AI-Enhanced ECG Predicts Hypertension, Related Risks

TOPLINE:

An artificial intelligence–ECG risk estimation model designed to predict incident hypertension (AIRE-HTN) identifies cases and stratifies the risk for adverse outcomes in addition to traditional markers.

METHODOLOGY:

• Researchers conducted a development and external validation prognostic cohort study in a secondary care setting to identify individuals at risk for incident hypertension.
• They developed AIRE-HTN, which was trained on a derivation cohort from the Beth Israel Deaconess Medical Center in Boston, involving 1,163,4src1 ECGs from 189,539 patients (mean age, 57.7 years; 52.1% women; 64.5% White individuals).
• External validation was conducted on 65,61src ECGs from a UK-based volunteer cohort, drawn from an equal number of patients (mean age, 65.4 years; 51.5% women; 96.3% White individuals).
• Incident hypertension was evaluated in 19,423 individuals without hypertension from the medical center cohort and in 35,8src6 individuals without hypertension from the UK cohort.

TAKEAWAY:

• AIRE-HTN predicted incident hypertension with a C-index of src.7src (95% CI, src.69-src.71) in both the cohorts. Those in the quartile with the highest AIRE-HTN scores had a fourfold increased risk for incident hypertension (P <.srcsrc1).
• The model’s predictive accuracy was maintained in individuals without left ventricular hypertrophy and those with normal ECGs and baseline blood pressure, indicating its robustness.
• The model was significantly additive to traditional clinical markers, with a continuous net reclassification index of src.44 for the medical center cohort and src.32 for the UK cohort.
• AIRE-HTN was an independent predictor of cardiovascular death (hazard ratio per 1-SD increase in score [HR], 2.24), heart failure (HR, 2.6src), myocardial infarction (HR, 3.13), ischemic stroke (HR, 1.23), and chronic kidney disease (HR, 1.89) in outpatients from the medical center cohort (all P <.srcsrc1), with largely consistent findings in the UK cohort.

IN PRACTICE:

“Results of exploratory and phenotypic analyses suggest the biological plausibility of these findings. Enhanced predictability could influence surveillance programs and primordial prevention,” the authors wrote.

SOURCE:

The study was led by Arunashis Sau, PhD, and Joseph Barker, MRes, National Heart and Lung Institute, Imperial College London, England. It was published online on January 2, 2src24, in JAMA Cardiology.

LIMITATIONS:

In one cohort, hypertension was defined using International Classification of Diseases codes, which may lack granularity and not align with contemporary guidelines. The findings were not validated against ambulatory monitoring standards. The performance of the model in different populations and clinical settings remains to be explored.

DISCLOSURES:

The authors acknowledged receiving support from Imperial’s British Heart Foundation Centre for Excellence Award and disclosed receiving support from the British Heart Foundation, the National Institute for Health Research Imperial College Biomedical Research Centre, the EJP RD Research Mobility Fellowship, the Medical Research Council, and the Sir Jules Thorn Charitable Trust. Some authors reported receiving grants, personal fees, advisory fees, or laboratory work fees outside the submitted work.

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.