Computed tomography angiography-derived extracellular volume fraction predicts early recovery of left ventricular systolic function after transcatheter aortic valve replacement

Abstract

Aims 

Recovery of left ventricular ejection fraction (LVEF) after aortic valve replacement has prognostic importance in patients with aortic stenosis (AS). The mechanism by which myocardial fibrosis impacts LVEF recovery in AS is not well characterized. We sought to evaluate the predictive value of extracellular volume fraction (ECV) quantified by cardiac CT angiography (CTA) for LVEF recovery in patients with AS after transcatheter aortic valve replacement (TAVR).

Methods and results 

In 109 pre-TAVR patients with LVEF <50% at baseline echocardiography, CTA-derived ECV was calculated as the ratio of change in CT attenuation of the myocardium and the left ventricular (LV) blood pool before and after contrast administration. Early LVEF recovery was defined as an absolute increase of ≥10% in LVEF measured by post-TAVR follow-up echocardiography within 6 months of the procedure. Early LVEF recovery was observed in 39 (36%) patients. The absolute increase in LVEF was 17.6 ± 8.8% in the LVEF recovery group and 0.9 ± 5.9% in the no LVEF recovery group (P < 0.001). ECV was significantly lower in patients with LVEF recovery compared with those without LVEF recovery (29.4 ± 6.1% vs. 33.2 ± 7.7%, respectively, P = 0.009). In multivariable analysis, mean pressure gradient across the aortic valve [odds ratio (OR): 1.07, 95% confidence interval (CI): 1.03–1.11, P: 0.001], LV end-diastolic volume (OR: 0.99, 95% CI: 0.98–0.99, P: 0.035), and ECV (OR: 0.92, 95% CI: 0.86–0.99, P: 0.018) were independent predictors of early LVEF recovery.

Conclusion 

Increased myocardial ECV on CTA is associated with impaired LVEF recovery post-TAVR in severe AS patients with impaired LV systolic function.

Introduction

Aortic stenosis (AS) is a highly prevalent primary valve disease in the developed world.¹ In patients with AS, left ventricular (LV) pressure overload leads to compensatory LV hypertrophy to maintain wall stress and cardiac output. Left untreated, severe AS can result in LV systolic dysfunction and is associated with adverse clinical outcomes.^2,3 By relieving LV afterload, surgical or transcatheter aortic valve replacement (TAVR) can improve LV function. However, a significant proportion of patients with low LV ejection fraction (EF) experience no improvement after TAVR.^4,5

Myocardial extracellular expansion and replacement fibrosis result in deteriorating LV systolic function and irreversible LV decompensation in patients with AS.^2,6,7 Cardiac magnetic resonance (CMR) has been frequently used to assess the expansion of the extracellular matrix with native T1 mapping and the extracellular volume fraction (ECV) by equilibrium CMR technique.^8,9 Increased ECV can be associated with replacement fibrosis detected by late gadolinium enhancement by CMR.² Iodinated contrast agents are extracellular tracer and an equilibrium technique similar to that used with CMR can be used for quantification of ECV by cardiac computed tomography angiography (CTA). ECV quantified by CTA from pre-contrast and delayed post-contrast CT images obtained before and after the initial transit of a contrast bolus through the myocardium has been shown to accurately reflect extracellular matrix and replacement fibrosis assessed by CMR and histological studies.^10–13 We hypothesized that CTA derived ECV might allow discrimination between TAVR patients who experience post-procedural LVEF recovery and those with impaired LVEF recovery. In this study, we evaluated the association between baseline ECV and early changes in LVEF after TAVR in severe AS patient with LV dysfunction.

Methods

Study population

We identified 109 patients who underwent TAVR at Cedars Sinai Medical Center from September 2016 to October 2018, (i) CTA acquired with both initial and delayed imaging for calculation of ECV, (ii) LVEF <50% on baseline echocardiography, and (iii) follow-up echocardiography within 6 months after TAVR. Patient medical history, demographic information, and laboratory values were obtained from electronic medical records. TAVR was performed via transfemoral or subclavian approach, using either balloon expandable (Edwards Sapien), or self-expanding (Medtronic Evolute or Corevalve) devices. The study was approved by the IRB at Cedars Sinai Medical Centre.

Cardiac CT protocol

All imaging was performed using a dual-source CT system (SIEMENS SOMATOM Definition Flash; SIEMENS Healthcare, Erlangen, Germany). Pre-contrast and delayed post-contrast prospectively electrocardiogram (ECG)-triggered acquisitions were performed using a standard coronary artery calcium scan protocol at end-systole during an inspiratory breath hold as previously described.^14,15 Delayed ECG-gated post-contrast CT was performed 5 min after contrast injection. For CTA, 100 mL bolus injection of iodine contrast, Omnipaque (GE Healthcare, Little Chalfont, Buckinghamshire, UK) was used with bolus triggering in the ascending aorta. CTA was performed with a collimation of 128 mm × 0.625 mm with a peak tube voltage of 120 kVp. Images were acquired craniocaudally, from the aortic arch to the diaphragm without dose modulation over the entire cardiac cycle with ECG gating at 10% intervals within the 0–90% of the cardiac cycle, then followed by a subsequent non–ECG-synchronized CT angiographic study of the chest, abdomen, and pelvis for simultaneous assessment of the access route. The mean radiation dose of pre-TAVR planning CT protocol was 17.5 ± 4.1 mSv.

CT image reconstruction

Pre-contrast and delayed post-contrast images for ECV quantification were reconstructed at 3.0 mm slices.¹² Contrast-enhanced CTA images were reconstructed at 0.6 mm slices with 0.3 mm overlap with iterative reconstruction. LV end-systolic volume (LVESV), LV end-diastolic volume (LVEDV), stroke volume, and myocardial mass were quantified from contrast-enhanced CTA using commercially available software, Syngovia (Siemens Healthineers, Erlangen, Germany).

ECV quantification

CT attenuation in Hounsfield units (HU) was measured in regions of interest (ROI) drawn in the septum and lateral wall in the basal and mid LV, on the axial image that showed greatest myocardial thickness and which did not have significant beam hardening or streak artefacts. Wall segments with poor endocardial definition were not included. Blood pool attenuation was measured using the same image, by placing an ROI in the LV cavity. ROIs drawn in the delayed post-contrast scan were placed at the same location as the pre-contrast scan using software based co-registration. In each patient, myocardial ECV, expressed as a percentage, was calculated as ECV_CT = (1 – haematocrit)×(ΔHU_myo/ΔHU_blood), where ΔHU_myo and ΔHU_blood are given by (HU_delayed – HU_early) of the myocardium and blood pool ROI^10–12 (Figure 1). Myocardial ECV was estimated by the mean ECV values of the septum and lateral wall. Among patients with prior myocardial infarction, the affected wall segment was excluded.

Echocardiography

All patients underwent transthoracic echocardiography before TAVR [median 14 days, interquartile range (IQR) 4–38 days]. The severity of pre-TAVR AS was assessed by the mean transvalvular pressure gradient (PG) and by aortic valve area (AVA) which was calculated using the continuity equation.¹⁶ LV function and chamber dimensions were measured using standard echocardiographic criteria.¹⁷ Severe AS was defined as AVA ≤1 cm² and AVA indexed to body surface area ≤0.6 cm²/m², associated with a transvalvular mean PG >40 mmHg. In patients with transvalvular mean PG ≤40 mmHg, severe AS was diagnosed by the presence of the following echocardiographic and CTA criteria adjudicated by investigators who were level 3 certified in echocardiography and cardiac CT: AVA ≤1 cm², AVA indexed to body surface area ≤0.6 cm²/m², and severe valve leaflet calcification defined as aortic valve calcium score of >1200 in women and >2000 in men and restricted leaflet opening by transoesophageal echocardiography.^18–20

Statistical analysis

LVEF recovery was defined as absolute increase of LVEF ≥ 10% on follow-up echocardiography compared with baseline.^5,21 Patients were divided into LVEF recovery or no LVEF recovery groups. The optimal cut-off value for ECV to predict EF recovery was determined using Liu's method which defines the optimal cut-point as the value at which the product of sensitivity and specificity is the highest.²² Continuous variables are expressed as mean ± standard deviation (SD) or median (IQR), and categorical variables are reported as counts with proportions. Continuous variables were compared using the Student’s t-test or Wilcox rank-sum test, and categorical variables were compared using the Pearson χ² test. Logistic regression analysis was used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) based on ECV, clinical, CTA, and echocardiographic imaging variables. Model fit for prediction of LVEF recovery by a model containing LVEDV and AV mean PG was compared with a model that included ECV in addition to LVEDV and AV mean PG. A reduction in residual deviance of the model after inclusion of ECV was considered to represent a better fit. In a sensitivity analysis, we restricted the sample population to include only patients with (i) ECV <40% (n = 90) or (ii) good agreement between septal and lateral ECV (n = 105). Variables with a P-value <0.1 by univariate analysis were included in multivariate analysis using backward stepwise logistic regression analysis. Inter- or intra-observer variability was assessed in randomly selected 15 patients by intraclass correlation coefficients (ICC) for agreement of ECV measurements. A two-tailed P value of <0.05 was considered statistically significant. All statistical analyses were performed using STATA Version 13 (StataCorp LP, College Station, TX, USA).

Results

Baseline characteristics and early LVEF changes pre- and post-TAVR

The mean age of study participants was 80 ± 9.7 years and 83 (76%) were male. Median interscan interval between baseline and follow-up echocardiogram was 43 days (IQR 35–56). In the overall population, LVEF improved from 32.2 ± 10.1% to 38.9 ± 13.3% after TAVR (P < 0.001). LVEF recovery was observed in 36% (39/109 patients) with a wide range in the degree of LVEF changes after TAVR (Figure 2). The absolute increase in LVEF was 17.6 ± 8.8% in LVEF recovery group and 0.9 ± 5.9% in no LVEF recovery group (P < 0.001). Patients with no LVEF recovery compared with those with LVEF recovery were more commonly male and had prior revascularization (Table 1). The LVEF recovery and no LVEF recovery groups had similar baseline LVEF, LVESV, diastolic function, and severity of mitral and tricuspid regurgitation. There were significant differences in AV mean PG, peak PG, and AV area index between LVEF recovery and no LVEF recovery groups. In the overall population, septal, lateral, and mean ECV were 31.8 ± 8.2%, 31.9 ± 9.3%, and 31.9 ± 7.4%, respectively. There was no significant difference between septal and lateral ECV (P = 0.914). Bland–Altman analysis revealed that there was a good agreement between septal and lateral wall ECV, and four outlier cases were identified (Supplementary data online, Figure S1). CTA-derived ECV was elevated in patients with no LVEF recovery group relative to those who with LVEF recovery group (33.2 ± 7.7 vs. 29.4 ± 6.1, P = 0.009).

Predictors of early LVEF recovery

Univariable and multivariable predictors of early LVEF recovery are shown in Table 2. In univariate logistic regression analysis, sex, prior history of coronary revascularization, AV mean PG, AV peak PG, AVA index, LVEDV, and ECV were associated with LVEF recovery. A multivariable model identified three imaging variables which were associated with early LVEF recovery: AV mean PG (OR: 1.07, 95% CI: 1.03–1.11, P: 0.001), LVEDV (OR: 0.99, 95% CI: 0.98–0.99, P: 0.035), and ECV (OR: 0.92, 95% CI: 0.86–0.99, P: 0.018, Table 2). Sex, history of prior coronary revascularization, and AVA index did not predict early LVEF recovery in multivariable analysis. We compared the fit of a model containing LVEDV and AV mean PG to the fit of a model that included ECV in addition to LVEDV and AV mean PG for testing the association between LVEF recovery vs. CTA and echocardiographic imaging variables. There was a discernible improvement to the model fit when ECV was included in the model (P = 0.013).

ECV threshold for prediction of LVEF recovery

An ECV of 30% provided the best discrimination between patients who experienced LVEF recovery and those who did not. There were 62 patients with ECV ≥30% and 47 with ECV <30%. Baseline LVEF was not different between patients with ECV ≥30% and those with ECV <30% (31.2 ± 10 vs. 33.7 ± 10.2, P = 0.128, Figure 3). A follow-up LVEF was lower in patients with ECV ≥30% compared with those with ECV <30% (36.8 ± 13.1 vs. 41.9 ± 13.2, P = 0.047). The proportion of patients who experienced LVEF recovery was lower among those with increased ECV (≥30%) compared with those with normal ECV (<30%) (27% vs. 47% P = 0.036). Every unit (%) of ECV increase over 30% threshold was associated with an 11% decrease in the likelihood of early LVEF recovery (OR: 0.89, 95% CI: 0.82–0.97, P: 0.012). In a sensitivity analysis, we restricted the sample population to include only patients with (i) ECV <40% or (ii) good agreement between septal and lateral ECV. Increased ECV defined as ≥30% still displayed significant association with LVEF recovery in both analyses (OR: 0.88, 95% CI: 0.80–0.98, P: 0.018 and OR: 0.92 95% CI: 0.86–0.99, P = 0.017, respectively).

Reproducibility of myocardial ECV measurement

Intra- and inter-observer agreement of ECV expressed as an ICC were 0.91 (95% CI 0.74–0.97) and 0.88 (95% CI: 0.64–0.96), respectively (both P < 0.001).

Discussion

In this single-centre study of severe AS patients with impaired LV function, we observed 64% of study participants did not experience early EF recovery after TAVR. Increased myocardial ECV derived by CTA was associated with reduced likelihood of early LVEF recovery after TAVR. Each percent increase in ECV over the threshold of 30% was associated with 11% reduction in likelihood of early LVEF recovery.

Previous investigators have shown that nearly a third of patients with severe AS have LV systolic dysfunction,^21,23 which often is associated with absence of LVEF recovery.^4,5,24 Furthermore, patients with severe AS in whom there is no LVEF recovery after aortic valve replacement experience worse clinical outcomes compared with those with LVEF recovery.^4,5,24. Une et al.²⁴ reported that 30% of severe AS patients did not experience LV recovery at five years after surgical AVR. Large TAVR trials reported that 30–49% of patients with baseline LV dysfunction experienced no early improvement in LVEF after TAVR.^4,5 In the current investigation, we observed a wide range of early changes of LVEF from −13% to 39%, and 64% of patients did not experience LVEF recovery after TAVR.

Prior studies reported that myocardial extracellular compartment expansion or replacement fibrosis may interfere with LVEF recovery.^2,25–27 Histopathologic assessment of myectomy specimens from patients with AS has shown that LV fibrosis and myocyte degeneration are associated with reduction in LVEF.²⁵ Azevedo et al.²⁶ also reported that the amount of myocardial fibrosis is associated with the degree of LV functional improvement after surgical AVR in patients with severe aortic valve disease. In addition, an increased ECV quantified by CMR has been shown to be an indicator of LV decompensation in patients with severe AS and provides independent prognostic information over LVEF.^2,27 ECV can be quantified using routine pre-TAVR procedural planning CTA with the addition of a delayed scan with minimal additional radiation exposure.¹² Prior studies have shown the feasibility of CTA-derived ECV measurement in various population such as patients with AS,¹⁰ heart failure,¹² post-chemotherapy,²⁸ and amyloidosis.²⁹ However, quantification of ECV by CTA has not been adapted for routine clinical practice. Furthermore, the association of ECV quantification with the use of CTA in predicting LVEF recovery post-TAVR has not been previously evaluated. We observed that LVEF recovery was less likely among those with an increased ECV above a threshold of 30%. We believe that the current study findings will add to the body of evidence regarding the clinical relevance of using CTA for ECV measurements in patients with AS.

In patients with AS, extracellular expansion (diffuse fibrosis) and replacement fibrosis, which is considered to represent an advanced stage of extracellular expansion, can be detected by CMR and quantified as ECV.^2,8,30 Higher ECV has been observed in patients with replacement fibrosis compared with those with extracellular expansion alone.² This stepwise increase in ECV in relation to the degree of fibrosis has not been well validated by CTA-derived ECV. Given the high correlation between CTA-derived and CMR-derived ECV,^11,12 we postulate that elevated CTA-derived ECV might also represent an advanced stage of myocardial fibrosis (replacement fibrosis). Further validation of our current findings with CMR or histology-based assessment of myocardial fibrosis will be necessary.

Although myocardial fibrosis is a prominent aetiology of increased ECV in patients with severe AS, in some patients, the cause of increased ECV may be related to cardiac amyloidosis, which has been reported to be present in nearly 25% of patients with severe AS.^31,32 ATTR amyloidosis is associated with very high ECV as calculated by CMR.^33,34 It is possible that some of the patients who had poor LVEF recovery with high ECV may have had ATTR amyloidosis. In this study population, there were 19 of 109 (17%) of patients with very high ECV (>40%) which is a non-specific but prominent characteristic of amyloidosis.³⁵ However, amyloidosis was confirmed in only one of these patients. Nonetheless, when these patients with very high ECV were excluded, the significance of elevated ECV remained unchanged. Further investigations are needed to explore the degree to which ATTR amyloidosis might be the underlying cause of impaired LVEF recovery after TAVR.

In addition to ECV, mean transvalvular PG was found to be associated with early EF recovery. The finding is consistent with findings from multiple prior observations that is low mean PG is associated with low likelihood of EF recovery after TAVR.^4,5 Baseline LVEDV was also associated with early LVEF recovery post-TAVR. Increased LVEDV was observed in patients with severe AS as a late myocardial response to elevated systolic wall stress that has shown to be associated with impaired LV systolic function.^24,36 Our results show that LV dilation signals irreversible LV remodelling and thus affects recovery of LV function after TAVR.

Limitations

This is a single-centre study with a relatively small sample size (n = 109); however, to our knowledge, this is the first and largest study that describes the association between CTA-derived ECV and EF recovery in patients with LV dysfunction undergoing TAVR. Given the retrospective and observational nature of the current study, we cannot discount the effect of unmeasured confounding factors, which might have influenced the clinical endpoints of this study. Among patients with concomitant coronary artery disease, information regarding completeness of revascularization and size of prior myocardial infarction, which are significant confounders and can affect LVEF recovery were not available and thus not included in the analysis. Clinical prognostic outcome data were not available in the current investigation; thus, our study was limited in evaluating the association between ECV and LVEF recovery vs. clinical outcomes. However, based on previous reports, one might extrapolate that patients without LVEF recovery could potentially experience poor outcomes relative to those with LVEF recovery.^4,5,24 LVEF recovery evaluation was limited to a 6-months follow-up period thus limiting our ability to predict long-term LVEF recovery. However, previous studies have shown that only a minority of patients without early EF recovery experience an LVEF improvement later in their clinical course.⁵ Although increase in ECV may reflect interstitial fibrosis induced by pressure overload in patients with AS, it may be also attributed to myocardial amyloid deposition, cardiomyopathy from ischaemic or non-ischaemic aetiology, and extracellular oedema as might be seen in myocarditis.^33,37,38 In our study, the optimal cut-point for increased ECV of ≥30% as a predictor of improvement would require confirmation in further study; however, this cut-point is the same as have been previously described with CMR.³⁹

Conclusion

Increased by ECV using CTA is associated with poor LV function recovery after TAVR in patients with severe AS and reduced baseline LVEF. This measurement, which can be made from the routine pre-TAVR CTA, might thus help to improve patient selection and optimization of the timing of TAVR.

Supplementary data

Supplementary data are available at European Heart Journal - Cardiovascular Imaging online.

Funding

Dr. Miriam and Sheldon G. Adelson Medical Research Foundation. B.T. was supported by R56HL131871 from NHLBI.

Conflict of interest: none declared.

References

Author notes