50 years; 6.3 for 50–59 years; 7.3 for 60–69 years; and 5.1 for ≥70 years). MSEA revealed alterations in amino acid metabolism, suggesting metabolic remodeling in AF. Conclusion: We developed a metabolomics-based model that stratified individuals according to the risk of incident AF with good predictive performance. This model may assist in the early identification of high-risk individuals and support targeted preventive strategies against AF, even after age stratification. Further studies are required to validate the predictive ability in different populations.

Introduction: The American Heart Association’s PREVENT equation now includes zip-code level Social Deprivation Index (SDI), highlighting the growing role of socio-environmental (SE) factors in risk prediction. However, concerns remain about fully capturing the breadth of exposures, particularly those prevalent in disadvantaged populations, within existing tools. This study sought to develop and evaluate a novel SE Risk Score using machine learning, and to compare its utility against SDI in the context of Coronary Artery Calcium (CAC) screening. Research Questions: Whether a novel machine learning-derived SE Risk Score incorporating more than 150 area level environmental pollution, social and economic variables, when added to CAC scores, improves Major Adverse Cardiovascular Events (MACE) prediction than SDI, in Blacks. Methods: We analyzed CAC scores, MACE outcomes, demographics, and census-tract SE variables from the CLARIFY registry (NCT04075162), a large prospective study of no-charge CAC testing (84,233 White and 7,940 Black participants). A SE Score was derived using an XGBoost machine learning model. This SE Score was then compared to zip-code level SDI. Analyses included Cox proportional hazards models, mediation analysis, and model performance evaluation (Harrell’s C-index, area under the curve (AUC), calibration metrics, Net Reclassification Improvement [NRI]) for models including CAC alone, CAC+SE Risk Score, and CAC+SDI. Results: Compared to Whites, Black participants had higher MACE (14.0% vs 6.4%), despite lower mean CAC (151.5 vs 175.5). Adding SE Score to CAC improved C-index from 0.681 to 0.712, while adding census-tract SDI yielded 0.705 and zip-code SDI 0.700 respectively. For Blacks, the AUC at Year 4 improved from 0.642 (CAC alone) to 0.669 (CAC+SE Risk Score), surpassing the 0.672 achieved with CAC+census-tract SDI ( Figures 1 and 2 ). The improvement in risk reclassification was more pronounced for Black individuals (Net Reclassification Improvement: 0.153) than for White individuals (0.081). SE factors mediated 47.12% of the relationship between race and MACE. Conclusion: The machine learning–derived SE Score outperformed SDI in predicting MACE, improving both discrimination and calibration. SE factors mediated the race–MACE link, and their inclusion with CAC scores significantly enhanced risk reclassification, particularly in Black individuals. More refined tools are needed to better assess and address risk in socially disadvantaged populations.

Background: While Fontan palliation offers improved long-term survival for patients with single ventricle congenital heart disease, many eventually develop Fontan circulatory failure. The Advanced Cardiac Therapies Improving Outcomes Network (ACTION) published consensus-based screening criteria to guide timely referral for advanced heart failure care. However, the performance of these criteria in identifying high-risk patients remains unexplored. Objective: To evaluate the performance of the ACTION referral criteria in identifying Fontan patients at risk for death or heart transplantation. Methods: This is a retrospective, single-center, case-control study of patients after Fontan palliation between 2000 and 2023. Cases were defined as patients who died or required heart transplantation or VAD support. Controls were matched 2:1 or 3:1 based on the year of Fontan surgery[MOU1] . Clinical data were reviewed to assess whether patients met one or more of the ACTION screening criteria. Results: We identified 53 Fontan failure cases and matched them with 151 controls. All cases met at least one ACTION criterion, compared to only 35.8% of controls. Patients with adverse outcomes were significantly more likely to meet criteria across all major domains, including ventricular dysfunction, Fontan pathway abnormalities, lymphatic complications, and extracardiac organ dysfunction. A full performance of ACTION criterion for Fontan failure is presented in Table 1. The sensitivity and negative predictive value of the ACTION screening tool were 100%, with area under the curve (AUC) of 0.82, suggesting strong performance as a screening tool intended to identify patients at risk. For each of the 4 listed ACTION criteria subdomains, the AUCs were similarly good, except for the extracardiac domain, which had a much lower AUC of 0.57. Conclusion: The ACTION screening criteria demonstrated a good performance in identifying Fontan patients at risk for adverse outcomes. These findings provide the first objective evidence supporting the clinical utility of this expert consensus-based tool and highlight its potential value in guiding timely referral to advanced heart failure care. Prospective multi-center validation is warranted.

Introduction: MitraClip (MC) placement for transcatheter edge-to-edge repair (TEER) is being increasingly utilized for treatment of secondary mitral regurgitation (SMR). The aim of this study was to analyze one-year outcomes after MC placement as well as the utility of pre-procedural EROA/LVEDV (effective regurgitation orifice area/left ventricular end diastolic volume) ratio in predicting outcomes. Methods: We analyzed patients who underwent MitraClip placement at three centers. Primary efficacy endpoint (PEE) was determined by absence of heart failure (HF) hospitalizations at one year follow up. One way ANOVA was used to determine variance of pre-procedural EROA/LVEDV ratio between the outcome groups. Multivariate regression analysis was done to identify pre-procedural echocardiographic parameters that held significant association with symptom improvement. ROC analysis was done between the ratio and symptom improvement. A two tailed p-value &lt; 0.05 was used to determine statistical significance. Results: Between January of 2021 and March 2024, 236 patients underwent MitraClip placement, of which 157 patients (66.52%) reached PEE. Mean effective regurgitant orifice area (EROA) was 0.38 cm 2 (0.32 - 0.44) in the symptom worsening group, compared to 0.52 cm 2 (0.43 - 0.61) in patients that reached PEE. The mean left ventricular end diastolic volume (LVEDV) in the symptom worsening group was 110.58 mL (95.83 - 125.32) compared to 74.56 mL (62.55 - 86.57) in the patients that reached PEE. A ratio between EROA and LVEDV (EROA/ LVEDV x 100) was measured to quantify disproportionate secondary mitral regurgitation (SMR). After multivariate regression analysis, the ratio was found to have a significant association with symptom improvement (OR: 1.059 (1.032 –1.087, p value &lt; 0.001). Forest plots depicted in Figure 1 . ROC curve analysis had an area under the curve (AUC) of 0.7417 (0.669- 0.816) making it a better predictor for symptom improvement (Figure 2). Youden Index was calculated at 0.184 with 87% sensitivity and 51.4% specificity (Index: 0.39048). Conclusions: After regressing for confounding factors, a preprocedural EROA/ LVEDV ratio (EROA/LVEDV x 100) &gt; 0.184 had 87% sensitivity in predicting symptom improvement. The ratio had significant association with symptom improvement group after multivariate regression analysis.

Background: Dialysis patients with both acute coronary syndrome (ACS) and atrial fibrillation (AF) face high thrombotic and bleeding risks. The optimal antithrombotic strategy and long-term outcomes remain unclear. Methods: We analyzed data from the CRUISE-R, a nationwide, multicenter, retrospective cohort in China. Patients were grouped by AF status. The primary endpoint was a composite of all-cause mortality, non-fatal myocardial infarction, and stroke. Baseline features, antithrombotic strategies, and outcomes were compared. betwee groups. Results: Among 1,147 patients, 102 (8.9%) had AF. Compared to those without AF, AF patients were older (65.1 vs. 61.4 years, p &lt; 0.001), had higher proBNP (30,826 vs. 21,153 pg/ml, p = 0.015), and larger left atrial diameter (43.6 vs. 40.9 mm, p = 0.001). Triple therapy (OAC + aspirin + clopidogrel) was more common in AF patients (9.8% vs. 0.96%, p &lt; 0.001) (Figure 1). Over 22.9 months of follow-up, 511 patients (44.6%) reached the primary endpoint—450 (43.1%) without AF vs. 61 (59.8%) with AF. AF patients had higher composite event and cardiac mortality rates (p &lt; 0.05), but stroke and bleeding rates were not significantly different (p = 0.18 and 0.056, respectively) (Figure 2). Cox regression analysis identified AF as an independent predictor of the primary endpoint (HR 1.46, 95% CI: 1.07–1.98, p = 0.016) and cardiac death (HR 1.52, 95% CI: 1.02–2.27, p = 0.038). Triple antithrombotic therapy was independently associated with an increased risk of bleeding (HR 5.88, 95% CI: 1.41–24.52, p = 0.015). Among the 106 patients with both ACS and AF, mean CHA2DS2-VASc and HAS-BLED scores were 3.8 and 3.4 (Figure 3A). However,both scores demonstrated limited predictive value for non-fatal stroke and bleeding events in this population, with area under the curve (AUC) values below 0.65 (Figure 3B). Conclusion: In dialysis patients with ACS, the presence of atrial fibrillation is associated with a significantly worse prognosis, including higher primary composite endpoint and cardiac mortality, although the incidence of stroke is not significantly increased. Triple antithrombotic therapy markedly raises the risk of bleeding and should be used with caution. Standard risk scores such as CHA2DS2-VASc and HAS-BLED have limited predictive power in this population. Further research is needed to develop more effective risk stratification tools for identifying high-risk subgroups prone to thrombotic and bleeding events.

Background: Atrial fibrillation (AF) recurrence remains a common and clinically significant complication following CA. Identifying reliable biomarkers for stratifying recurrence risk is essential to guide post-ablation management and improve long-term outcomes. The systemic immune-inflammation index (SII), a composite marker derived from peripheral platelets, neutrophils, and lymphocytes, has emerged as a potential predictor of cardiovascular events. Methods: Embase, Medline (via PubMed), Scopus, and Web of Science were systematically searched from inception to May 2025 to identify studies evaluating the accuracy of the SII in predicting AF recurrence after CA. Pooled mean difference (MD) and area under the curve (AUC), along with 95% confidence intervals (CIs), were calculated using the inverse variance method under a random-effects model. Pooled sensitivity and specificity, with corresponding 95% CIs, were estimated using a bivariate random-effects model with Meta-DiSc V2. Results: Six studies comprising 543 patients with recurrence and 1,462 without recurrence were included. AF recurrence was significantly associated with higher SII levels (MD: 201.9; 95% CI: 79.34, 324.47; p=0.001). The pooled AUC for SII was 0.71 (95% CI: 0.64, 0.79; p&lt;0.00001), indicating moderate discriminative ability. Diagnostic accuracy analysis confirmed SII as a reliable predictor of recurrence, with a pooled sensitivity of 64% (95% CI: 0.50, 0.75), specificity of 71% (95% CI: 0.63, 0.77), positive likelihood ratio (PLR) of 2.09 (95% CI: 1.82, 2.40), and negative likelihood ratio (NLR) of 0.53 (95% CI: 0.41, 0.68). The 95% prediction ellipse area was 0.155. Conclusion: The SII demonstrates potential as a biomarker for identifying patients at elevated risk of AF recurrence following CA. In line with current trends emphasizing inflammation’s role in arrhythmogenesis, these findings support incorporating immune-inflammatory profiling into post-ablation risk assessment. Although not yet incorporated into existing clinical guidelines, the evidence underscores the need for its consideration in future guideline updates as part of a multi-parametric approach that integrates clinical, electrophysiological, imaging, and biomarker-based data to guide post-ablation management and improve patient outcomes.

Background: In patients with paradoxical low-gradient severe aortic stenosis (AS) based on aortic valve area (AVA) at rest (&lt;1cm2) and normal left ventricular ejection fraction (LVEF &gt;50%), the role of dobutamine stress echocardiography (DSE) is unclear. While current guidelines recommend assessing aortic valve calcium score in this context, DSE may still offer clinical value. Methods: We conducted a retrospective study on patients with paradoxical low gradient AS (LVEF &gt;50% AVA &lt;1cm 2 , MG &lt; 40 mmHg) at rest who underwent DSE after Institutional Review Board approval. Data was extracted using an internal database. Patients &gt; 18 yrs old with LVEF &gt;50%, mean gradient (MG) &lt; 40 mmHg, and valve area &lt; 1 cm 2 at rest were included. Kruskal-Wallis test was used to identify which resting variables were associated with severe AS (MG &gt;40 mm Hg and Valve area &lt;1 cm 2 ) with DSE. Results: A total of 42 patients were included in the study and variables were reported as means (±SD). The mean age of patients was 78 years (±10) and 60% were female. The mean resting LVEF was 60% (±5%), MG at rest was 28 mmHg (±7) while it was 40 mmHg (±14) with DSE, and mean aortic valve area at rest was 0.9 cm 2 (±0.09) while it was 1.1 cm 2 (±0.23) with DSE. Of the 42 patients, 11 (26%) were found to have severe AS during DSE, defined by a MG &gt;40 mmHg and AVA &lt;1.0 cm 2 . Baseline AVA and MG at rest were significantly associated with severe AS on DSE (p=0.008 and p=0.024, respectively), while age, stroke volume index, and baseline LVEF were not. Univariate regression analysis showed that the area under the curve (AUC) for predicting severe AS was 0.73 for resting MG and 0.77 for resting AVA. A resting AVA cutoff of 0.78 cm 2 provided optimal sensitivity (0.63) and specificity (0.90). For resting MG, a cutoff of 29 mmHg gave a sensitivity of 0.91 and specificity of 0.68 for identifying patients with severe AS. Conclusion: Our study indicated that DSE remains useful in identifying severe AS in substantial number of patients (26%) who exhibit paradoxical low gradient severe AS. Additionally, AVA at rest and baseline mean gradient were statistically the reliable parameters to determine those who may benefit significantly from undergoing DSE. This is one of the first studies to our knowledge in the literature suggesting that DSE may still be useful in identifying severe AS patients despite “paradoxical low gradient severe AS” at rest. Further research is warranted to validate our findings.

Background: Evaluating nocturnal regulation of cardiac autonomic nervous system (CANS) with frequency-domain heart rate variability (HRV) is relevant, especially in patients with sleep apnea. So far, the association between very low frequency (VLF) of HRV and major adverse cardiovascular and cerebrovascular events (MACCEs) has not been sufficiently elucidated. Aims: We seek to evaluate the impact of VLF of HRV on MACCEs in patients with hypertension. Methods: This prospective study enrolled 2,061 hypertensive patients from 2017–2021 monitored with nocturnal Holter electrocardiography and type III home sleep apnea testing. Nocturnal HRV was defined by the variation in normal-to-normal intervals at night, and evaluated based on frequency-domain spectra, including high frequency (HF) (0.15–0.5 Hz), low frequency (LF) (0.04–0.15 Hz), VLF (0.0033–0.04 Hz), and the LF/HF ratio. Restricted cubic spline analysis via Cox regression was used to explore the prognostic effects of different HRV indices. The cut-off point for VLF was set at 2,566 ms 2 . Stabilized inverse probability weighting (IPW) based on propensity scores was used to balance baseline characteristics between low-VLF (&lt;2566 ms 2 ) and high-VLF groups (≥2566 ms 2 ). Results: Over a median follow-up of 39 months, 1,867 patients aged 58.4±11.2 years (77.4% men) were included in final analysis. MACCEs occurred in 15.3%. Among 4 HRV indices, only VLF showed a stable nonlinear prognostic association. Compared with patients with high-VLF group, the low-VLF group had a 42–47% higher risk of MACCEs compared to the high-VLF group, depending on the cohort (unadjusted or IPW). In obese patients, those with low-VLF vs high-VLF showed an unadjusted HR of 1.55 for MACCEs (95% CI, 1.21–1.99; P =0.001), which persisted after IPW adjustment (HR 1.38 [95% CI, 1.06–1.81]; P =0.016). Similarly, patients with OSA in the low-VLF group (vs high-VLF group) had an increased risk of MACCEs (unadjusted HR 1.56, 95% CI 1.20–2.02, P =0.001; adjusted HR 1.38, 95% CI 1.07–1.77, P =0.013). The ratio of nocturnal-to-baseline VLF had an area under the curve (AUC) of 0.962 in predicting respiratory event-related cardiac cycle changes during short time intervals. Conclusion: Lower nocturnal VLF of HRV may increase the risk of MACCEs in patients with hypertension, particularly those with obesity and sleep apnea. suggesting a pathophysiological mechanism that links cardiovascular events to impaired regulation of cardiopulmonary coupling by the CANS.