Sex-based differences in left ventricular remodeling in patients with chronic aortic regurgitation: a multi-modality study

This is a single center retrospective cohort study of consecutive patients with at least moderate chronic AR by echocardiography, who underwent CMR within 90 days of the qualifying echocardiogram between December 1, 2005 and December 3, 2019. Patients with an LVEF < 50%, acute AR, concomitant valvular lesions moderate or greater in severity, type A dissection or aortic valve endocarditis, congenital heart disease, prior cardiac surgery, hypertrophic cardiomyopathy, cardiac amyloidosis/sarcoidosis, pericardial disease, radiation heart disease, pulmonary hypertension, restrictive/dilated/ischemic cardiomyopathy were excluded. Patients with inadequate views on echocardiography for assessment of AR or LV dimensions/volume were excluded. Clinical data were entered into the electronic medical record prospectively and extracted for analysis. This study was approved by our Institutional Review Board and granted a waiver for informed consent.

Comprehensive echocardiographic examination was performed for all patients using commercially available machines (Phillips Healthcare, Best, the Netherlands; Siemens Healthineers, Erlangen, Germany; or General Electric Healthcare, Chicago, Illinois, USA). LV dimensions (wall thickness, volumes and mass) were measured and calculated based on the 2015 American Society of Echocardiography (ASE) guidelines [15]. Cardiac chamber sizes were indexed to both body surface area (BSA) and height [16]. AR severity rating was stratified as mild, moderate, or severe in accordance with the ASE guidelines [17]. Additionally a hierarchical, multi-parametric method was employed, as part of the extensive efforts within our echocardiographic laboratory to decrease interobserver variability [5] (Fig. 1a).

CMR was performed based on institutional protocols (1.5 T Avanto, Siemens Healthineers, Erlangen, Germany, from 2005 to 2008, or 1.5 T Achieva or 3 T Ingenia (Phillips Healthcare, Best, the Nethernands, from 2009 to 2015). Scan parameters for balanced steady state (bSSFP) imaging on the 1.5 T scanner are as follows: TR 2.7 ms, TE 1.4 ms, acquired matrix 180 × 183, flip angle 70°, temporal resolution 30 ms. Phase contrast imaging parameters for flow quantification on the 1.5 T scanner are as follows: TR 4.8 ms, TE 2.9 ms, acquired matrix 128 × 120, flip angle 12°, temporal resolution 39 ms. Scan parameters for the 3 T scanner are as follows: bSSFP imaging: TR 2.8 ms, TE 1.4 ms, acquired matrix 192 × 177, flip angle 45°, temporal resolution 53 ms. Phase contrast imaging parameters for flow quantification on the 3 T scanner are as follows: TR 4.7 ms, TE 2.7 ms, acquired matrix 128 × 110, flip angle 12°, temporal resolution 47 ms. Epicardial and endocardial borders were traced at end-diastole and end-systole, excluding the papillary muscles, from the short axis stack bSSFP cine images to determine LV volumes and LVEF. Sphericity index was calculated as the ratio of LV end-diastolic volume to the volume of a sphere with diameter equal to the LV length measured at end-diastole in the 4 chamber view, with values closer to 1 indicating greater sphericity. Relative wall thickness (RWT) was calculated as (septal wall thickness + posterior wall thickness)/LV end-diastolic dimension) [18]. Flow quantification of AR was assessed utilizing phase contrast velocity encoded imaging acquired at the mid-ascending aorta at the level of the right pulmonary artery. This was based on prior investigations demonstrating no significant difference in regurgitant volume (RV) and fraction (RF) performed at the sinotubular junction and mid ascending aorta in a cohort which included ~ 40% non-tricuspid aortic valves and eccentric AR [19]. The contour of the aorta was traced on phase images throughout the cardiac cycle to measure antegrade and retrograde flow, which were then used to calculate RV and RF. At the time of image acquisition our lab defined severity of AR as the following: mild as a RF =< 20%, moderate as a RF of 21–39%, and severe as a RF >= 40% [20, 21]. Phase-contrast flow data and LV mass and volumes were measured, as previously described [21]. Post-processing analysis was performed with cvi42 (Circle Cardiovascular Imaging, Calgary, Alberta, Canada). The presence of holodiastolic flow reversal (HFR) in the descending thoracic aorta was also recorded and defined as persistent flow reversal ≥ 10 ml/s throughout diastole [22] (Fig. 1b).

Clinical outcomes of aortic valve surgery (repair or replacement), and death were assessed for all patients until September 30, 2018. Outcomes were assessed by review of electronic medical records, or by phone call for patients not seen within our system in the prior three months using a script approved by our Institutional Review Board with two attempts at contact. Patients lost to follow up were censored at the time of last known contact.

Continuous variables are expressed as mean ± standard deviation and categorical variables are expressed as percent. Differences in LV volumes indexed by both BSA and height, based on aortic valve morphology and sex. Non-linear regression models were built to assess the effect of RF on LV dimensions and volumes based on sex. Bland–Altman plots were constructed to assess differences in measurements between echocardiography and CMR by sex. A multivariable linear regression model was constructed to determine independent predictors of LV remodeling. Linear and nonlinear effects were considered for each continuous predictor and two-way interactions with gender were included. Sensitivity/specificity, and Youden’s index, were reported for various cut-points of CMR predictors. Statistical analyses were performed using SAS (version 9.4, SAS Institute, Cary, NC, USA). Significance was based on a p-value of < 0.05.

During the follow up period, 156 patients (121 men and 35 women) underwent aortic valve surgery. Severe AR was the primary indication for aortic valve surgery in 119 patients (97 men and 22 women). Mean time from the CMR to date of surgery was 321 days, 62.2% within 6 months, with no difference between men and women. Adjusting for age, HR, RF, and aortic valve type, women who underwent aortic valve surgery with a primary indication of severe AR had significantly smaller LVEDVI (women: 91.3 ± 27.6 ml/m² vs men: 143.2 ± 47.6 ml/m², p = 0.003), LVESVI (women: 37.8 ± 14.1 ml/m² vs men: 64.1 ± 30.4 ml/m², p = 0.041), and LV length (women: 8.2 ± 0.8 mm vs men: 9.9 ± 1.2 mm, p = 0.001), but LVEDDI and LVESDI by CMR were similar: women: 2.9 ± 0.4 cm/m² vs men: 3.1 ± 0.6 cm/m², (p = 0.754), and women: 1.8 ± 0.4 cm/m² vs men: 2.1 ± 0.6 cm/m², (p = 0.7 and 0.48, respectively).

Our study demonstrates significant sex-based differences in LV remodeling, and the impact of sex on the accuracy of echocardiography to quantify LV remodeling in the presence of significant chronic AR. One prior study examining differences in sex-based LV remodeling by CMR in AR also demonstrated limited compensatory LV dilation in women with chronic AR compared to men [23]. However, a significant proportion of patients within this cohort had moderate or greater aortic stenosis (27%) or mitral regurgitation (23%) confounding the ability to assess the impact of isolated AR on LV remodeling. Our current study, builds upon Kammerlander et al.’s findings, and further defines sex differences in phenotypic geometrical differences in LV remodeling and sex-based differences in LVESVDI thresholds and echocardiographic LVESDI margins of error. Given the improved accuracy of CMR quantification of LVR, CMR evaluation in our study defined a more marked increase in LVEDVI and LVESVI for both men and women in the setting of increasing AR severity, when compared to echocardiography. Additionally LV length increased with increasing AR severity, though the rate of increase was significantly higher in men. However, upon further comprehensive assessment of LV remodeling beyond LV volumes by CMR, our findings demonstrate that evaluation of LV remodeling is incompletely described by LV diameters or LV volumes in isolation. In our study, the integration of LV diameters, LV length, wall thickness, LV volumes, and LV sphericity provided a comprehensive assessment which elucidated distinct sex-based phenotypes of LV remodeling in response to chronic AR.

While previous research have also demonstrated sex-based differences in LVR in response to increased afterload, such as severe aortic stenosis, there are limited data regarding phenotypic differences in remodeling in response to AR, based on sex. Women in our study demonstrated significantly smaller CMR derived LV volumes, even when indexed to BSA. This significant difference was seen at all levels of AR severity, and women demonstrated a blunted increase in LV volumes with increasing AR severity. Furthermore, women in our study surprisingly demonstrated larger LVEDDI after adjusting for age, blood pressure, RF, and aortic valve morphology, while demonstrating significantly smaller LV length. Furthermore, when CMR measurements were indexed to height, rather than BSA, indexed LV diameters were significantly larger in women compared to men, despite having significantly smaller indexed LV volumes.

Additionally men had consistently higher sphericity volume index than women after adjusting for relative wall thickness. These findings suggests that male LV remodeling is more consistent with a spherical phenotype and female LV remodeling is more consistent with a conical phenotype.

Similar differences in phenotypic geometric remodeling have previously been described in an animal model of chronic overload [24]. Ashikaga et al. demonstrated that chronic volume overload resulted in conical LV geometry in the early phase with an isolated increase in basal diameters, and that LV remodeling progressed to a more spherical geometry in the late phase, with increases in both apical and basal dimensions. Thus, our findings suggest that women may demonstrate a phenotype that is more consistent with early remodeling, with a blunted response in the ability for further dilation (Fig. 6). We postulate that this blunted compensatory response in further LV dilation may result in increased wall stress in the setting of chronic volume overload, and may explain the presence of more significant symptoms in the female patients included in our study. Further mechanistic studies are needed to determine if the blunted response to LV compensation/remodeling and differences in LV geometry in the setting of significant AR result in increased LV wall stress, diastolic dysfunction and, consequently, symptoms.

Previous studies have demonstrated sex differences in referral patterns for surgical aortic valve intervention, such that men tend to be referred for aortic valve surgery due to a dilated LV or depressed LVEF, whereas women tend to be referred in the setting of symptoms and often at an older age than their male counterparts [6]. Our study is the first to demonstrate that women are unlikely to develop similar extent and pattern of LV dilation in response to chronic AR. The current guideline recommendations for referring asymptomatic patients with chronic AR include LVESD > 5.0 cm or LVEDD > 6.5 cm/ > 7.0 cm based on echocardiographic data [3, 25]. Recently, Nesius et al. demonstrated differences in CMR vs echocardiographic measurements of LV diameters, such that echocardiography underestimated CMR derived LVEDD and LVESD by 6.6 mm (p < 0.001, CI 5.8–7.7) and 5.9 mm (p < 0.001, CI 4.1–7.6), respectively [26]. Our data confirms these findings of systematic underestimation of LV diameters by echocardiography, but further demonstrates significant sex differences in accuracy of LV diameter based on imaging modality. In our study, differences in LV volumes and dimensions were accentuated in the setting of increasing volumes and dimensions, and there was also a significant interaction between magnitude of LV dimensions and sex such that the magnitude of the difference in measurements in LVEDDI and LVESDI between echocardiography and CMR increased more markedly for females than males with larger LV dimensions. The etiology for the increased error in echocardiographic measurements in women is likely multifactorial. Figure 7 illustrates factors that may impact accuracy of LV diameter measurements. This figure presents the comparison of the basal slice in 2 patients with severe AR and resultant severe LV dilation, and demonstrates that the female pattern of dilation resulted in an ellipsoid shape in the basal slice. As a result, potential echocardiographic acquisition of the parasternal long/LV outflow tract image (i.e. just above or below the dotted line) can result in significant variation in measurements. On the other hand, the short axis basal slice in the male patient demonstrates symmetric dilation resulting in a circular shape that is less prone to error in measurements with differing image acquisition planes. Furthermore, significant trabeculations are seen along the basal inferolateral wall (Blue arrow) in the female patient, and, the echocardiographic image fails discern the trabeculations from the compacted myocardium during systole, resulting in a significantly smaller LVESD measurement compared to CMR. Lastly, women may have a higher prevalence of less optimal acoustic windows or image quality due to breast attenuation, which may obscure the ability to definitively identify the correct endocardial border.

Our study findings have important clinical implications, as two recent studies demonstrated increase in adverse post-operative outcomes and survival for patients who underwent surgical AVR with LVESDI 2.0–2.5 and > 2.5 cm/m² [27, 28]. These findings suggest that the current guideline recommendations should be updated to include LVESDI with a threshold of 2.0–2.5 cm/m². Thus, underestimation of LVESDI by echocardiography may lead to delayed surgical intervention and worse outcomes. Because men demonstrated a consistent underestimation of LV diameters by echocardiography, a simple correction factor could be derived for echocardiography-derived LV diameters in men. However, the difference between LV diameters by echocardiography and CMR increased with increasing dimension for women in our study, and women with LVESDI by CMR > 2.5 cm/m² demonstrated an underestimation of LVESDI by echocardiography by an average of 0.70 ± 0.35)cm/m². This significant discrepancy in LVESDI measurements demonstrates the impact of sex on the accuracy of LVESDI measurements in a range that has critical importance in regard to surgical referral and association of adverse post-surgical outcomes. Our findings, in conjunction with recent CMR studies by Kammalander et al. [23] and Malahfji el al. [29] demonstrate the unique ability for CMR to more accurately characterize sex based differences in adverse LV remodeling and resultant impact on adverse outcomes, thus suggesting that CMR may be the best suited modality to derive sex-based thresholds for surgical referral in patients with significant AR. Further studies are needed to determine if the higher prevalence in symptoms at lower severity of chronic AR in women may be attributed to differences in compensatory remodeling, with resultant increased wall stress and elevated LV filling pressures. Additionally, large multi-center studies will be required to determine if this difference in compensatory LV remodeling is associated with adverse clinical outcomes and if women would benefit from alternative surgical referral criteria.

While this study is one of the largest comparative studies of echocardiography versus CMR in patients with chronic AR, this was a single-center retrospective cross-sectional study examining patients referred for echocardiography and CMR at the discretion of the ordering provider. Echocardiographic contrast was utilized in a minority of the studies (3.6%) and thus conclusions cannot be made regarding whether the use of endocardial border definition contrast to better define the endocardial borders for measurement of LV volumes would alter the findings of sex-based differences. Furthermore, patients with an eGFR < 30 ml/min/1.73 m² and implantable cardiac devices were excluded from our study. Therefore, we cannot exclude the presence of selection bias. Because these CMR studies were clinically indicated studies, the findings of the study may have influenced the decision to refer patients for surgical intervention. This study included a relatively small number of women; however, given the superior reproducibility of CMR measurements of LV remodeling our study is adequately powered to assess differences in LVR assessed by CMR, based on previously published data [14, 30].