Predicting paravalvular leak after transcatheter mitral valve replacement using commercially available software modeling

• Published in: Journal of cardiovascular computed tomography Vol. 14; no. 6; pp. 495 - 499
• Main Authors: Morris, Michael F., Pena, Alejandro, Kalya, Aneesh, Sawant, Abhishek C., Lotun, Kapildeo, Byrne, Timothy, Fang, H. Kenith, Pershad, Ashish
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.11.2020
• Subjects: Aged; Aged, 80 and over; Arizona; Cardiac Catheterization - adverse effects; Computed tomography; Computed Tomography Angiography; Coronary Angiography; Female; Heart Valve Prosthesis Implantation - adverse effects; Humans; Male; Middle Aged; Mitral regurgitation; Mitral Valve - diagnostic imaging; Mitral Valve - physiopathology; Mitral Valve - surgery; Mitral Valve Insufficiency - diagnostic imaging; Mitral Valve Insufficiency - etiology; Mitral Valve Insufficiency - physiopathology; Paravalvular leak; Patient-Specific Modeling; Predictive Value of Tests; Retrospective Studies; Risk Assessment; Risk Factors; Severity of Illness Index; Software; Software modeling; Treatment Outcome; Arizona; Mitral regurgitation; Computed tomography; Paravalvular leak; Software modeling
• DOI: 10.1016/j.jcct.2020.04.007
• ISSN: 1934-5925; 1876-861X

There is limited data identifying patients at risk for significant mitral regurgitation (MR) after transcatheter mitral valve replacement (TMVR). We hypothesized that software modeling based on computed tomography angiography (CTA) can predict the risk of moderate or severe MR after TMVR. 58 consecutive patients underwent TMVR at two institutions, including 31 valve-in-valve, 16 valve-in-ring, and 11 valve-in-mitral annular calcification. 12 (20%) patients developed moderate or severe MR due to paravalvular leak (PVL). The software model correctly predicted 8 (67%) patients with significant PVL, resulting in sensitivity of 67%, specificity 96%, positive predictive value 89%, and negative predictive value 86%. There was excellent agreement between CTA readers using software modeling to predict PVL (kappa 0.92; p < 0.01). On univariate analysis, CTA predictors of moderate or severe PVL included presence of a gap between the virtual valve and mitral annulus on the software model (OR 48; p < 0.01), mitral annular area (OR 1.02; p 0.01), and % valve oversizing (OR 0.9; p 0.01). On multivariate analysis, only presence of a gap on the software model remained significant (OR 36.8; p < 0.01). Software modeling using pre-procedural CTA is a straightforward method for predicting the risk of moderate and severe MR due to PVL after TMVR.