Prevalence of low-flow state by automated machine-learning 3D echocardiography in patients with moderate-to-severe aortic stenosis and normal ejection-fraction

Background: In aortic valve stenosis (AS), measurement of stroke volume index (SVi) by the 2D Doppler- technique is required to calculate the aortic valve area (AVA) and determine flow status but is prone to systematic errors. Purpose: To investigate the prevalence of low-flow (LF) state (SVi ≤ 35 ml/m2) in patients with AS and concurrent SVi quantification by a validated 3D left ventricular volumetric method and standard 2D Doppler methods and its potential repercussions on flow status reclassification. Methods: Consecutive patients with moderate or severe AS (≤1.5 cm2 by continuity equation) and ejection fraction ≥ 50% underwent concurrent Dynamic Heart Model (DHM) evaluation with larger settings of the boundary detection sliders (end-diastolic position = 60/60; end-systolic position = 30/30). Results: We included 57 patients (median 78 years [70-85]); 40% were women. The mean AVA was 1.03 ± 0.37 cm2, median 1.00 [0.73-1.20] cm2. We found a significant but modest correlation between SVi assessed by DHM and 2D Doppler (Pearson corr.=0.48, p < 0.001). In the 25 patients with severe AS (AVA < 1 cm2) subgroup, 75% of patients categorized as LF by 2D Doppler were reclassified as normal flow by DHM. The observed proportion of overall agreement in these patients was 60% (Cohen's kappa = 0.178, p = 0.238), without significant correlation (Pearson coeff.: 0.358; p = 0.166). Conclusions: In patients with moderate or severe AS and normal EF, the prevalence of LF state defined as SVi ≤ 35 ml/m2 may be significantly lower than previously reported when the LV volumetric metrics by DHM are employed, and the AVA is close to the replacement referral cut-off.