Daily Cardiology Research Analysis
Three high-impact cardiology studies stand out today: an RCT meta-analysis shows sodium–glucose cotransporter-2 inhibitors reduce atrial arrhythmias and sudden cardiac death, and two large-scale AI-ECG studies demonstrate robust prediction of hypertension and atrial fibrillation with downstream links to major cardiovascular events. Together, they advance therapeutic decision-making and enable scalable, low-cost risk stratification from routine ECGs.
Summary
Three high-impact cardiology studies stand out today: an RCT meta-analysis shows sodium–glucose cotransporter-2 inhibitors reduce atrial arrhythmias and sudden cardiac death, and two large-scale AI-ECG studies demonstrate robust prediction of hypertension and atrial fibrillation with downstream links to major cardiovascular events. Together, they advance therapeutic decision-making and enable scalable, low-cost risk stratification from routine ECGs.
Research Themes
- AI-enabled ECG biomarkers for cardiometabolic risk
- Cardiometabolic therapies and arrhythmic outcomes (SGLT2 inhibitors)
- Population-scale predictive modeling and clinical risk stratification
Selected Articles
1. Sodium-Glucose Cotransporter-2 Inhibitors and Arrhythmias: A Meta-Analysis of 38 Randomized Controlled Trials.
Across 38 RCTs including 88,704 patients, SGLT2 inhibitors reduced incident atrial arrhythmia (OR 0.85) and sudden cardiac death (OR 0.72) versus control, with no effect on ventricular arrhythmia or cardiac arrest. Benefits were observed over a mean 1.6-year follow-up.
Impact: This synthesis of RCT data links SGLT2 inhibitors to reductions in atrial arrhythmia and sudden cardiac death, outcomes not typically primary endpoints in the original trials. It informs therapeutic selection beyond glycemic and heart failure effects.
Clinical Implications: In patients with type 2 diabetes, heart failure, or CKD, SGLT2 inhibitors may be preferential when arrhythmic risk is a concern, while recognizing that ventricular arrhythmia risk is unchanged. Clinicians should not assume antiarrhythmic protection against ventricular events.
Key Findings
- Across 38 RCTs (n=88,704), SGLT2i reduced incident atrial arrhythmia (OR 0.85, 95% CI 0.75-0.98, P=0.02).
- SGLT2i reduced sudden cardiac death (OR 0.72, 95% CI 0.55-0.94, P=0.02).
- No significant effect on ventricular arrhythmia (OR 1.03) or cardiac arrest (OR 0.94).
- Mean follow-up across trials was 1.6 years.
Methodological Strengths
- Meta-analysis restricted to randomized controlled trials with large aggregate sample size (n=88,704).
- Random-effects modeling and predefined outcomes across cardiometabolic populations.
Limitations
- Arrhythmia and SCD were often secondary endpoints with heterogeneous ascertainment across trials.
- No individual patient data; limited ability to explore subgroup mechanisms or dose-response.
Future Directions: Individual patient-level meta-analyses and prospective, arrhythmia-focused RCTs are needed to confirm mechanisms, quantify AF burden reduction, and define subgroups with maximal benefit.
BACKGROUND: Sodium-glucose cotransporter-2 inhibitors (SGLT2i) have shown promising results in reducing hospitalizations from heart failure (HF) and cardiovascular mortality. However, their effect on arrhythmia and sudden cardiac death (SCD) is not well established. OBJECTIVES: The authors sought to evaluate the association between SGLT2i and the risk of arrhythmias and SCD in patients with type 2 diabetes mellitus, HF, or chronic kidney disease. METHODS: We performed a systematic literature search on PubMed, EMBASE, and Scopus for relevant randomized controlled trials from inception until February 10, 2023. ORs and 95% CIs were pooled using a random effect model. RESULTS: A total of 38 randomized controlled trials with 88,704 patients (48,435 in the SGLT2i group and 40,269 in the control group) were included in the study. The mean age of patients among SGLT2i and control groups was 56.8 and 56.7 years, respectively. The mean follow-up duration was 1.6 years. Pooled analysis of primary and secondary outcomes showed that SGLT2i significantly reduced the risk of incident atrial arrhythmia (OR: 0.85 [95% CI: 0.75-0.98], P = 0.02), SCD (OR: 0.72 [95% CI: 0.55-0.94], P = 0.02) compared with placebo. However, the risk of ventricular arrhythmia (OR: 1.03 [95% CI: 0.84-1.26], P = 0.77) and cardiac arrest (OR: 0.94 [95% CI: 0.72-1.23] P = 0.67) was comparable between both groups of patients. CONCLUSIONS: SGLT2i therapy was associated with an overall lower risk of atrial arrythmia and SCD in patients with type 2 diabetes mellitus and/or HF or chronic kidney disease. However, SGLT2i therapy was not associated with a lower risk of ventricular arrhythmia.
2. A deep learning digital biomarker to detect hypertension and stratify cardiovascular risk from the electrocardiogram.
Using 752,415 ECGs for training and 56,760 for external validation, HTN-AI identified hypertension with AUROC 0.80 (internal) and 0.77 (external). The model’s hypertension probability independently associated with mortality, heart failure, MI, stroke, and aortic dissection/rupture after multivariable adjustment.
Impact: Demonstrates a scalable, low-cost digital biomarker from routine ECGs to flag undiagnosed hypertension and stratify future CVD risk, addressing measurement gaps in blood pressure surveillance.
Clinical Implications: HTN-AI could triage patients for confirmatory ambulatory/home BP monitoring, prompt earlier lifestyle/pharmacologic treatment, and augment risk stratification without additional sensors.
Key Findings
- Internal and external AUROC for hypertension detection were 0.803 and 0.771, respectively.
- Model-predicted hypertension probability associated with mortality (HR/SD 1.47), HF (2.26), MI (1.87), stroke (1.30), and aortic dissection/rupture (1.69), all p<0.001.
- Training set: 752,415 ECGs from 103,405 adults; external validation: 56,760 adults.
Methodological Strengths
- Very large, real-world ECG datasets with external validation across institutions.
- Competing risk (Fine-Gray) analyses with multivariable adjustment linking model output to incident events.
Limitations
- Retrospective development; generalizability outside the health systems requires further validation.
- Black-box model interpretability and potential dataset/ascertainment biases inherent to EHRs.
Future Directions: Prospective implementation studies to test clinical workflows (screen–confirm–treat), impact on BP control and CVD outcomes, and fairness audits across demographics.
Hypertension is a major risk factor for cardiovascular disease (CVD), yet blood pressure is measured intermittently and under suboptimal conditions. We developed a deep learning model to identify hypertension and stratify risk of CVD using 12-lead electrocardiogram waveforms. HTN-AI was trained to detect hypertension using 752,415 electrocardiograms from 103,405 adults at Massachusetts General Hospital. We externally validated HTN-AI and demonstrated associations between HTN-AI risk and incident CVD in 56,760 adults at Brigham and Women's Hospital. HTN-AI accurately discriminated hypertension (internal and external validation AUROC 0.803 and 0.771, respectively). In Fine-Gray regression analyses model-predicted probability of hypertension was associated with mortality (hazard ratio per standard deviation: 1.47 [1.36-1.60], p < 0.001), HF (2.26 [1.90-2.69], p < 0.001), MI (1.87 [1.69-2.07], p < 0.001), stroke (1.30 [1.18-1.44], p < 0.001), and aortic dissection or rupture (1.69 [1.22-2.35], p < 0.001) after adjustment for demographics and risk factors. HTN-AI may facilitate diagnosis of hypertension and serve as a digital biomarker of hypertension-associated CVD.
3. Utilizing 12-lead electrocardiogram and machine learning to retrospectively estimate and prospectively predict atrial fibrillation and stroke risk.
In 1.78 million patients, CNN models using 12-lead ECG achieved AUROC 0.99 for AF diagnosis, 0.86 for past AF estimation, and 0.85 for future AF prediction. Individuals flagged as having past or future AF had higher risks of stroke, HF hospitalization, MI, and death.
Impact: Shows that a single ECG can retrospectively and prospectively flag AF risk, enabling targeted screening and anticoagulation strategies to prevent embolic stroke at scale.
Clinical Implications: ECG-based AI can prioritize prolonged rhythm monitoring and timely anticoagulation decisions in high-risk individuals before overt AF, potentially reducing stroke burden.
Key Findings
- CNN models achieved AUROC 0.99 (diagnosis), 0.86 (past AF), and 0.85 (future AF) from 12-lead ECG.
- Predicted past/future AF associated with higher risks of stroke, HF hospitalization, MI, and mortality.
- Model interpretability suggested contributions from QRS in V1/aVL/aVR and ECG features like low R-wave amplitude and flattened T waves.
Methodological Strengths
- Extremely large, multi-branch health system dataset enabling robust model training and validation.
- Linked clinical outcomes demonstrating prognostic value beyond AF detection.
Limitations
- Observational design with potential EHR biases; external generalizability outside system remains to be proven.
- Outcome follow-up duration not detailed; intervention impact on stroke prevention not tested.
Future Directions: Randomized screening trials using AI-ECG triage for extended monitoring and anticoagulation, plus external validations across diverse healthcare systems.
BACKGROUND: The stroke risk in patients with subclinical atrial fibrillation (AF) is underestimated. By identifying patients at high risk of embolic stroke, health-care professionals can make more informed decisions regarding anticoagulation treatment to prevent stroke. The main aim of this study was to forecast the risk of AF both retrospectively and prospectively. METHODS: The research used a dataset of patients who had received a standard 12-lead electrocardiogram (ECG) at the seven branches of Chang Gung Memorial Hospital between October 2007 and December 2019. Using convolutional neural network (CNN) ECG models, the study classified the risk of AF development both retrospectively and prospectively in 1,776,968 patients by analyzing their 12-lead ECG. The study also examined the risk of stroke, hospitalization for heart failure (HF), myocardial infarction (MI), and death among patients with predicted AF versus that of those with normal sinus rhythm. RESULTS: The CNN models could be used to accurately diagnose AF, assess the risk of past AF episodes, and predict the risk of future AF episodes with high accuracy, as shown by areas under the receiver operating characteristic curve of 0.99, 0.86, and 0.85, respectively. Patients who were estimated to have had past AF or predicted to have future AF were at a higher risk of developing stroke, HF hospitalization, MI, and mortality. The ECGs of patients with predicted AF tended to exhibit lower R-wave amplitudes and flattened T waves. Additionally, we observed that the QRS complexes in leads V1, aVL, and aVR were highly weighted in predicting AF in the CNN models. CONCLUSIONS: The CNN models were effective for estimating the past and future risk of AF by analyzing 12-lead ECG. Patients with predicted AF had a higher risk of developing stroke, hospitalization for HF, MI, and death. By using this AF prediction model, physicians may be able to identify patients who should be screened for AF and taking action to prevent stroke and manage cardiovascular risk.