Publications

The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack-Lipid-Lowering Trial (ALLHAT-LLT) was a pragmatic randomized trial conducted in 513 clinical centers across North America from 1994 to 2002. ALLHAT-LLT, which used standard of care (usual care) as the control treatment, did not find an effect of assignment to pravastatin on all-cause death. We aimed to use the trial data to estimate the effect of adhering to assigned treatment, that is, the per-protocol effect. Moderately hypercholesterolemic (low-density lipoprotein cholesterol 100-189 mg/dL), hypertensive participants ≥55 years old were included and randomized to pravastatin (40 mg/day unless contraindicated) or usual care (lipid-lowering treatment discouraged unless clinically indicated). Of 10 355 individuals randomized, we included 9741 with complete baseline data. One-third of individuals in the pravastatin arm were nonadherent to the assigned treatment strategy at year 5. After adjusting for factors related to adherence, the 5-year risk difference comparing pravastatin to usual care was -3.0% (-4.9%, -1.3%) and the risk ratio was 0.79 (0.68, 0.91) for death. Findings were consistent across different definitions of treatment strategies. Our results suggest that the use of pravastatin under full adherence was protective against death, while assignment to pravastatin was not.

BACKGROUND: Cardiac MRI (CMR) markers of myocardial fibrosis and infiltration are diagnostically and prognostically important in cardiomyopathies. A noninvasive ECG correlate of CMR interstitial fibrosis measurements (extracellular volume [ECV] and native T1) could assist in diagnosis, risk stratification, and tracking disease progress. The spatial ventricular gradient (SVG) is a vectorcardiographic (VCG) measure of electrical heterogeneity obtained from a 12-lead ECG. The link between the SVG and CMR-derived myocardial interstitial fibrosis is unknown.

METHODS: Retrospective study of patients referred for CMR from 2018-2022 at a single academic center, with an ECG performed within 30 days. VCGs were constructed from 12-lead ECGs, and SVG vector components were calculated. ECV and T1 values were regressed on SVG components, demographics, and ECG parameters using linear regression.

RESULTS: In total, 345 patients met inclusion criteria: 55% male, median age 60.2 (P25-P75 47.4-69.9) years, and median LVEF 57 (P25-P75 44-63)%. Median SVG magnitude was 39.5 (P25-P75 28.2-58.2) mV∙ms and SVG magnitude was inversely correlated with ECV and T1 (p < 0.001 for both). In a multivariable model, SVG magnitude remained independently associated with ECV and T1 after adjustment for body mass index, LVEF, age, and sex (p < 0.01). Patients with amyloid cardiomyopathy had the most abnormal values of ECV and T1; no amyloid patient had SVG magnitude > 50 mV∙ms.

CONCLUSION: SVG magnitude is correlated with CMR-derived ECV and native T1. Patients with high SVG magnitude were not observed to have a large burden of diffuse fibrosis or infiltration.

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.

BACKGROUND: Early detection of cardiovascular disease is a global public health priority. Artificial intelligence (AI)-enabled stethoscopes offer robust performance characteristics in point-of-care detection of heart failure, atrial fibrillation, and valvular heart disease (VHD). We conducted a pragmatic, cluster-randomised controlled implementation trial to determine the real-world effect and implementation challenges of AI-stethoscopes.

METHODS: UK primary care practices were cluster randomised 1:1 to intervention (training and implementation in use of AI-stethoscopes in routine care) or control (routine care). Given the nature of the intervention, masking of participants (practices, clinicians, and patients) was not feasible. During cardiac examinations, the AI stethoscope recorded 15 s of single-lead electrocardiogram and phonocardiogram signals for input to three AI algorithms that returned binary predictions for the presence or absence of reduced left ventricular ejection fraction (≤40%), atrial fibrillation, and VHD (all with regulatory approval). The primary endpoint was incidence of any newly coded diagnosis of heart failure (all subtypes), expressed per 1000 patient-years (incidence rate ratio [IRR]), derived from a UK National Health Service Secure Data Environment. A coprimary endpoint stratified detection of heart failure by place of diagnosis (community-based vs via hospital admission). Secondary endpoints included atrial fibrillation and VHD detection rates, performance characteristics of the AI-stethoscope, use rates, and clinician-reported implementation barriers and enablers.

FINDINGS: Between Oct 30, 2023, and May 22, 2024, 205 practices were randomly assigned (96 to the intervention arm [701 933 registered patients] and 109 to the control arm [851 242 registered patients]). Intervention practices recorded 12 725 patient examinations with the AI-stethoscope, across 972 clinical users. Intention-to-treat analysis found heart failure detection did not differ between groups (IRR 0·94 [95% CI 0·86-1·02]); with no difference in community-based or hospital-based diagnoses (p>0·05).

INTERPRETATION: Implementation of an AI stethoscope in routine primary care did not significantly increase detection of heart failure or increase community-based diagnosis after 12 months of implementation. AI stethoscope use was independently associated with significantly higher detection rates of heart failure, as well as atrial fibrillation and VHD. This randomised controlled implementation trial establishes a pragmatic design with randomisation that generates real-world data essential for understanding and overcoming the barriers to implementation of innovation in health care.

FUNDING: National Institute for Health and Care Research, British Heart Foundation, and Imperial Health Charity.

BACKGROUND: Early prediction of atrial fibrillation (AF) is crucial for reducing adverse outcomes. While artificial intelligence-enhanced electrocardiogram (AI-ECG) analysis shows promise in predicting AF, most approaches require digital ECG signals, limiting their application in settings where ECGs are stored as images.

OBJECTIVE: We aimed to develop and validate an image-based AI-ECG approach for predicting incident AF across multiple datasets.

METHODS: We used 1,163,401 ECGs from 189,539 patients in the Beth Israel Deaconess Medical Center (BIDMC) dataset and 70,655 ECGs from 65,610 participants in the United Kingdom (UK) Biobank. The AI-ECG model was trained on ECG images processed to 310x868 pixels.

RESULTS: The model achieved C-statistics of 0.754 (95% confidence interval [CI]: 0.747-0.761) in the BIDMC dataset and 0.723 (95% CI: 0.704-0.741) in the UK Biobank for predicting incident AF. Performance was maintained across key subgroups including outpatients, women, and non-white individuals. Compared with the CHARGE-AF risk score, the AI-ECG model showed superior performance (c-statistic 0.696 vs 0.667, P < .05) and provided significant additive value when combined (c-statistic 0.711, P < .0001). The model also performed well on smartphone-photographed ECGs (c-statistic 0.736). Saliency mapping indicated the model primarily focused on P-wave morphology and PR interval regions.

CONCLUSION: This image-based approach enables AI-ECG prediction of AF in settings without digital ECG infrastructure and provides additive value to known clinical risk scores.

BACKGROUND: The COVID-19 pandemic presented unprecedented challenges to hospital system and critical care resources, leading to significant changes to operations and patient care. There are limited national data on these changes and instances of unsanctioned deviations from patient care, yet understanding the COVID response is key to future preparedness efforts. We sought to understand how hospitals and states navigated scarcity during COVID-19, particularly in the absence of a declaration of crisis standards of care.

METHODS: Between February 2022 and September 2022 we conducted 34 interviews with 36 leaders of U.S. states' COVID-19 planning and response efforts. Interviews were transcribed verbatim and verified. We analyzed interviews using iterative inductive thematic analysis for descriptions of resource scarcity and changes to policies and procedures to prevent rationing lifesaving care.

RESULTS: Nearly all participants described equipment and personnel scarcity in their home institution or state during COVID-19. Hospitals across regions and states developed formal and informal coordination processes for load and resource sharing in response to influxes of high-acuity patients, avoiding formal rationing of lifesaving resources in many regions. Participants also described unsanctioned patient triage, early discharge, and patients counseled to accept less aggressive care (e.g., premature transition to hospice) in states that had not declared crisis standards of care.

CONCLUSIONS: Extending limited resources and inter-institutional collaboration helped avoid formal rationing. Yet, patient care was unquestionably impacted due to scarcity, both real and perceived. Reports of using hospital triage protocols to deny patients lifesaving care outside of formally recognized crisis conditions and attempts to nudge patients to accept less-resource-intensive care are concerning. This may have had disproportionate effects on older adults, individuals with disabilities, and racial and ethnic minoritized groups. To avoid unsanctioned deviations from standard practice in future health emergencies, we recommend that transparent and equitable triage protocols are implemented with robust oversight.

While the cardiometabolic benefits of exercise volume and intensity are well established, the clinical significance of exercise timing remains poorly understood, largely due to the limitations of short-term accelerometry. We leveraged minute-level heart rate data from 14,489 participants from the All of Us Research Program to define habitual exercise timing over a one-year period. Compared to daytime exercise, habitual morning exercise was associated with lower odds of coronary artery disease (OR 0.69; CI 0.55-0.87), hypertension (OR 0.82; CI 0.72-0.94), type 2 diabetes (OR 0.70; CI 0.58-0.85), hyperlipidemia (OR 0.79; CI 0.69-0.90), and obesity (OR 0.65; CI 0.55-0.77). These associations were independent of total physical activity volume and remained consistent across hour-of-day analyses, with the lowest risk nadir occurring between 07:00-08:00 for coronary artery disease. These findings suggest that exercise timing may represent a distinct, underappreciated dimension of exercise behavior linked to cardiometabolic health.

OBJECTIVE: The optimal management of patients with asymptomatic carotid stenosis (AsxCS) and symptomatic carotid stenosis (SxCS) is controversial and includes intensive medical management (ie, best medical therapy [BMT]) with or without an additional carotid revascularization procedure (ie, carotid endarterectomy [CEA], transfemoral carotid artery stenting [TFCAS] or transcarotid artery revascularization [TCAR]). The aim of this international, expert-based, multispecialty Delphi consensus document was to reconcile the conflicting views regarding the optimal management of AsxCS and SxCS patients.

METHODS: A three-round Delphi consensus process was performed including 63 experts from Europe (n = 37) and the United States (n = 26). A total of six different clinical scenarios were identified involving patients with either AsxCS or SxCS. For each scenario, five treatment options were available: (i) BMT alone, (ii) BMT plus CEA, (iii) BMT plus TFCAS, (iv) BMT plus TCAR, and (v) BMT plus CEA/TFCAS/TCAR. Differences in treatment preferences between US and European participants were assessed using Fisher's exact test, and odds ratios were used to quantify the magnitude and direction of association. Consensus was achieved when >70% of the Delphi consensus participants agreed on a therapeutic approach.

RESULTS: Most participants concurred that BMT alone is not adequate for the management of a 70-year-old fit male or female patient with 80% to 99% AsxCS (52/63 [82.5%] and 45/63 [71.5%], respectively). In contrast, most panelists would opt for BMT alone for an 80-year-old male AsxCS patient with several comorbidities (48/63 [76.2%]). The majority of participants would opt for BMT plus a carotid revascularization procedure for an 80-year-old male SxCS patient with a recent ipsilateral cerebrovascular event, an ipsilateral 70% to 99% SxCS, and a 5-year predicted risk of ipsilateral ischemic event of 10% (54/63 [85.7%]), 15% (59/63 [93.6%]), or 20% (63/63 [100%]). The opinion of US-based participants varied from that of Europe-based respondents in some scenarios.

CONCLUSIONS: The panel agreed that BMT alone is insufficient for most patients with SxCS, and that select subgroups of AsxCS patients may also benefit from revascularization, especially when high-risk features are present. Patients should be stratified according to their predicted stroke risk, as well as their individual clinical, anatomical, and imaging features and should be treated accordingly.