Publications

BACKGROUND: Patients with cardiovascular (CV) diseases are increasingly frail but rarely represented in trials. Understanding effect modification by frailty on CV trials is critical as it could help define treatment strategies in frail patients.

OBJECTIVES: This meta-analysis aims to assess the implications of frailty on CV outcomes in clinical trials.

METHODS: Randomized controlled trials examining the effects of frailty in the context of CV trials were included (CRD42024528279). Outcomes included a composite of major adverse cardiac events (MACE), all-cause mortality, CV mortality, hospitalizations, and frailty-specific outcomes (physical, quality of life, and frailty scores). HRs and 95% CIs were pooled for clinical endpoints, and standardized mean differences (SMDs) were calculated for frailty-specific outcomes.

RESULTS: Thirty unique randomized controlled trials were included with a pooled total of 87,711 participants. Frail patients had a significantly increased risk of MACE (HR: 2.33 [95% CI: 1.87-2.91], P < 0.001, I2 = 83%), all-cause mortality (HR: 2.34 [95% CI: 1.80-3.05], P < 0.01, I2 = 75%), CV mortality (HR: 1.76 [95% CI: 1.60-1.93], P < 0.001, I2 = 0%), and hospitalizations (HR: 2.38 [95% CI: 1.65-3.43], P < 0.001, I2 = 92%) compared to nonfrail patients. In the frailest group, trial interventions decreased MACE (HR: 0.81 [95% CI: 0.74-0.88], P < 0.001, I2 = 0%) and hospitalization (HR: 0.81 [95% CI: 0.72-0.90], P < 0.001, I2 = 0%) risks with no significant difference in mortality risk (P > 0.05) compared with the control group. Trial interventions significantly improved physical (SMD: 0.15, 0.04-0.26) and quality of life (SMD: 0.15, 0.09-0.21) but not frailty scores (P > 0.05).

CONCLUSIONS: While frailty prognosticated a higher risk of CV events and mortality, frailty did not reduce treatment efficacy. CV trial interventions appear beneficial even in the frailest group.

The PRIME (Proposed Requirements for Cardiovascular Imaging-Related Machine Learning Evaluation) 2.0 checklist is an updated, domain-specific framework designed to standardize the development, evaluation, and reporting of artificial intelligence (AI) applications in cardiovascular imaging. This update specifically responds to rapid advances from traditional machine learning to deep learning, large language models, and multimodal generative AI. The updated checklist was developed through a modified Delphi process by an international panel of clinical and technical experts. In contrast to general AI reporting guidelines, it delivers detailed, practical recommendations on all critical aspects of AI research and builds upon the original 7-domain framework by incorporating cardiovascular imaging-specific complexities such as cardiac motion, imaging artifacts, and interobserver variability. By promoting transparency and rigor, PRIME 2.0 can serve as a vital resource for researchers, clinicians, peer reviewers, and journal editors working at the forefront of AI in cardiovascular imaging.

Hypertensive heart disease (HHD) is a major contributor to cardiovascular (CV) morbidity and mortality. Once primarily seen in older adults, recent data suggest a rising burden among younger populations. National Center for Health Statistics mortality data for United States adults aged 15 to 44 from 1999 to 2024 were analyzed. Age-adjusted mortality rates were calculated overall and by demographic subgroup, including sex, race, ethnicity, age group, rural and urban residence, state, and Census region. The proportion of HHD mortality relative to other CV disease (CVD) deaths were examined. Joinpoint regression identified annual percent changes and inflection points. From 1999 to 2024, there were 119,264 HHD-related deaths among young adults. HHD mortality rose from 1.3 (95% CI, 1.23-1.36) to 6.3 (95% CI, 6.12-6.40), with the sharpest increase from 2018 to 2021. Males experienced greater HHD mortality over the study period (increasing from 1.76 to 9.13 deaths per 100,000 person-years) than females (0.76-3.31 deaths per 100,000 person-years). Differences were also noted by race and ethnicity, with non-Hispanic Black individuals experiencing greater HHD mortality that non-Hispanic White and Hispanic individuals. Age-related and geographic differences were also observed. The proportionate HHD mortality increased from 3.8% in 1999 to 16.8% in 2024. Sustained increases in HHD mortality were observed after the COVID-19 pandemic relative to prepandemic levels. HHD-related mortality among young adults in the United States has risen significantly, with differences noted by sex, race and ethnicity, age, rural and urban residence, state, and Census region. The growing share of HHD deaths among CVD deaths in young adults signals its increasing role in premature CVD mortality. In conclusion, these trends underscore the urgent need for early prevention, equitable care, and targeted strategies to reduce HHD in young adults.

BACKGROUND: Abnormal ventricular depolarization, evident as a broad QRS complex on an ECG, is traditionally categorized into left bundle-branch block (LBBB) and right bundle-branch block or nonspecific intraventricular conduction delay. This categorization, although physiologically accurate, may fail to capture the nuances of diseases subtypes.

METHODS: We used unsupervised machine learning to identify and characterize novel broad QRS phenogroups. First, we trained a variational autoencoder on 1.1 million ECGs and discovered 51 latent features that showed high disentanglement and ECG reconstruction accuracy. We then extracted these features from 42 538 ECGs with QRS durations >120 milliseconds and employed a reversed graph embedding method to model population heterogeneity as a tree structure with different branches representing phenogroups.

RESULTS: Six phenogroups were identified, including phenogroups of right bundle-branch block and LBBB with varying risk of cardiovascular disease and mortality. The higher risk right bundle-branch block phenogroup exhibited increased risk of cardiovascular death (adjusted hazard ratio [aHR], 1.46 [1.30-1.63], P<0.0001) and all-cause mortality (aHR, 1.24 [1.16-1.33], P<0.0001) compared with the baseline phenogroup. Within LBBB ECGs, tree position predicted future cardiovascular disease risk differentially. Additionally, for subjects with LBBB undergoing cardiac resynchronization therapy, tree position predicted cardiac resynchronization therapy response independent of covariates, including QRS duration (adjusted odds ratio [aOR], 0.47 [0.25-0.86], P<0.05).

CONCLUSIONS: Our findings challenge the current paradigm, highlighting the potential for these phenogroups to enhance cardiac resynchronization therapy patient selection for subjects with LBBB and guide investigation and follow-up strategies for subjects with higher risk right bundle-branch block.

INTRODUCTION: Complete heart block (CHB) is a life-threatening condition that can lead to ventricular standstill, syncopal injury, and sudden cardiac death, and current electrocardiography (ECG)-based risk stratification (presence of bifascicular block) is crude and has limited performance. Artificial intelligence-enhanced electrocardiography (AI-ECG) has been shown to identify a broad spectrum of subclinical disease and may be useful for CHB.

OBJECTIVE: To develop an AI-ECG risk estimator for CHB (AIRE-CHB) to predict incident CHB.

DESIGN, SETTING, AND PARTICIPANTS: This cohort study was a development and external validation prognostic study conducted at Beth Israel Deaconess Medical Center and validated externally in the UK Biobank volunteer cohort.

EXPOSURE: Electrocardiogram.

MAIN OUTCOMES AND MEASURES: A new diagnosis of CHB more than 31 days after the ECG. AIRE-CHB uses a residual convolutional neural network architecture with a discrete-time survival loss function and was trained to predict incident CHB.

RESULTS: The Beth Israel Deaconess Medical Center cohort included 1 163 401 ECGs from 189 539 patients. AIRE-CHB predicted incident CHB with a C index of 0.836 (95% CI, 0.819-0.534) and area under the receiver operating characteristics curve (AUROC) for incident CHB within 1 year of 0.889 (95% CI, 0.863-0.916). In comparison, the presence of bifascicular block had an AUROC of 0.594 (95% CI, 0.567-0.620). Participants in the high-risk quartile had an adjusted hazard ratio (aHR) of 11.6 (95% CI, 7.62-17.7; P < .001) for development of incident CHB compared with the low-risk group. In the UKB UK Biobank cohort of 50 641 ECGs from 189 539 patients, the C index for incident CHB prediction was 0.936 (95% CI, 0.900-0.972) and aHR, 7.17 (95% CI, 1.67-30.81; P < .001).

CONCLUSIONS AND RELEVANCE: In this study, a first-of-its-kind deep learning model identified the risk of incident CHB. AIRE-CHB could be used in diverse settings to aid in decision-making for individuals with syncope or at risk of high-grade atrioventricular block.

BACKGROUND: Multiple risk scores and biomarkers have been proposed for the prediction of atrial fibrillation (AF), but it is unknown how these compare with each other and whether they could be combined.

OBJECTIVE: This study aimed to evaluate and compare approaches for incident AF prediction.

METHODS: The artificial intelligence-enhanced electrocardiogram risk estimator-AF (AIRE-AF), a convolutional neural network with a discrete-time survival loss function, was developed to predict incident AF. It was trained using a dataset of 1,163,401 electrocardiograms from 189,539 patients from the Beth Israel Deaconess Medical Center and externally validated in the UK Biobank (n = 38,892). AIRE-AF was compared with other risk prediction approaches including the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE)-AF, a clinical risk score.

RESULTS: In the Beth Israel Deaconess Medical Center cohort, AIRE-AF predicted incident AF with a C-index of 0.750 (0.743-0.758). AIRE-AF was superior to CHARGE-AF, left atrial size, and N-terminal pro-B-type natriuretic peptide. The addition of CHARGE-AF and left atrial size provided a minor improvement in performance (C-index improvement 0.017). There was no additive value of N-terminal pro-B-type natriuretic peptide in combination with AIRE-AF. The single best-performing single predictor in the volunteer population (UK Biobank) was CHARGE-AF (C-index 0.750 [0.734-0.769]). The best-performing combination of 2 predictors was AIRE-AF and CHARGE-AF (C-index 0.768 [0.743-0.792]). The addition of a polygenic risk score to AIRE-AF and CHARGE-AF provided a further significant improvement in performance (C-index 0.791 [0.766-0.816]).

CONCLUSION: We present the first comprehensive evaluation of methodologies for predicting incident AF. Risk prediction with a model including AIRE-AF and CHARGE-AF resulted in similar performance to more complex models.