Impact of Right Ventricular Surface Area-to-Volume Ratio on Ventricular Remodeling After Pulmonary Valve Replacement

Optimal reverse remodeling of the right ventricle (RV), a sentinel goal of pulmonary valve replacement (PVR) in patients with repaired tetralogy of Fallot, is not fully predicted by volume-based pre-PVR parameters. Our objectives were to characterize novel geometric RV parameters in patients receiving PVR and in controls, and to identify associations between these parameters and chamber remodeling post-PVR. Secondary analysis was performed on cardiac magnetic resonance (CMR) data from 60 patients enrolled in a randomized trial of PVR with and without surgical RV remodeling. 20 healthy age-matched subjects served as controls. The primary outcome was optimal post-PVR RV remodeling (end-diastolic volume index (EDVi) ≤ 114 ml/m2 and ejection fraction (EF) ≥ 48%) vs. suboptimal remodeling (EDVi ≥ 120 ml/m2 and EF ≤ 45%). RV geometry was markedly different at baseline in PVR patients compared with controls, with lower systolic surface area-to-volume ratio (SAVR) (1.16 ± 0.26 vs.1.44 ± 0.21 cm2/mL, p < 0.001) and lower systolic circumferential curvature (0.87 ± 0.27 vs. 1.07 ± 0.30 cm− 1, p = 0.007) but similar longitudinal curvature. In the PVR cohort, higher systolic SAVR was associated with higher RVEF both pre- and post-PVR (p < 0.001). Among PVR patients, 15 had optimal and 19 had suboptimal remodeling post-PVR. Multivariable modeling showed that among the geometric parameters, higher systolic SAVR (OR 1.68 per 0.1 cm2/mL increase; p = 0.049) and shorter systolic RV long-axis length (OR 0.92 per 0.1 cm increase; p = 0.035) were independently associated with optimal remodeling. Compared with controls, PVR patients have lower SAVR and lower circumferential but not longitudinal curvature. Higher pre-PVR systolic SAVR is associated with optimal remodeling post-PVR.


Introduction
Surgical repair of tetralogy of Fallot (rTOF) and similar physiology can be accomplished with very low early morbidity and mortality [1]. While the childhood years are relatively uneventful, these patients are prone to complications such as arrhythmias, low exercise tolerance, heart failure, and early death in their adult years [2][3][4]. Dysfunction of the as mortality and sustained ventricular tachycardia showed that while certain traditional volumetric measures such as RV EDVi were not associated with these important clinical outcomes [8], RV dysfunction was independently associated with poor clinical outcomes after PVR [9]. Therefore, it is still unclear what parameters of the RV should be used to guide when a patient should be referred for a PVR in order to optimize the likelihood of a favorable clinical outcome.
It has long been recognized that the geometric shape of the RV influences its mechanical efficiency. This has been demonstrated in studies of LV shape, and its impact on functional outcomes [10][11][12]. The RV, however, poses a more difficult challenge, due to its complex shape even in the normal heart, and a paucity of data regarding abnormalities in its shape in disease states such as rTOF. Some authors have described these abnormalities using 3D-echocardiography [13] and cardiac magnetic resonance (CMR) [14][15][16]. However, these studies focused on RV shape at a single time point, did not evaluate changes in RV geometry following PVR, and did not examine pre-PVR markers of adverse RV remodeling after PVR. RV geometric parameters such as local wall curvatures, surface area, and surface areato-volume ratio can be assessed using CMR as previously described by Tang et al. [17]. Accordingly, the present study was designed to (i) describe the changes in parameters of RV geometry following PVR, and (ii) identify associations between pre-PVR geometric parameters and post-PVR RV remodeling.

Study Design
We conducted a secondary analysis of aggregate CMR data from patients enrolled in a prospective single center, unblinded, randomized clinical trial of PVR with and without surgical remodeling of the RVOT [18]. The present study was approved by the Institutional Review Board at Boston Children's Hospital. The pre-PVR CMR and the latest available post-PVR CMR study prior to any reintervention were used for analysis.

Subjects
Of the 64 patients enrolled in the surgical RV remodeling trial, 60 qualified for the current study. The rationale for choosing this cohort was the availability of extensive, prospectively collected clinical, laboratory, and imaging data before and after PVR. Eligibility criteria for the trial included rTOF or similar physiology with chronic pulmonary regurgitation (PR) measured by CMR with a PR fraction of ≥ 25% and 2 or more of the following: [1] RV EDVi ≥ 160 mL/m 2 ; [2] RV end-systolic volume (ESV) index (ESVi) ≥ 70 mL/m 2 ; [3] RV EF ≤ 45%; [4] LV EDVi ≤ 65 mL/m 2 ; [5] RV outflow aneurysm (defined as focal dilation with akinetic or dyskinetic wall motion); or [6] clinical criteria (e.g., exercise intolerance, symptoms and signs of heart failure, or cardiac medications). Exclusion criteria for the trial included [1] severe RVOT obstruction (defined as peak systolic gradient ≥ 60 mm Hg by cardiac catheterization); [2] severe RV hypertension at systemic or higher level; [3] additional sources of RV volume overload other than pulmonary or tricuspid valve regurgitation (e.g., partially anomalous pulmonary venous connection); and [4] contraindications to preoperative CMR.
Given the novel nature of the RV shape parameters evaluated in this study, we analyzed RV shape parameters in 20 age-matched patients with no heart disease who underwent CMR in our laboratory. Examples of indication for CMR included suspected structural anomaly (e.g., anomalous pulmonary venous connection or aberrant coronary origin), where the intracardiac anatomy and function were found to be normal. The rationale for including a control group was to provide a frame of reference for the values measured in the trial cohort.

Data Acquisition and Modeling
The imaging techniques and volumetric measurements employed in this study have been previously published [19]. Briefly, CMR scans were performed on a 1.5 Tesla scanner using a standardized imaging protocol [4]. The RV and LV were imaged using ECG-gated, breath-held steady state free precession cine sequences in the ventricular long-and shortaxis planes (12-14 equidistant slices covering the ventricles from base to apex; slice thickness 6-8 mm; interslice gap 0-2 mm; 30 frames per cardiac cycle). Three-dimensional RV/LV geometry and computational mesh were constructed following the procedures described previously [20] (Fig. 1).

Measures of RV Geometry
We measured traditional volumetric parameters of both ventricles, namely EDV, ESV, and EF as well as RV and LV mass as previously described by our center [21]. The methodology of measuring the novel parameters of RV geometry were described by Tang et al. [17] and are included in the online addendum. These included: [1] Local wall curvatures (longitudinal and circumferential); [2] wall thickness; [3] maximum RV long-axis length; [4] surface area of the inner (endocardial) surface of the ventricular mesh model; and [5] calculated RV surface area-to-volume ratio (SAVR).
All measurements and ratios were recorded in end-diastole (maximal volume) and in peak systole (minimal volume).

Outcomes
The primary outcome was optimal vs. suboptimal post-PVR RV reverse remodeling as defined in the original clinical trial [18]. Optimal reverse remodeling was defined as RV EDVi ≤ 114 mL/m 2 and RV EF ≥ 48% (normal RV size and global systolic function); suboptimal reverse remodeling was defined as RV EDVi ≥ 120 mL/m 2 and RV EF ≤ 45% (at least mild RV dilation or dysfunction). As these bounds are non-contiguous, a group of 'intermediate' patients fulfilling neither the optimal nor suboptimal definitions remained.

Statistical Analysis
Descriptive statistics include mean ± standard deviation (SD) and median with interquartile range (IQR) for continuous variables, as appropriate. Categorical data are described as frequency with percentage. Baseline demographics of controls and PVR patients, and baseline clinical characteristics of the PVR patients stratified by outcome status were compared. A Fisher exact test was used for categorical variables, the Wilcoxon rank sum test for ordinal variables and for continuous variables that were not normally distributed, and Student's two-sample t test for normally distributed continuous variables. ANOVA and the Kruskal-Wallis test were used for comparisons among multiple groups. Logistic regression was used to estimate univariate associations between predictors and outcomes in the PVR patients. Continuous CMR parameters were also categorized according to data tertile to examine potential nonlinear effects. Tertile variables that were significant in multivariable modeling were collapsed into two categories where indicated to achieve parsimony. Classification and regression tree analysis (CART) was employed to confirm or refine the binary threshold for such predictors. Since time from PVR to the post-PVR CMR (ranging from 4 months to 3.9 years) could be a confounding factor, we first determined if this time interval was correlated with both the predictors and the outcome. Since no significant correlation was found, it was therefore not included as a covariate. Multivariable logistic regression was utilized to identify

Changes in RV Geometry with PVR
The follow-up CMR used to assess RV reverse remodeling was performed 1.7 ± 1.1 years after PVR. RV reverse remodeling was variable, with 15 patients (25%) experiencing optimal reverse remodeling, 19 suboptimal reverse remodeling (32%), and 26 intermediate reverse remodeling (43%). Table 4 summarizes changes in RV geometry after independent predictors of optimal vs. suboptimal outcome for the PVR patients. A p-value of < 0.05 was considered significant for all comparisons. Analyses were performed with SAS version 9.4 (SAS Institute, Inc., Cary, NC) and R version 3.5.2.

Results
There were no differences between the demographic characteristics of the PVR and control groups ( Table 1). Table 2 summarizes the patient characteristics of the PVR group, stratified by optimal versus suboptimal post-PVR outcome, as well as the 'intermediate' patients who did not meet either of the defined outcomes. There were no differences among groups in the baseline demographics, surgical and

Predictors of Outcome
The preoperative variables associated with an optimal postoperative outcome in univariate analysis are summarized in Table 5. In addition to traditional volume-based parameters, higher SAVR in systole was predictive of optimal RV reverse remodeling (odds ratio were independent predictors of optimal reverse remodeling (c-statistic = 0.84, Table 5). Importantly, correlation analysis demonstrated a significant positive association between RV EF and systolic SAVR (Fig. 2).

Relationship Between SAVR and Ventricular Volume
To explore the hypothesis that the association between higher SAVR and optimal RV reverse remodeling reflects that patients with higher SAVR had smaller RV volumes, we controlled for RV EDVi by including it as a covariate in the novel-parameter multivariable model described in Table 5. The analysis showed that the effect size (odds ratio) for systolic SAVR diminished by less than 20% (from 1.68 to 1.41), but was no longer significant (p = 0.24) due to the correlation between SAVR and RV EDVi.

Discussion
Amongst CMR-derived novel parameters of RV geometry, we identified higher RV systolic SAVR and shorter systolic RV long-axis length as independent predictors of optimal RV reverse remodeling following PVR. Several studies have tried to identify predictors of optimal RV reverse remodeling following PVR and showed that pre-PVR RV volumetric parameters were useful markers of early post-PVR RV reverse remodeling [5,7]. However, with longer follow-up, it was found that the early volumetric improvements were not maintained [22]. More concerning was the finding that RV volumetric parameters did not predict an improvement PVR stratified by outcome groups. RV volumes and RV surface area decreased significantly with a commensurate increase in systolic and diastolic SAVR (SAVR in diastole 0.84 ± 0.12 to 1.04 ± 0.16 cm 2 /mL, p < 0.001; SAVR in systole 1.16 ± 0.26 to 1.28 ± 0.31 cm 2 /mL, p < 0.001).
Patients who experienced optimal reverse remodeling had a higher diastolic SAVR at baseline and had a more pronounced increase in diastolic SAVR post-PVR compared to those who had suboptimal reverse remodeling (change in diastolic SAVR 0.24 ± 0.10 cm 2 /mL vs. 0.15 ± 0.11 cm 2 /mL, p = 0.026).
A pre-PVR systolic SAVR value of 1.03 discriminated between patients with and without optimal reverse remodeling. Of those with pre-PVR systolic SAVR ≥ 1.03, 72% (13/18) experienced optimal reverse remodeling whereas ESVi, indexed end-systolic volume; EF, ejection fraction LV, left ventricle; SAVR, surface area-to-volume ratio; C-curvature, circumferential curvature; L-curvature, longitudinal curvature RV shape but importantly also their changes after PVR, and their contribution to prediction of RV reverse remodeling following PVR.
The shape of a pumping chamber affects its mechanical efficiency. This has been well demonstrated for the relation between LV shape and its systolic function by both theoretical [11] and clinical analyses [10,12]. Similar analyses for the RV have been difficult to perform due to its more complex shape. CMR combined with computational mesh analysis provides a unique window into RV geometry in health and disease. We found that the C-curvature of the RV in key clinical outcomes such as death or sustained ventricular tachycardia [9]. Thus, RV volumetric assessment alone cannot explain the variance in outcomes following PVR. A more nuanced understanding of the RV geometry is therefore needed to better explain the observed variance. Attempts have been made to describe the geometry of the RV beyond volumetry. These have focused on analysis of RV wall curvatures [13,14,16], and localized strain [16], as well as vorticity, and a principal component analysis approach toward shape analysis [23][24][25][26]. In the present study, we sought to investigate not just the abnormalities of  Table 1 * P < 0.001; ^ P < 0.05 for the comparison of Post-PVR vs. Pre-PVR.
PVR, pulmonary valve replacement; repaired tetralogy of Fallot; RV, right ventricle; EDVi, indexed end-diastolic volume; ESVi, indexed endsystolic volume; EF, ejection fraction; LV, left ventricle; SAVR, surface area-to-volume ratio 0.92 (0.86, 0.99) 0.035 Variables listed in the univariate analysis section are those that had a significant association with the outcome. Measures not significantly associated with optimal outcome by univariate analysis are listed in an online Supplemental Table 2. Multivariable model 1 utilized all predictor parameters. Multivariable model 2 used as candidates only the novel RV geometric measures OR, odds ratio; PVR, repaired tetralogy of Fallot; RV, right ventricle; EDVi, indexed end-diastolic volume; ESVi, indexed end-systolic volume; EF, ejection fraction LV, left ventricle; SAVR, surface area-to-volume ratio distended and globular RV in the short-axis (i.e., free wall to septum) dimension, but not as much distension and flattening occurs in the long-axis (i.e., base-to-apex) dimension.
We also describe a novel parameter-the ratio of ventricular surface area to ventricular volume (i.e., the surface area-to-volume ratio or SAVR). A sphere is the shape with in PVR patients was significantly lower than in controls. A lower curvature means a more gradual shape change in the chosen direction (Fig. 3). Interestingly, the L-curvature did not significantly differ between PVR patients and controls. This suggests that the adverse remodeling seen in PVR patients subjected to chronic volume loading leads to a more

Limitations
In order to study the differences between the most optimal patients (no RV dilation and no RV dysfunction postoperatively) and suboptimal patients (those that have both RV dilation and RV dysfunction post-operatively), as well as to retain consistency with the outcome definitions from the original trial, our primary outcome analysis excluded a group of patients with intermediate post-operative outcomes. This may have decreased the power of the study to detect smaller differences. Additionally, the cohort of PVR patients was restricted to those who did not have contraindications to CMR. This restriction excluded patients with pacemakers or defibrillators, which may lead to selection bias. Thus, the findings of this study may not be generalizable to such patients. We did not perform reproducibility analyses on the novel parameters measured in this study. The study included several non-rTOF PVR patients. This may lead to heterogeneity in RV geometry due to slightly variant native anatomies and size/location of VSD patches, although the indications for PVR were uniform in the entire cohort. The patients included in the original clinical trial from which these data are drawn were recruited between the years 2004-2008. At that time, the RV volume threshold for referral for PVR was higher than it is today. Resultantly, baseline RV EDV is significantly higher in these patients compared to contemporary patients being referred for PVR.

Conclusions
This study identified novel markers of RV geometry and characterized adverse remodeling in PVR patients as compared with controls. Patients had a more globular RV shape in the short-axis plane and a lower mean SAVR compared with controls. When restricted to novel parameters, systolic SAVR, and lower systolic RV long-axis length were associated with optimal reverse remodeling after PVR. These insights into the characteristics of normal RV shape and its alterations in rTOF may lead to refinement of referral indications for PVR as well as surgical RV remodeling strategies during PVR.
Acknowledgements We thank Dr. Heng Zuo for his help with data processing and Ms. Kai-Ou Tang for artwork.
Author Contributions Nikhil Thatte participated in the study design, data analysis, and wrote the first draft of the manuscript. Lynn Sleeper participated in the study design, data analysis, and provided critical revisions to the manuscript. Minmin Lu participated in the statistical lowest SAVR for a given volume of fluid and is therefore the most efficient way to contain that fluid within the least amount of area of the covering surface. SAVR is seen in nature by the spontaneous formation of spherical shapes such as soap bubbles. While this is efficient for a structure designed to be a container, it is not ideal for a pump. The walls of the ventricular pump provide the motive force to the contained fluid (blood) and, therefore, it is more likely to be beneficial to have an increased SAVR to generate optimal pump function. The positive correlation we found in this study between RV EF and SAVR (Fig. 2) lends support to this notion. We postulate that an RV with a lower SAVR represents a more spherical ventricle and potentially a less efficient pump. This is easier to visualize for the LV with the poorly functional, globular shaped chamber in dilated cardiomyopathy being an example. While the diseased RV does not assume a completely globular shape, our study shows that the adverse remodeling of the RV in PVR has the effect of creating a shape with a low C-curvature and low SAVR.
The original clinical trial from which these data were drawn compared the impact on RV EF of RV remodeling surgery [18]. The primary remodeling strategy was an extended longitudinal excision of RVOT scar beyond the previously placed patch. The trial had a negative result. The authors postulated that, amongst other factors, targeted modification of RV geometry may have been necessary in order to obtain a benefit from remodeling surgery. They noted that a predetermined geometric goal was not established while undertaking the remodeling procedure, due to paucity of understanding of the baseline geometry and desired geometric outcome. Our study may inform both the pre-PVR assessment of RV geometry in PVR and establish targets for surgical remodeling of the RV during PVR. Specifically, tailoring RV remodeling toward creating a ventricle that is more tightly curved in the circumferential dimension (higher C-curvature) and has higher SAVR as the geometric goal of RV remodeling during PVR may lead to improved postoperative RV size and function. One approach toward this type of surgical remodeling is the use of "contracting bands" which serve to tighten the free wallto-septum dimension of the RV, effectively increasing the C-curvature (making the ventricle more tightly curved in the circumferential dimension) [27].
This study focused on post-PVR ventricular remodeling rather than clinical outcomes. Future research is warranted to explore the impact of these pre-PVR RV geometric parameters on post-PVR clinical outcomes, and to analyze the associations between geometric parameters and novel RV function metrics such as strain and energetics biomarkers.