Daily Cardiology Research Analysis
Analyzed 212 papers and selected 3 impactful papers.
Summary
Three impactful cardiology studies highlight how advanced analytics and imaging can sharpen decision-making. An AI-derived EuroTR score accurately predicts 1-year outcomes after tricuspid transcatheter edge-to-edge repair, a massive ML model anticipates aortic stenosis risk years ahead using routine data, and intravascular imaging–guided PCI shows outcomes comparable to CABG in diabetics with complex disease.
Research Themes
- AI/ML-driven risk stratification and prediction in cardiovascular care
- Imaging-guided revascularization strategies in complex coronary disease
- Translating predictive analytics into patient selection and outcomes
Selected Articles
1. Predicting Outcomes in Patients With Tricuspid Regurgitation Undergoing Transcatheter Edge-to-Edge Repair Using an Artificial Intelligence-Derived Risk Score: The EuroTR Risk Score.
Using extreme gradient boosting on 1,826 patients from a European registry, the EuroTR score accurately predicted 1-year mortality after T-TEER (C-index 0.741) and outperformed established tools (EuroScore, TRI-SCORE). Risk gradients were clinically meaningful and consistent across subgroups, and higher EuroTR scores correlated with death, HF hospitalization, or persistent NYHA III+ symptoms.
Impact: This is a rigorously validated AI risk tool tailored to T-TEER, addressing a major evidence gap in patient selection and benchmarking against widely used risk models.
Clinical Implications: EuroTR can inform Heart Team decisions, consent, and trial enrichment by identifying patients at high risk of poor outcomes after T-TEER, potentially guiding peri-procedural optimization and postprocedural follow-up intensity.
Key Findings
- Derivation (n=1,225) and validation (n=601) cohorts yielded a total of 1,826 patients.
- EuroTR achieved Harrell’s C-index 0.741 (95% CI: 0.699-0.783) in validation, outperforming EuroScore and TRI-SCORE.
- High vs low risk groups showed HR 4.26 (95% CI: 2.71-6.67; P<0.001) for 1-year mortality after T-TEER.
- Risk rank correlated with a composite of 1-year death, HF hospitalization, or persistent NYHA III+ (30.6% to 85.5% from lowest to highest risk ranks).
- Performance was consistent across atrial vs non-atrial TR, TRILUMINATE-eligible vs -noneligible, and with or without CIED leads.
Methodological Strengths
- Large, multicenter registry with independent external validation cohort
- Benchmarking against established risk scores and use of advanced ML (XGBoost)
Limitations
- Observational registry design with potential residual confounding and selection bias
- Black-box ML model with limited interpretability for feature contributions at the bedside
Future Directions: Prospective impact studies and randomized workflow integration trials should test whether EuroTR-guided selection improves outcomes, and explore model explainability and calibration across health systems.
BACKGROUND: Risk stratification for tricuspid valve transcatheter edge-to-edge repair (T-TEER) is paramount in the decision-making process to appropriately select patients with severe tricuspid regurgitation. OBJECTIVES: The aim of this study was to develop and validate an artificial intelligence-driven risk score, the EuroTR (European Registry of Transcatheter Repair for Tricuspid Regurgitation) score, to predict 1-year mortality in patients undergoing T-TEER. METHODS: The EuroTR score was developed using data from the EuroTR registry, comprising 1,225 patients in the derivation cohort and 601 patients in the validation cohort. On the basis of 18 clinical, laboratory, echocardiographic, and hemodynamic parameters, an extreme gradient boosting algorithm was trained and independently validated against established risk models. RESULTS: Among the entire study cohort (N = 1,826), the overall 1-year survival rate was 82.1% (95% CI: 80.1%-84.2%), with no significant differences between the derivation and validation cohorts.
2. Three-year risk prediction of aortic stenosis using routine medical records: derivation and validation in 919 954 individuals from two cohorts.
Across 919,954 individuals from two health systems, a 3-year ML model (ASrisk) using routine labs and vitals identified higher odds of echo abnormalities even in undiagnosed individuals. Those with ASrisk >0.95 had an 11-fold enrichment for new AS diagnoses and averaged a 0.42 cm² reduction in aortic valve area over 3 years.
Impact: This study demonstrates scalable, externally validated ML risk prediction for AS using readily available EHR features, enabling targeted echocardiography and earlier intervention.
Clinical Implications: Health systems could flag high ASrisk individuals for timely echocardiography and surveillance, potentially reducing diagnostic delay and optimizing TAVR referral timing.
Key Findings
- Model derived and validated in 919,954 individuals (MSDW: 429,996; UK Biobank: 489,958).
- Increasing ASrisk associated with higher odds of echo markers of AS severity in undiagnosed individuals.
- ASrisk >0.95 conferred an ~11-fold enrichment for incident AS diagnoses within 3 years.
- High ASrisk was associated with an average aortic valve area decline of 0.42 cm² over 3 years.
Methodological Strengths
- Very large, multi-cohort derivation and external validation
- EHR-based features enable scalable, low-cost implementation across systems
Limitations
- Observational design with reliance on EHR coding and potential misclassification
- Model utility not tested in randomized screening or outcome-improvement trials
Future Directions: Prospective studies should assess whether ASrisk-guided echocardiography improves time to diagnosis and outcomes, and evaluate thresholds, calibration drift, and health-economic impact.
AIMS: All-cause mortality ranges between 33% and 42% for individuals with untreated moderate to severe aortic stenosis (AS). Transcatheter aortic valve replacement makes this a treatable condition, if identified early. Machine learning-based tools show great promise to predict cardiovascular outcomes. METHODS AND RESULTS: We developed and validated a machine learning model for 3-year prediction of AS risk (ASrisk) using serum biomarkers and vital sign measurements. We then evaluated the tool's capacity to identify diagnoses of AS sequelae, echocardiographic outcomes in individuals not diagnosed with AS, as well as enrichment and 3-year aortic valve area reduction in individuals with high ASrisk. Among 919 954 participants, 429 996 were from the Mount Sinai Data Warehouse (MSDW) [2179 (0.5%) AS cases] and 489 958 were from the UK Biobank [5066 (1%) AS cases].
3. Clinical Impact of Diabetes Mellitus After Intravascular Imaging-Guided PCI vs Coronary Artery Bypass Grafting.
Among 3,402 diabetics with left main or 3-vessel disease, overall PCI had higher 3-year events than CABG. However, IVI-guided PCI achieved comparable risk to CABG for the composite of all-cause death, nonfatal MI, or stroke (11.4% vs 12.4%; HR 0.88; P=0.525), a finding supported by propensity-matched analyses.
Impact: Challenges the long-standing paradigm that CABG is categorically superior in diabetics with complex disease by showing IVI-guided PCI can achieve comparable outcomes.
Clinical Implications: When robust intravascular imaging guidance is used, PCI may be a viable alternative to CABG for selected diabetics with left main/3-vessel disease, emphasizing strict adherence to IVI-guided optimization.
Key Findings
- In 3,402 DM patients with LM or 3-vessel disease, overall PCI had higher 3-year composite events than CABG.
- IVI-guided PCI vs CABG: 11.4% vs 12.4% for the 3-year composite (HR 0.88; 95% CI 0.58-1.32; P=0.525).
- Propensity-matched analyses yielded consistent results favoring comparability between IVI-guided PCI and CABG.
- Angiography-guided PCI had worse outcomes than CABG, underscoring the benefit of IVI guidance.
Methodological Strengths
- Individual patient-level data with large sample size and propensity score matching
- Direct comparison of IVI-guided PCI, angiography-guided PCI, and CABG in complex disease
Limitations
- Nonrandomized comparison across modalities with potential residual confounding and selection bias
- Derived from combined trial and institutional registries; generalizability may vary
Future Directions: Randomized trials of IVI-guided PCI vs CABG in diabetics with LM/3VD are warranted, with standardized IVI protocols and patient-centered outcomes.
BACKGROUND: Coronary artery bypass grafting (CABG) is the standard revascularization method for patients with diabetes mellitus (DM) and complex coronary artery disease. However, with significant advances in percutaneous coronary intervention (PCI), particularly the use of intravascular imaging (IVI), it is uncertain whether contemporary IVI-guided PCI can achieve clinical outcomes comparable with those of CABG. OBJECTIVES: The aim of this study was to compare the clinical outcomes of IVI-guided PCI, angiography-guided PCI, and CABG in DM patients with left main or 3-vessel disease. METHODS: A total of 3,402 DM patients with left main or 3-vessel disease were included from individual patient-level data of the RENOVATE-COMPLEX-PCI trial and the institutional registries of Samsung Medical Center (n = 6,962). The primary outcome, which was a composite of all-cause death, nonfatal myocardial infarction, or stroke at 3 years was compared among IVI-guided PCI, angiography-guided PCI, and CABG.