De-identified subject data from the LAMPOON investigational device exemption trial (NCT03015194) were retrospectively evaluated under an Institutional Review Board approved protocol (4). All subjects in that trial (N = 30) were considered at prohibitive risk for TMVR-related LVOT obstruction based on a predicted end-systolic neo-LVOT area of <200 mm². Subjects underwent a valve-in-ring or valve-in-mitral annular calcification procedure with the SAPIEN 3 valve (Edwards Lifesciences, Irvine, CA, United States) with retrograde LAMPOON prior to valve implantation. The LAMPOON technique and 30-day clinical outcomes have been previously reported (2, 4). Briefly, this technique involves three main steps: (1) anterior leaflet traversal with a guidewire, followed by (2) leaflet laceration by electrifying and pulling the guidewire from base to tip, immediately followed by (3) TMVR. Complete details on the utilized retrograde technique and newer LAMPOON technique iterations were recently described (5).

Inclusion criteria for this study were: (1) the availability of a post-procedure, multi-phase CT scan with contrast-enhancement, and (2) a splayed anterior leaflet that was visible on post-procedure CT and exposed cells of the implanted transcatheter valve. Of the thirty subjects in the LAMPOON investigational device exemption trial, sixteen subjects did not have adequate visualization of the anterior mitral leaflet on the post-procedure CT scan due to image quality, and two subjects did not have an available post-procedure CT due to patient death. Of the remaining twelve subjects, four had CTs in which we could not visualize exposed cells of the transcatheter valve, despite a successful anterior leaflet laceration. Thus, eight subjects were selected for the present study who had a splayed anterior leaflet which exposed open cells of the implanted transcatheter valve, as verified on post-procedure CT. These eight subjects were at risk of a fixed LVOT obstruction from TMVR due to a small predicted neo-LVOT, and not a dynamic obstruction due to a long anterior leaflet.

Contrast-enhanced, multi-phase CT scans were acquired after the procedure, either pre-discharge or within 30-days of follow-up. CT images were reconstructed in 10% intervals over the cardiac cycle within <1.0-mm slice thickness. Mimics 20.0 (Materialize, Leuven, Belgium) image post-processing software was used to generate 3-D models of the left ventricle and aorta using previously described segmentation techniques (6). The left ventricular volume was segmented from the CT images for all phases across the cardiac cycle, and the time rate of change in volume was calculated to provide time-varying flow rate. This method for deriving flow rate from left ventricular volumes has been described previously (7).

Two different 3D models were created for each subject using the post-procedure CT datasets. The first model represented “TMVR with LAMPOON.” In this model, the actual implanted transcatheter valve and splayed anterior leaflet were reconstructed in 3D from the post-procedure CT (Figure 1, left). These reconstructions of the implanted valve therefore incorporated the actual valve deployment depth, angulation, and expansion characteristics. The positions of transcatheter valve cells exposed by LAMPOON were identified visually on the post-procedure CT, and these exposed cells were intentionally left patent in this model.

The second model represented “TMVR without LAMPOON.” In this model, the exposed cells of the transcatheter valve were artificially closed, mimicking a completely intact anterior leaflet geometry (Figure 1, right). Since the model of “TMVR with LAMPOON” mimicked the performed interventional procedure, the model of “TMVR without LAMPOON” was considered to mimic a control “virtual procedure.” Both models were designed to be anatomically identical other than the presence or absence of the splayed anterior leaflet.

Computed tomography datasets were utilized to measure the neo-LVOT and skirt neo-LVOT area. The neo-LVOT area was measured in a plane perpendicular to the LVOT and located at the narrowest point along the residual LVOT after TMVR (Figure 1, right). Following LAMPOON, the transcatheter valve skirt still protrudes into the LVOT, creating a narrowing at the transcatheter valve skirt called the “skirt” neo-LVOT (8). Figure 1 illustrates differences between the neo-LVOT and skirt neo-LVOT. The skirt neo-LVOT area was measured in a plane perpendicular to the LVOT and located at the level of the transcatheter valve skirt (Figure 1, left). Both neo-LVOT and skirt neo-LVOT area measurements were first predicted on baseline CT, and then measured on post-procedure CT.

Computational fluid dynamics (CFD) is a method used to simulate fluid flow with computer modeling. In this study, we utilized a validated CFD workflow for simulating patient-specific LVOT hemodynamics based on cardiac CT data (7). Patient-specific CFD simulations were performed for both modeled conditions (TMVR with and without LAMPOON) for each subject. The output of these simulations provided spatially varying velocity and pressure data across the entire left ventricle and aorta. The workflow used for performing the CFD simulations has been previously described and validated against clinical measurements from Doppler echocardiography (7). The average difference between CFD-derived and echocardiography-derived peak LVOT velocity measurements was <10%, indicating a good agreement between the CFD simulations and clinical measurements. This CFD workflow is briefly summarized below.

All CFD simulations were performed for the peak-systolic phase of the cardiac cycle. The cardiac phase of the post-procedure CT scan which most closely coincided with the peak systolic flow rate was selected for creating a 3D model of the left heart anatomy and implanted transcatheter valve. For each 3D model, an inlet was placed at the left ventricular apex, and an outlet was placed at the sino-tubular junction. 3-Matic 12.0 (Materialize, Leuven, Belgium) computer aided design software was used to place a virtual wall at the level of the bioprosthetic valve leaflets to prevent regurgitant flow. A volumetric mesh was then generated using ANSYS Fluent Meshing (Ansys, Inc. Canonsburg, PA, United States). This mesh included polyhedral elements with an edge length of 0.5 mm, and a prismatic boundary layer mesh (10 layers with geometric growth of 1.05) which was applied on the left ventricular wall. The total number of mesh elements per patient model was on average ∼1,800,000 elements. Transition-to-turbulence characteristics in the flow field were accounted for using the scale-adaptive simulation turbulence model. Turbulence boundary conditions included (1) hydraulic diameter (which is the inlet diameter, D_H) and (2) turbulence intensity (approximated using 0.16 × Re_D^–0.125, where Re_D is the Reynolds number based on the patient-specific flow rate and the hydraulic diameter).

The time-varying left ventricular volume curve (obtained from the post-procedure CT scan) was used to calculate a peak systolic flow rate, which was then prescribed as the inlet boundary condition. The outlet boundary condition was set to a zero-reference pressure. All structures were assumed to be rigid and non-permeable, and a no-slip boundary condition was applied at the walls. Blood was modeled as a single-phase Newtonian fluid with a density of 1060 kg/m³ and dynamic viscosity of 0.0034 poise. The 3-D Navier-Stokes equations were solved in ANSYS Fluent 19.0 using the SIMPLE scheme for pressure-velocity coupling. A bounded second-order implicit transient formulation was utilized with a time step of 0.001 s. The solution was considered converged when residuals in momentum and turbulence variables declined below 10^–4. Three-thousand time steps were simulated, with the last time step being used for data analysis.

Hemodynamic metrics were extracted from each CFD simulation, including the peak LVOT velocity (V_peak), peak LVOT pressure gradient (ΔP_peak), and peak LVOT effective orifice area (EOA_peak). To calculate ΔP_peak, the ventricular pressure was derived from an average pressure across the model inlet (LV apical plane), and the aortic pressure was derived from an average pressure across the model outlet (sino-tubular junction plane). ΔP_peak was then calculated as the difference between the ventricular and aortic pressures. EOA_peak was calculated as the ratio of the peak volumetric flow rate (Q_peak) to the peak outflow velocity (V_peak) (9).

All data analyses were performed with the MedCalc statistical software package (MedCalc, Ostend, Belgium). Simulated hemodynamics for each subject were compared between the two modeled conditions (TMVR with and without LAMPOON) using the paired samples t-test or Wilcoxon test for parametric and non-parametric data, respectively. Statistical significance was defined as p < 0.05.

Baseline, procedural and post-procedural characteristics are shown in Table 1. All subjects were considered to be at prohibitive risk of LVOT obstruction based on small ventricular anatomy and small LVOT clearance (end-systolic neo-LVOT area <200 mm² as predicted on baseline CT), and therefore were considered candidates for the LAMPOON procedure. All subjects included were female, likely due to the selection of patients with small ventricular anatomy for LAMPOON. The indication for mitral valve replacement was mitral regurgitation in three subjects, and mitral stenosis in five subjects.

All subjects underwent successful midline anterior leaflet laceration (LAMPOON procedure) immediately followed by investigational use of the SAPIEN 3 transcatheter heart valve in the mitral position. The setting of valve replacement was valve-in-ring for two subjects, and valve-in-mitral annular calcification for six subjects. Following valve implantation, all subjects had trace or no para-valvular regurgitation. The number of transcatheter valve cells exposed by the splayed anterior mitral leaflet ranged from 1 to 3 cells. The combined area of the exposed cells ranged from 41 to 88 mm², and the area of a single exposed cell was on average 36 ± 7 mm². Valve deployment depth (percent deployed in left ventricle) ranged from 39 to 85%. At follow up, the peak systolic flow rate (calculated from the post-procedure CT-derived LV volumes) ranged from 13 to 38 L/min.

Simulations of left ventricular flow patterns are depicted by color maps (red = highest velocity, blue = lowest velocity) with streamlines for a representative subject (Subject 5) (Figure 2). For the condition of the virtual control procedure (TMVR without LAMPOON), flow acceleration occurred at the level of the neo-LVOT, resulting in a narrowed outflow jet (Figure 2, left panel). In this condition, the neo-LVOT was shown to be the predominant geometric narrowing for systolic flow. All forward flow was directed through the neo-LVOT, causing an increase in velocity at this location due to the reduced flow area. Additionally, there was a region of reduced flow adjacent to the transcatheter valve due to the narrowed outflow jet, as noted in the neo-LVOT short-axis view (Figure 2, upper left image).

For the condition of the performed interventional procedure (TMVR with LAMPOON), flow occurred both through the neo-LVOT and through exposed cells of the transcatheter valve (Figure 2, right panel). LAMPOON provided an additional source of flow into the LVOT, leading to a reduction in outflow velocity due to the increased flow area. The region of reduced flow adjacent to the valve in the case of TMVR without LAMPOON was decreased in the case of TMVR with LAMPOON, thus reflecting flow closer to normal flow physiology (Figure 2, upper right image). Following TMVR with LAMPOON, the main impediment to forward flow was observed to be the transcatheter valve skirt which still projected into the native LVOT.

Simulations of left ventricular pressure fields are depicted by color maps (red = highest pressure, blue = lowest pressure) for a representative subject (Subject 5) (Figure 3). For the condition of the virtual control procedure (TMVR without LAMPOON), there was a substantial pressure drop which occurred at the location of the neo-LVOT, corresponding to the location of geometric narrowing. For the condition of the performed interventional procedure (TMVR with LAMPOON), the simulated pressure drop was reduced at the neo-LVOT due to the increase in flow area from the splayed anterior leaflet. In this subject, the splayed anterior leaflet from LAMPOON exposed three cells of the transcatheter valve, corresponding to a total increase in flow area of 87 mm².

As compared to TMVR without LAMPOON, TMVR with LAMPOON resulted in lower V_peak, lower ΔP_peak, and higher EOA_peak by 14 ±7% (p = 0.006), 31 ±17% (p = 0.01), and 17 ±9% (p = 0.002), respectively (Figure 4). Following LAMPOON, V_peak decreased from 2.6 ±1.1 m/s to 2.2 ±0.9 m/s, ΔP_peak decreased from 17.5 ±19.6 mmHg to 10.0 ± 8.8 mmHg, and EOA_peak increased from 1.6 ±0.6 cm² to 1.8 ±0.7 cm² (p < 0.05 for all comparisons). Subject 7 had a substantially larger simulated LVOT gradient than the other subjects and could be considered an outlier (Figure 4A). However, when we excluded this subject from the analysis, we observed the same significant trend for the effect of LAMPOON on simulated LVOT gradient (p = 0.02).

LAMPOON was most effective in subjects who had a small post-procedure neo-LVOT area, as measured on post-procedure CT (Figure 5). The largest hemodynamic improvement with LAMPOON was observed in Subject 7 who had the smallest post-procedure neo-LVOT area across all subjects (49 mm²). This subject experienced a 32 mmHg reduction in the simulated LVOT gradient following LAMPOON. 3D flow patterns in this subject were also shown to substantially improve following LAMPOON, as evidenced by more centrally directed flow along the LVOT (Supplementary Video 1).

As the post-procedure neo-LVOT area increased beyond 250 mm², LAMPOON began to demonstrate a smaller hemodynamic effect. For all subjects with a post-procedure neo-LVOT area greater than 250 mm², the simulated LVOT gradient reduced by less than 2 mmHg following LAMPOON. This finding is also illustrated in Figure 6 which shows simulations of flow patterns in two representative subjects with different neo-LVOT dimensions.

This was a retrospective study with a small sample size. As adoption of the LAMPOON technique expands, larger datasets will be needed to corroborate the findings of this study. The modeling of blood flow in this study using CFD may be limited given the assumptions of fixed left ventricular walls with zero wall velocity during peak systole, and steady flow boundary conditions. Since the pressure gradient across a normal aortic valve should be negligible relative to the gradient across an obstructed LVOT, we did not model an aortic valve in our simulations for simplicity. The anterior mitral leaflet was modeled depending on whether the leaflet could be reliably segmented on the post-procedure CT dataset. The posterior mitral leaflet was not modeled since this leaflet is not affected by the LAMPOON procedure and should not impact flow in the LVOT. The value of our CFD workflow is to provide a simplified, rapid approach to simulate LVOT hemodynamics. This method has been previously validated and has demonstrated good accuracy (<10% error) when compared to clinical hemodynamic measurements (7). Despite the discussed limitations, this study provides unique insight into the hemodynamic effect of LAMPOON which would have been impossible to study in a clinical setting.