Product: Us2.v1 Company: Us2.ai
Artificial intelligence-based echocardiographic assessment for monitoring disease progression in transthyretin cardiac amyloidosis
European Journal of Heart Failure, 2025
Abstract
Aims
In transthyretin amyloid cardiomyopathy (ATTR-CM), reduced stroke volume (SV) portends a poor prognosis. Artificial intelligence (AI) enables rapid, standardized assessment of left ventricular outflow tract velocity-time integral (LVOT-VTI), which is a reliable surrogate for SV. We investigated longitudinal changes in AI-derived LVOT-VTI as outcome predictors in ATTR-CM.
Methods and results
Consecutive patients with ATTR-CM underwent baseline and 12 ± 1 month transthoracic echocardiography between 2007 and 2021. Scans were processed by an AI platform for fully automated measurements including LVOT-VTI. Changes in echocardiographic variables were related to all-cause mortality in a landmark analysis using multivariable Cox models adjusting for clinical covariates (age, sex, TTR genotype, atrial fibrillation status, New York Heart Association class and National Amyloidosis Centre stage). Time-dependent receiver-operating characteristic analysis identified the optimal threshold of LVOT-VTI change. A total of 752 patients (74 ± 9 years; 88% men; 66% wild-type) were followed for a median of 3.3 years (interquartile range 2.1–5.0 years), during which 334 (44.4%) died. Among changes in echocardiographic parameters over 12 months, only LVOT-VTI change remained independently prognostic (adjusted hazard ratio [HR] per 1% decrease 0.994, p = 0.025). A ≥5% decrease (n = 377 patients, 50%) independently predicted all-cause mortality (adjusted HR 1.41, 95% confidence interval 1.13–1.76; p = 0.003), and improved risk reclassification (integrated discrimination improvement = 0.012; continuous net reclassification improvement = 0.21, both p < 0.001).
Conclusions
A ≥5% decrease of AI-derived LVOT-VTI over 12 months, a simple indicator of SV loss, is independently associated with worse outcome in ATTR-CM. Routine monitoring of this automated AI metric may guide earlier therapeutic escalation and is a possible endpoint for future trials.
