Daily Cardiology Research Analysis
Three studies shape current cardiology practice and research: a massive multi-task deep learning model using routine 12‑lead ECGs accurately predicts 1‑year major adverse cardiovascular events and long‑term risk; a randomized trial in VA‑ECMO shows no 1‑year benefit of early routine left ventricular unloading; and a post hoc analysis indicates QFR‑based deferral of PCI is riskier than FFR‑based deferral. Together, they impact prevention, critical care strategy, and physiology‑guided PCI.
Summary
Three studies shape current cardiology practice and research: a massive multi-task deep learning model using routine 12‑lead ECGs accurately predicts 1‑year major adverse cardiovascular events and long‑term risk; a randomized trial in VA‑ECMO shows no 1‑year benefit of early routine left ventricular unloading; and a post hoc analysis indicates QFR‑based deferral of PCI is riskier than FFR‑based deferral. Together, they impact prevention, critical care strategy, and physiology‑guided PCI.
Research Themes
- AI-driven ECG risk prediction for MACEs
- Strategy optimization in VA-ECMO (LV unloading)
- Physiology-guided PCI: QFR vs FFR for deferral safety
Selected Articles
1. A multitask deep learning model utilizing electrocardiograms for major cardiovascular adverse events prediction.
Using 2.82 million ECGs and external validation, ECG‑MACE accurately predicted 1‑year first‑ever HF, MI, ischemic stroke, and mortality and outperformed Framingham risk scores for longer‑term outcomes. The model also stratified 10‑year incidence, highlighting its potential for scalable preventive risk screening from routine ECGs.
Impact: This is one of the largest ECG-based prognostic AI studies with external validation and strong performance, offering a low-cost, widely deployable risk stratification tool.
Clinical Implications: Health systems could integrate ECG‑MACE into EHRs to flag high‑risk patients for early preventive interventions, targeted diagnostics, or closer follow‑up, even before traditional risk factors accumulate.
Key Findings
- Trained on 2,821,889 12‑lead ECGs with external validation (n=113,224), the model achieved AUROCs: HF 0.90, MI 0.85, ischemic stroke 0.76, mortality 0.89.
- Outperformed Framingham risk scores for 5‑year MACEs and 10‑year mortality prediction.
- Over 10 years, model‑positive patients had substantially higher incidence ratios (HF 15.28; MI 7.87; IS 4.74; mortality 13.18) than model‑negative patients.
Methodological Strengths
- Very large multi-cohort dataset with external validation
- Multitask learning with long-term incidence stratification analyses
Limitations
- Retrospective design without prospective clinical deployment
- Generalizability across devices/health systems and model interpretability need further assessment
Future Directions: Prospective, multi-center deployment trials to assess clinical utility, workflow integration, outcomes impact, and health equity; calibration across vendors and regions; interpretable AI methods.
Deep learning analysis of electrocardiography (ECG) may predict cardiovascular outcomes. We present a novel multi-task deep learning model, the ECG-MACE, which predicts the one-year first-ever major adverse cardiovascular events (MACE) using 2,821,889 standard 12-lead ECGs, including training (n = 984,895), validation (n = 422,061), and test (n = 1,414,933) sets, from Chang Gung Memorial Hospital database in Taiwan. Data from another independent medical center (n = 113,224) was retrieved for external validation. The model's performance achieves AUROCs of 0.90 for heart failure (HF), 0.85 for myocardial infarction (MI), 0.76 for ischemic stroke (IS), and 0.89 for mortality. Furthermore, it outperforms the Framingham risk score at 5-year MACEs and 10-year mortality prediction. Over 10-year follow-ups, the model-predicted-positive group exhibits significantly higher MACE incidences than the model-predicted-negative group (relative incidence ratio: HF: 15.28; MI: 7.87; IS: 4.74; mortality: 13.18). Using solely ECGs, ECG-MACE effectively predicts one-year events and exhibits long-term anticipation. It provides potential applications in preventive medicine.
2. Early left ventricular unloading after venoarterial extracorporeal membrane oxygenation: 1-year outcomes of the EARLY-UNLOAD randomized clinical trial.
In 116 cardiogenic shock patients on VA‑ECMO, early routine LV unloading (transseptal LA cannulation within 12 h) did not improve 1‑year all‑cause mortality, HF rehospitalization, or composite outcomes versus a conventional rescue approach.
Impact: Provides high-quality randomized evidence against routine early LV unloading on VA‑ECMO at 1 year, informing critical care strategies.
Clinical Implications: Routine early LV unloading after VA‑ECMO initiation should not be expected to improve 1‑year outcomes; selective rescue unloading based on hemodynamics remains appropriate pending further evidence.
Key Findings
- Early LV unloading vs conventional rescue: 1‑year all‑cause mortality 56.9% vs 57.1% (HR 0.97; 95% CI 0.60–1.58; p=0.887).
- No significant differences in cardiac/non‑cardiac mortality, HF rehospitalization (HR 1.17; 95% CI 0.43–3.24; p=0.758), or composite outcomes.
- Follow‑up completeness was high (98.3% at 1 year).
Methodological Strengths
- Randomized clinical trial with prespecified 1‑year endpoints
- High follow‑up completeness and clinically relevant outcomes
Limitations
- Single‑center, open‑label design and modest sample size may limit power
- Allowance of rescue unloading may attenuate between‑group differences
Future Directions: Larger, multicenter RCTs to test specific unloading modalities, timing, and patient phenotypes; exploration of mechanistic endpoints (LV distension, pulmonary edema, biomarkers).
AIMS: The long-term effects of early left ventricular (LV) unloading after venoarterial extracorporeal membrane oxygenation (VA-ECMO) remain unclear. METHODS AND RESULTS: The EARLY-UNLOAD trial was a single-centre, investigator-initiated, open-label, randomized clinical trial involving 116 patients with cardiogenic shock (CS) undergoing VA-ECMO. The patients were randomly assigned to undergo either early routine LV unloading by transseptal left atrial cannulation within 12 h after randomization or the conventional approach, which permitted rescue transseptal cannulation in case of an increased LV afterload. The pre-specified secondary endpoints at 1 year included all-cause mortality, cardiac mortality, non-cardiac mortality, rehospitalization for heart failure (HF), and the composite of all-cause mortality or rehospitalization for HF. At 1 year, data for 114 of 116 patients (98.3%) were available for analysis. All-cause death had occurred in 33 of 58 patients (56.9%) in early group and 32 of 56 patients (57.1%) in conventional group {hazard ratio [HR], 0.97 [95% confidence interval (CI), 0.60 to 1.58], P = 0.887}. There was no significant difference in cardiac or non-cardiac mortality. Among 61 survivors at 30 days, the incidence of rehospitalization for HF at 1 year was comparable between two groups [HR, 1.17 (95% CI 0.43 to 3.24), P = 0.758]. The incidence of the composite outcome of all-cause mortality or rehospitalization for HF also did not differ between the groups [HR, 1.01 (95% CI 0.69 to 1.76), P = 0.692]. CONCLUSION: Among patients with CS undergoing VA-ECMO, early routine LV unloading did not improve clinical outcomes at 1 year of follow-up. REGISTRATION: ClinicalTrials.gov: NCT04775472.
3. Coronary revascularisation deferral based on quantitative flow ratio or fractional flow reserve: a post hoc analysis of the FAVOR III Europe trial.
In a post hoc analysis of FAVOR III Europe, QFR‑based deferral of revascularization was associated with higher 1‑year MACE than FFR‑based deferral in patients with complete study‑lesion deferral (adjusted HR 2.07, p=0.03), with a non‑significant trend in the broader ‘any‑lesion deferral’ group.
Impact: These results caution against relying on QFR alone to defer PCI and will influence physiology‑guided decision‑making and guideline discussions.
Clinical Implications: For deferral decisions, FFR remains the safer standard; centers using QFR should consider confirmatory FFR, especially when planning complete lesion deferral.
Key Findings
- Complete study‑lesion deferral: MACE 5.6% (QFR) vs 2.8% (FFR); adjusted HR 2.07 (95% CI 1.07–4.03); p=0.03.
- Any‑lesion deferral subgroup: MACE 5.6% (QFR) vs 3.6% (FFR); adjusted HR 1.55 (95% CI 0.88–2.73); p=0.13.
- Deferral based on QFR >0.80 was less safe than FFR >0.80 in terms of 1‑year MACE in complete‑deferral patients.
Methodological Strengths
- Large deferred subcohort derived from a randomized trial
- Adjusted hazard analyses with clinically relevant endpoints
Limitations
- Post hoc, non-randomized comparison of deferral strategies within trial arms
- Residual confounding and selection effects cannot be excluded
Future Directions: Prospective studies to define when QFR can safely defer PCI and to integrate hybrid strategies (QFR screening with confirmatory FFR) and patient-level thresholds.
BACKGROUND: Safe deferral of revascularisation is a key aspect of physiology-guided percutaneous coronary intervention (PCI). While recent evidence gathered in the FAVOR III Europe trial showed that quantitative flow ratio (QFR) guidance did not meet non-inferiority to fractional flow reserve (FFR) guidance, it remains unknown if QFR might have a specific value in revascularisation deferral. AIMS: We aimed to evaluate the safety of coronary revascularisation deferral based on QFR as compared with FFR. METHODS: Patients randomised in the FAVOR III trial in whom PCI was deferred in at least one coronary artery, based on QFR or FFR>0.80, were included in the present substudy. The primary outcome was the 1-year rate of major adverse cardiac events (MACE), with results reported for two subsets of deferred patients: (1) any study lesion deferral and (2) complete study lesion deferral. RESULTS: A total of 523 patients (55.2%) in the QFR group and 599 patients (65.3%) in the FFR group had at least one coronary revascularisation deferral. Of these, 433 patients (82.8%) and 511 (85.3%) patients, respectively, had complete study lesion deferral. In the "complete study lesion deferral" patient group, the occurrence of MACE was significantly higher in QFR-deferred patients as compared with FFR-deferred patients (24 [5.6%] vs 14 [2.8%], adjusted hazard ratio [HR] 2.07, 95% confidence interval [CI]: 1.07-4.03; p=0.03). In the subgroup of "any study lesion deferral", the MACE rate was 5.6% vs 3.6% (QFR vs FFR), adjusted HR 1.55, 95% CI: 0.88-2.73; p=0.13. CONCLUSIONS: QFR-based deferral of coronary artery revascularisation resulted in a higher incidence of 1-year MACE as compared with FFR-based deferral.