Transcatheter mitral valve replacement (TMVR) is emerging in the context of annular calcification (valve-in-MAC; ViMAC), failing surgical mitral annuloplasty (mitral-valve-in-ring; MViR) and failing mitral bioprosthesis (mitral-valve-in-valve; MViV). A notorious risk of TMVR is neo left ventricular outflow tract (neo-LVOT) obstruction. Three-dimensional computational models (3DCM) are derived from multi-slice computed tomography (MSCT) and aim to predict neo-LVOT area after TMVR. Little is known about the accuracy of these neo-LVOT predictions for various mitral phenotypes. Preprocedural 3DCMs were created for ViMAC, MViR and MViV cases. Throughout the cardiac cycle, neo-LVOT dimensions were semi-automatically calculated on the 3DCMs. We compared the predicted neo-LVOT area on the preprocedural 3DCM with the actual neo-LVOT as measured on the post-procedural MSCT. Across 12 TMVR cases and examining 20%-70% of the cardiac phase, the mean difference between predicted and post-TMVR neo-LVOT area was -23 ± 28 mm2 for MViR, -21 ± 34 mm2 for MViV and -73 ± 61 mm2 for ViMAC. The mean intra-class correlation coefficient for absolute agreement between predicted and post-procedural neo-LVOT area (throughout the whole cardiac cycle) was 0.89 (95% CI 0.82-0.94, p < 0.001) for MViR, 0.81 (95% CI 0.62-0.89, p < 0.001) for MViV, and 0.41 (95% CI 0.12-0.58, p = 0.002) for ViMAC. Three-dimensional computational models accurately predict neo-LVOT dimensions post TMVR in MViR and MViV but not in ViMAC. Further research should incorporate device host interactions and the effect of changing hemodynamics in these simulations to enhance accuracy in all mitral phenotypes.
Read full abstract