By Lynda Williams, medwireNews Reporter
medwireNews: Research suggests that an electrocardiography (ECG)-based artificial intelligence risk estimator for hypertension (AIRE-HTN) may help to identify patients with an increased risk of developing hypertension and its related adverse outcomes.
Arunashis Sau (Imperial College London, UK) and co-workers used a convolutional neural network technique to create the AIRE-HTN based on 1,163,401 ECGs from 189,539 patients (52.1% women, mean age 57.7 years) who attended the Beth Israel Deaconess Medical Center in Boston, Massachusetts, USA.
The model was first trained to identify incident hypertension, defined as a hypertension diagnosis at baseline or within 30 days of ECG, and then to predict incident hypertension, based on an AIRE-HTN score where a higher value is associated with a greater risk over the following 5 years.
Overall, 32.2% of the 19,423 patients without hypertension at baseline were diagnosed with hypertension over an average follow-up of 6.8 years. Using their first outpatient ECG information, the model predicted incident hypertension with a C index of 0.70, where 0.5 equals a random prediction and 1.0 a perfect prediction.
The AIRE-HTN was applied to a validation cohort of 65,610 UK Biobank participants (51.5% women, mean age 65.4 years) with ECGs, 4.3% of whom were diagnosed with incident hypertension over an average of 4.0 years. When 35,806 of these participants were assessed by AIRE-HTN for risk of incident hypertension, the C index was “maintained” at 0.7, the authors report in JAMA Cardiology.
Sensitivity analysis confirmed that the model was effective when assessing hypertension risk in the validation cohort patients with normal left ventricular (LV) mass by cardiac magnetic resonance imaging (C index=0.70), or with a blood pressure (BP) below 120/80 mmHg without use of antihypertensive medications (C index=0.71).
The AIRE-HTN was “significantly additive to existing clinical markers in predicting incident hypertension”, the researchers say, with a C index of 0.75 versus a C index of 0.73 based on age, sex, BP, smoking status, diabetes and ethnicity.
The AIRE-HTN was also able to independently predict among patients with or without pre-existing cardiovascular or kidney disease the likelihood of cardiovascular death (hazard ratio [HR]=2.24 for each standard deviation increase in AIRE-HTN score), heart failure (HR=2.60), myocardial infarction (HR=3.13), ischaemic stroke (1.23), intracranial haemorrhage (HR=1.25) and chronic kidney disease (HR=1.89), after adjusting for clinical markers.
Furthermore, analysis found that the AIRE-HTN score for predicting adverse outcomes was significantly and positively associated with key patient characteristics including LV wall thickness and mass, diastolic function and filling pressure, aortic dimensions, carotid intima-media thickness, BP, arterial stiffness and adiposity. Peak heart rate during exercise and birthweight were negatively associated with the AIRE-HTN score.
“The consistent model performance in a group of individuals with a normal baseline BP and no LV [hypertrophy] suggests that AIRE-HTN may be identifying the latent biology of hypertension pathogenesis and not just simply identify individuals with milder hypertension at baseline”, suggest Sau and co-workers.
They conclude: “Using this model to predict incident hypertension and its link to cardiovascular outcomes could help identify at-risk patients for whom more active surveillance and lifestyle interventions might be recommended at an earlier stage in the life course.”
News stories are provided by medwireNews, which is an independent medical news service provided by Springer Healthcare Ltd. © 2025 Springer Healthcare Ltd, part of the Springer Nature Group
JAMA Cardiol 2025; doi:10.1001/jamacardio.2024.4796