Author + information
- Received April 7, 2017
- Revision received July 5, 2017
- Accepted July 6, 2017
- Published online October 18, 2017.
- Oren Zusman, MDa,b,∗ (, )
- Gregg S. Pressman, MDc,
- Shmuel Banai, MDb,d,
- Ariel Finkelstein, MDb,d and
- Yan Topilsky, MDb,d
- aDepartment of Medicine E, Rabin Medical Center, Petah-Tiqva, Tel Aviv, Israel
- bSackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel
- cEinstein Medical Center, Philadelphia, Pennsylvania
- dDivision of Cardiovascular Diseases and Internal Medicine, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel
- ↵∗Address for correspondence:
Dr. Oren Zusman, Department of Internal Medicine E, Rabin Medical Center, Zabotinsky 39, Petah-Tiqva, Hamerkaz 4941492, Israel.
Objectives To describe patients with severe symptomatic aortic stenosis with normal flow and low gradients and determine whether they benefit from intervention.
Background Severe symptomatic aortic stenosis is a progressive disease with high mortality. Although surgical aortic valve replacement (SAVR) or transcatheter aortic valve replacement (TAVR) are indicated for patients with high gradients (>40 mm Hg) or low gradients due to low flow, the approach for patients with normal flow and low gradients is poorly defined.
Methods Consecutive adult patients who underwent echocardiography between 2012 and 2015 at Tel-Aviv Medical Center and had an aortic valve area of ≤1.0 cm2, a mean gradient of <40 mm Hg, a stroke volume index of >35 ml/m2, and symptoms formed the study group. Patients designated for intervention (SAVR or TAVR) had their procedure within 6 months of the echocardiogram; the others were treated conservatively. The endpoints were all-cause mortality and cardiac-related mortality.
Results During the study period, 1,358 patients with an aortic valve area of ≤1.0 cm2 and symptoms were identified; 34% of these had normal flow and low gradient aortic stenosis and 303 were included. After mean follow-up of 652 days, 60 patients (20%) had died, with overall mortality rates of 28%, 10%, and 12% for conservatively treated, TAVR, and SAVR patients, respectively (p < 0.001). Using Cox regression with adjustment for other variables, TAVR was associated with improved survival versus conservative treatment (hazard ratio [HR]: 0.49; 95% confidence interval [CI]: 0.26 to 0.93; p = 0.03), and lower cardiac mortality (HR: 0.3; 95% CI: 0.10 to 0.74; p = 0.007) with no significant difference for SAVR versus TAVR. After propensity score matching of TAVR and conservatively treated patients, 25 of 94 (28%) conservatively treated and 10 of 94 (12%) TAVR patients had died (p = 0.016). In the matched cohort, Cox regression showed that TAVR had a significant association with improved survival (HR: 0.42; 95% CI: 0.20 to 0.86; p = 0.03).
Conclusions Symptomatic patients with an aortic valve area of ≤1.0 cm2, normal flow, and low gradient may benefit from intervention as opposed to conservative treatment.
Aortic stenosis (AS) is a progressive disease; once symptoms develop, the prognosis is poor (1). The standard of care for patients with severe symptomatic AS is valve replacement, be it surgical aortic valve replacement (SAVR) or by transcatheter aortic valve replacement (TAVR) (2). The definition of severity and, therefore, the indication for intervention is based on imaging parameters, usually obtained by transthoracic echocardiography (3). The European Society of Cardiology and American Heart Association/American College of Cardiology guidelines (2,4) define “severe AS” when the aortic valve area (AVA) is <1.0 cm2 and the mean transvalvular pressure gradient is ≥40 mm Hg. However, some patients have low gradients despite having severe stenosis by AVA (5). This can be due to low flow across the valve associated with either a reduced or preserved ejection fraction (6–11). Other patients display normal flow and gradients <40 mm Hg (normal flow low gradient [NFLG]) despite severe stenosis by AVA. It has been demonstrated that the AVA needed to generate a gradient of 40 mm Hg is closer to 0.8 cm2 than to 1.0 cm2, and that there can be discrepancies among echocardiographic parameters used to determine need for intervention (5). This research addresses the group of patients with normal flow, low gradient, and an AVA of ≤1.0 cm2 for whom there is conflicting evidence regarding the need for intervention (12–14).
We identified all adult patients who underwent echocardiography between 2012 and 2015 at Tel-Aviv Medical Center who had an AVA of ≤1.0 cm2 (by continuity equation) and documented symptoms (angina, dyspnea, or syncope). Patients were identified through the computerized echocardiographic records, and data were extracted retrospectively. Patients with no symptoms, incomplete Doppler echocardiogram, previous aortic valve replacement (AVR) or TAVR, or severe concomitant valvular disease were excluded. History, coexisting conditions, and clinical examination were recorded at baseline by patients’ personal physicians at our institution. Clinical management was determined independently by the patients’ personal physicians using all information available. Baseline clinical and follow-up data were obtained by a review of the medical records.
Patients were further defined according to flow and gradient status, with a gradient of <40 mm Hg being considered as “low gradient,” and those with a stroke volume index (SVI) of ≥35 ml/m2 being considered “normal flow.” To ascertain the accuracy of the Doppler echocardiographic measurements of stroke volume we used the following steps: 1) we used the clearest on-axis image of the left ventricular outflow tract (LVOT) providing the largest LVOT diameter; 2) we measured the LVOT diameter inner-edge-to-inner-edge from the base of the right coronary cusp anteriorly to the commissure posteriorly; and 3) we corroborated the measures of LVOT diameter by relating it to body surface area (within 2 mm of predicted for their body surface area using the formula: LVOT diameter = ([5.7 × body surface area] +12). If the measured diameter was >2 mm smaller or larger than the predicted value, we reassessed the 2-dimensional echocardiographic LVOT measurement with the modified Teichholz method, measuring the apical to the septal bulge in patients with normal left ventricular geometry and by biplane Simpson method in patients with segmental wall abnormalities to estimate the LVOT stroke volume.
Patients designated for intervention underwent their procedure within 6 months of the index echocardiogram; others were treated conservatively. Patients having intervention after the 6-month period were counted with the conservative group and censored at the time of intervention (i.e., any additional mortality benefit due to intervention was neglected). An effort was made to identify the reason for choice of treatment strategy. In patients who underwent interventions (either TAVR or SAVR), eligibility for TAVR versus SAVR was determined by our heart team. For cases in which calculated Society of Thoracic Surgeons score was ≥8, TAVR is the procedure of choice. In cases with a Society of Thoracic Surgeons score of <8, TAVR is elected if surgical risk is considered high based on the basis of frailty measures not included in the standardized risk scores, and SAVR elected in the rest of patients.
The primary outcome was all-cause mortality, which was determined from Interior Ministry records. As a secondary analysis, we defined cardiac-related death as arrhythmic sudden death and heart failure–related death, and noncardiac death separately, according to the criteria of Hinkle and Thaler (15). Time to event started from the date of the index echocardiogram for conservatively treated patients, and from the date of intervention for TAVR/SAVR patients. We also identified subsequent echocardiographic exams. The study was approved by the Institutional Review Board of the Tel-Aviv Sourasky Medical Center.
Interobserver and intraobserver variability
Interobserver variability was assessed by comparing the readings made by another independent reader in 15 randomly selected patients. Intraobserver variability was determined by having the observer who measured the data remeasure the parameters. The degrees of interobserver and intraobserver variability were determined using the Bland-Altman method and the within-subjects coefficient of variation. The within-subjects coefficient of variation (calculated as a ratio of the standard deviation of the measurement difference to the mean value of all measurements) provides a scale-free estimate of variation expressed as a percentage. We measured the intraobserver and interobserver reproducibility for the LVOT diameter, the LVOT VTI, the aortic valve VTI, the calculated SVI, and the calculated AVA.
Continuous, normally distributed variables are presented as mean ± SD and compared using the Student t test. Ordinal and/or non-normally distributed variables are presented as median and interquartile range and compared using the Wilcoxon rank-sum test. Normality was assessed using the Shapiro-Wilk test and visual inspection of quantile-quantile (QQ) plots. Categorical variables are compared using the chi-square test. Correlation between AVA, AVA index, and transaortic gradient was examined and presented using the Pearson r correlation coefficient. Patients with missing information were to be excluded from the analysis. A Cox proportional hazard model was fitted for all-cause, long-term mortality. We examined multicollinearity using variation inflation factors, with a result of >2.5 considered high. All available variables were entered into the model and selected based on backward stepwise algorithm using the Akaike information criterion. Proportionality of odds was tested based on Schoenfeld residuals. We sought to preserve at least 1:10 events per variable. To better adjust for confounders, we used propensity score matching. The propensity score was estimated using logistic regression with the goal of creating the best balanced groups; collinear parameters, noise variables, and parameters that hindered balance were removed. No stepwise selection process was applied. The propensity score was estimated using logistic regression with all variables entered into the model, and then matching was performed using nearest neighbor (i.e., matching 2 closest propensity score pairs) with a 1:1 ratio between the treated and control groups and without replacement. To further decrease disparity in pairs, matching was restricted by a caliper of 0.7 of the SD of the propensity score. Assessment of balance was performed using the methods described by Hansen and Bowers (16), and by inspecting resulting standardized mean differences. A standard mean difference of less than 0.2 was considered small. All analyses were performed using R (17) with relevant packages (18–20).
A total of 1,358 patients had an AVA of ≤1.0 cm2. Of those, 40% had high gradient AS, 26% had low flow low gradient AS, and 34% had NFLG AS, of which 303 were included for analysis after applying exclusion criteria (Figure 1). Baseline characteristics, drug therapy, and echocardiographic data are presented in Table 1. In general, SAVR patients were younger and had fewer comorbidities than TAVR or conservatively treated patients. The correlation of AVA with mean gradient was modest (r = 0.26), with the AVA index having a slightly higher correlation (r = 0.38). Of the conservatively treated patients, 64 (43%) had intervention deferred out of a conscious cardiologist decision because they did not meet with current guidelines for severe AS, 53 patients (35%) were treated so due to perceived high risk/benefit ratio (considering that transaortic gradients were not severely elevated), and 32 (22%) were treated conservatively due to patient preference. Of the conservatively treated patients with subsequent echocardiography data (n = 61), 33 (54%) retained NFLG status, 15 (25%) transitioned to low flow low gradient AS, and 13 patients (21%) progressed to a high gradient. Averaged for 1 year, AVA was reduced by an average of 0.1 ± 0.2 cm2, whereas the mean gradient increased by an average of 4.0 ± 8.1 mm Hg. Out of the conservative group, 3 patients underwent TAVR (after 432, 628, and 830 days from the index echocardiographic examination), and 1 patient underwent SAVR (372 days after the index echocardiographic examination). These patients were censored at the date of the procedure.
After a mean follow-up time of 652 days, 60 patients had died: 43 of 149 (29%) conservatively treated patients, 13 of 114 (11%) TAVR patients, and 4 of 40 (10%) SAVR patients (p < 0.001 for the difference between all; p = 1.0 for TAVR vs. SAVR). Kaplan-Meier curves are presented in Figure 2. By the log-rank test, the association of type of intervention with survival was significant (p = 0.002) due to differences between conservatively treated patients and SAVR/TAVR (p = 0.56 for difference between the latter two). Cox regression with adjustment for age, ejection fraction, chronic kidney disease, congestive heart failure, and previous stroke showed a significant association of intervention type with mortality with a HR of 0.49 (95% CI: 0.26 to 0.93; p = 0.03) for TAVR versus conservative treatment, and an HR of 0.55 (95% CI: 0.17 to 1.75; p = 0.31) for SAVR versus conservative treatment. There was no difference for SAVR versus TAVR (p = 0.85) (Table 2).
To further account for bias, we proceeded with matching against the propensity score as described. Matching produced 2 equal groups of 94 patients in the TAVR and the conservative groups each, with overall difference between group’s characteristics nonsignificant (p = 0.5). The C-statistic for the propensity score model was 0.80. Characteristics of both groups are presented in Table 3. In the matched cohort, 25 of 94 conservatively treated patients (27%) and 11 of 94 TAVR treated patients (12%) had died (p = 0.016). In Cox regression analysis in the matched cohort, type of intervention also had a significant association with mortality in unadjusted analysis (HR: 0.42; 95% CI: 0.2 to 0.86; p = 0.04) and in an adjusted one (HR: 0.45; 95% CI: 0.22 to 0.92; p = 0.03) (Table 4). Last, in a secondary outcome analysis, TAVR treatment compared with conservative therapy was associated with a significant decrease in cardiac death (HR: 0.3; 95% CI: 0.10 to 0.74; p = 0.007), but not in noncardiac death (HR: 0.55; 95% CI: 0.22 to 1.25; p = 0.2).
Interobserver and intraobserver variability
Comparison of intraobserver parameters showed good agreement between measurements: LVOT diameter (mean difference 0.13 ± 1.3 mm; r = 0.95; p = 0.9), LVOT VTI (mean difference −1.2 ± 0.3 cm; r = 0.94; p = 0.2), aortic valve VTI (mean difference −1.06 ± 0.8 cm; r = 0.95; p = 0.2), calculated SVI (mean difference −2.3 ml/m2; r = 0.95; p = 0.2), and AVA (mean difference −0.03 cm2; r = 0.93; p = 0.1). The Bland-Altman plot showed a random scatter of points around 0, indicating no systematic bias or measurement error proportional to the measurement value. Measurement variability (within-subject coefficient of variation) for intraobserver differences was as follows: LVOT diameter, 0.6%; LVOT VTI, 0.5%; aortic valve VTI, 0.8%; SVI, 0.9%; and AVA, 1.0%.
Comparison of interobserver parameters showed good agreement between measurements: LVOT diameter (mean difference 0.8 ± 1.4 mm; r = 0.95; p = 0.5), LVOT VTI (mean difference 0.74 ± 0.3 cm; r = 0.91; p = 0.3), aortic valve VTI (mean difference −1.0 ± 0.9 cm; r = 0.93; p = 0.3), calculated SVI (mean difference 1.9 ml/m2; r = 0.90; p = 0.2), and AVA (mean difference 0.03 cm2; r = 0.89; p = 0.1). The Bland-Altman plot showed a random scatter of points around 0, indicating no systematic bias or measurement error proportional to the measurement value. Measurement variability (within-subject coefficient of variation) for interobserver differences was as follows: LVOT diameter, 1.8%; LVOT VTI, 1.1%; aortic valve VTI, 0.9%; SVI, 1.5%; and AVA, 1.6%.
In this study, we examined patients with severe symptomatic AS (as defined by an AVA of ≤1.0 cm2) who had a low gradient despite normal flow. The major findings are: 1) NFLG patients comprise up to one-third of the total symptomatic population with an AVA of ≤1.0 cm2; 2) approximately one-half of NFLG patients transitioned to other flow-gradient groups during follow-up; and 3) intervention was associated with reduced mortality and cardiac-related mortality versus conservative treatment in these patients.
According to guidelines, patients with NFLG AS do not meet requirements for AVR, whether surgical or via TAVR. The 1-year mortality rate in the patients with NFLG who were treated conservatively in our study was 20%, compared with 51% at 1 year in high gradient patients in the PARTNER (Placement of AoRTic TraNscathetER Valve) trial (21) or 50% in low flow low gradient patients (8). In other studies dealing with NFLG patients, mortality has ranged from 10% to 40% (12,13,22).
Our study found intervention to be associated with better survival than conservative therapy among NFLG patients, similar to results found in another study that evaluated patients with low gradient severe AS (12). This finding is in contrast with other research that found no benefit of AVR in this patient group (13), or similar long-term mortality when comparing asymptomatic NFLG patients with those with moderate AS (22). However, these studies included asymptomatic patients with very few TAVR patients. By contrast, our study focused only on symptomatic patients and included a substantial number of TAVR patients.
Because there is discordance between AVA and mean gradient cutoffs (5), it has been suggested that the AVA corresponding with a 40 mm Hg transvalvular gradient is closer to 0.8 cm2 than 1.0 cm2. The question is then what should be changed—should we require a smaller AVA to define “severe AS” as several authors suggest (23), or should we lower the mean gradient requirement? The results of our study suggest that for symptomatic patients with an AVA of ≤1.0 cm2 and normal flow, individual judgment on appropriateness of intervention might be more important than adherence to a specific gradient cutoff. This strategy is reasonable, because patients with a mean gradient of 30 to 35 mm Hg can progress to higher gradients in <1 year (24,25) and, because gradients, as with other biologic processes, exist on a spectrum and do not neatly fit into categories defined by simple cutoffs.
Our study has several limitations. Because of its observational, retrospective nature, it is subject to selection bias and its results imply association and not effect. We attempted to control for this by exclusion of patients who otherwise would not be considered for intervention (such as those with active malignancy), and by using adjusted analysis with propensity scores, which resulted in much more balanced albeit smaller groups. In addition, we have examined cause of death as another measure to support the notion that these patients benefit from intervention directly. Although the groups have differences, the conservative and TAVR groups are quite similar in terms of age, sex, and cardiac and noncardiac comorbidities. Nevertheless, these results should be interpreted with caution, but might serve as incentive to explore this issue in future, prospective randomized trials.
COMPETENCY IN MEDICAL KNOWLEDGE: Patients with symptomatic normal flow (>35 ml/m2), low gradient (mean gradient <40 mm Hg) and a calculated AVA of ≤1.0 cm2 (NFLG AS) benefitted from intervention (either TAVR or surgical AVR) compared with conservative treatment. This benefit remained significant even after adjustment for clinical variables and other echocardiographic parameters. These results suggest that, although not indicated in current guidelines, for symptomatic patients with NFLG AS, individual judgment on appropriateness of intervention, or tighter follow-up might be more important than adherence to a specific gradient cutoff.
TRANSLATIONAL OUTLOOK: Additional larger randomized prospective studies are needed to validate our findings and test whether performance of TAVR or surgical AVR may improve outcome in broad unselected patients with NFLG AS. Furthermore, because current assessment of severity of AS by AVA, peak velocity, and gradients do not seem to convey the full complexity of the process, different parameters should be sought to yield a better threshold for intervention.
The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- aortic stenosis
- aortic valve area
- aortic valve replacement
- left ventricular outflow tract
- normal flow low gradient
- surgical aortic valve replacement
- stroke volume index
- transcatheter aortic valve replacement
- Received April 7, 2017.
- Revision received July 5, 2017.
- Accepted July 6, 2017.
- 2017 American College of Cardiology Foundation
- Joint Task Force on the Management of Valvular Heart Disease of the European Society of Cardiology (ESC),
- European Association for Cardio-Thoracic Surgery (EACTS),
- Vahanian A.,
- Alfieri O.,
- Andreotti F.,
- et al.
- Nishimura R.A.,
- Otto C.M.,
- Bonow R.O.,
- et al.
- Dayan V.,
- Vignolo G.,
- Magne J.,
- Clavel M.-A.,
- Mohty D.,
- Pibarot P.
- Levy F.,
- Laurent M.,
- Monin J.L.,
- et al.
- Tribouilloy C.,
- Lévy F.,
- Rusinaru D.,
- et al.
- Tribouilloy C.,
- Rusinaru D.,
- Maréchaux S.,
- et al.
- Clavel M.-A.,
- Dumesnil J.G.,
- Capoulade R.,
- Mathieu P.,
- Sénéchal M.,
- Pibarot P.
- Clavel M.-A.,
- Pibarot P.,
- Dumesnil J.G.
- Ozkan A.,
- Hachamovitch R.,
- Kapadia S.R.,
- Tuzcu E.M.,
- Marwick T.H.
- Kang D.-H.,
- Jang J.Y.,
- Park S.-J.,
- et al.
- Pellikka P.A.
- Hinkle L.E.,
- Thaler H.T.
- ↵(2015) R Core Team R: A language and environment for statistical computing (R Foundation for Statistical Computing, Vienna, Austria).
- ↵Harrell FE. rms: Regression Modeling Strategies. R package version 5.1-0. Available at: https://CRAN.R-project.org/package=rms. Accessed January 1, 2017.
- Ho D.E.,
- Imai K.,
- King G.,
- Stuart E.A.
- Jander N.,
- Minners J.,
- Holme I.,
- et al.
- Zoghbi W.A.
- Eveborn G.W.,
- Schirmer H.,
- Heggelund G.,
- Lunde P.,
- Rasmussen K.
- Otto C.M.,
- Burwash I.G.,
- Legget M.E.,
- et al.