Author + information
- Received August 28, 2000
- Revision received November 28, 2001
- Accepted January 30, 2002
- Published online April 17, 2002.
- Jeremy J. Pereira, MB, BS*,
- Michael S. Lauer, MD, FACC*,
- Mohammad Bashir, MB, BS*,
- Imran Afridi, MD, FACC*,
- Eugene H. Blackstone, MD, FACC†,
- William J. Stewart, MD, FACC*,
- Patrick M. McCarthy, MD†,‡,
- James D. Thomas, MD, FACC* and
- Craig R. Asher, MD, FACC*,* ()
- ↵*Reprint requests and correspondence:
Dr. Craig R. Asher, The Cleveland Clinic Foundation, Desk F-15, 9500 Euclid Avenue, Cleveland, Ohio 44195 USA
Objectives We sought to assess whether aortic valve replacement (AVR) among patients with severe aortic stenosis (AS), severe left ventricular (LV) dysfunction and a low transvalvular gradient (TVG) is associated with improved survival.
Background The optimal management of patients with severe AS with severe LV dysfunction and a low TVG remains controversial.
Methods Between 1990 and 1998, we evaluated 68 patients who underwent AVR at our institution (AVR group) and 89 patients who did not undergo AVR (control group), with an aortic valve area ≤0.75 cm2, LV ejection fraction ≤35% and mean gradient ≤30 mm Hg. Using propensity analysis, survival was compared between a cohort of 39 patients in the AVR group and 56 patients in the control group.
Results Despite well-matched baseline characteristics among propensity-matched patients, the one- and four-year survival rates were markedly improved in patients in the AVR group (82% and 78%), as compared with patients in the control group (41% and 15%; p < 0.0001). By multivariable analysis, the main predictor of improved survival was AVR (adjusted risk ratio 0.19, 95% confidence interval 0.09 to 0.39; p < 0.0001). The only other predictors of mortality were age and the serum creatinine level.
Conclusions Among select patients with severe AS, severe LV dysfunction and a low TVG, AVR was associated with significantly improved survival.
Symptomatic severe valvular aortic stenosis (AS), with a high transvalvular gradient (TVG) and valve area ≤0.75 cm2, is associated with a high mortality for medically treated patients (1–3), although survival is improved after aortic valve replacement (AVR) (3). When left ventricular (LV) dysfunction develops due to excessive afterload and wall stress and the TVG remains high, the results of AVR remain acceptable (4–6).
For patients with severe AS in the presence of severe LV dysfunction and a mean TVG ≤30 mm Hg, the benefits of AVR remain controversial. Only a small number of select patients with these characteristics, often from older series, have had their outcome reported after AVR, with conflicting results (5,7–12)and with limited power to determine the predictors of survival. The largest series of 52 surgical patients reported by Connolly et al. (12)showed a marked improvement in functional class among 30-day survivors, although the perioperative mortality rate was 21%. Furthermore, no studies have assessed survival in the absence of AVR among this high-risk cohort.
Therefore, we reviewed our institution’s experience to test the hypothesis that among patients with severe AS with severe LV dysfunction and a low TVG, AVR results in improved survival.
From the echocardiographic and surgical databases of our institution, we identified all patients from January 1, 1990 to November 20, 1998 who had an aortic valve area (AVA) ≤0.75 cm2, LV ejection fraction (LVEF) ≤35% and mean TVG ≤30 mm Hg. Patients were excluded if they had more than moderate (>2+) aortic regurgitation by echocardiography, had undergone valve replacement or repair previously or required any valve replacement in addition to AVR during the operation. Patients who underwent concomitant coronary artery bypass graft surgery (CABG) or mitral or tricuspid valve repair were eligible.
Sixty-eight consecutive patients received AVR (AVR group), and 95 consecutive patients did not receive AVR (control group). Of these 95 patients, 5 were excluded because of the presence of life-threatening noncardiac conditions, and 1 was excluded because of heart transplantation, resulting in 89 patients in the control group.
Preoperative clinical data, echocardiographic results, cardiac catheterization hemodynamic data, native coronary anatomy (if catheterization was performed) and operative data were obtained by reviewing the medical records and data bases. Survival from the date of echocardiography was obtained by using the Social Security Death Index of all patients (13,14)and was validated at clinical follow-up in 74% of patients in the AVR and control groups. Follow-up was for a minimum of six months (range 0.5 to 7.5 years). Late postoperative New York Heart Association (NYHA) and Canadian Cardiovascular Society (CCS) functional classes were also assessed in 44 of 46 long-term survivors in the AVR group, by telephone interview with the patient or physician, mail questionnaire or review of the medical records.
Comprehensive two-dimensional echocardiography and preoperative Doppler transthoracic echocardiography were performed in all patients. Left ventricular ejection fraction and right ventricular function were determined by visual estimation. This method of assessing LVEF has been widely employed (6,12,15), and its validity has been confirmed in several studies (16–19), including one that suggests superiority of visual estimation of LVEF over quantitative methods (18). Moreover, when we assessed whether the visually estimated LVEF among the 89 patients in the control group, stratified by LVEF <20% or ≥20%, predicted survival, the survival curves diverged immediately and were considerably worse (p = 0.05) among patients with LVEF <20%, further supporting the validity of this method.
In addition, as an internal validation, among the 157 patients in the AVR and control groups, we quantified, by the volumetric method (20), LVEF in 42 (27%) randomly selected echocardiographic studies. Among these patients, LVEF was visually estimated at 23 ± 7% and calculated to be 24 ± 9% (r = 0.60, p < 0.001). Furthermore, in those studies in which LVEF was calculated, only five patients had LVEF >35% (with peak and mean gradients of 42 and 25 mm Hg), and of these patients, none had LVEF ≥45%.
Regional LV wall motion abnormalities were categorized according to a standard 16-segment model (20). Aortic valve hemodynamic data were assessed using standard methods, and the AVA was calculated by the continuity equation (21). Mitral and aortic regurgitation was semi-quantitated from 0 (none) to 4+ (severe) (22,23). A relative wall thickness ratio (24)and AVA index (AVA/body surface area) were calculated.
Coronary artery disease (CAD) was defined as ≥50% lumen diameter narrowing of the left main or major epicardial vessels. Multivessel CAD was defined as either left main or two or three major epicardial vessel disease. Aortic valve area was calculated from the Gorlin equation (25). Cardiac output was determined by either the Fick or thermodilution method.
The type and size of the aortic prosthesis used in AVR, concomitant CABG or mitral or tricuspid valve repair, as well as the aortic cross-clamp time and cardiopulmonary bypass time, were recorded in the AVR group. In-hospital deaths were defined as deaths before hospital discharge.
Development of AVR propensity scores
Using propensity analysis (26), a logistic regression model (27)was created, where AVR was the dependent variable, and 20 observed and plausible correlates of AVR acted as independent variables, including LVEF, right ventricular systolic function, LV end-diastolic dimension and wall thickness, severity of mitral regurgitation, mean aortic TVG and AVA at echocardiography, age, gender, race, NYHA and CCS functional classes, syncope, diabetes mellitus, hypertension, previous myocardial infarction or CABG, peripheral vascular disease, chronic airway obstruction and serum creatinine. The most important correlates of receiving AVR were younger age, higher mean TVG and male gender. A nonparsimonious propensity score for AVR was generated among the 157 patients who had severe AS with severe LV dysfunction and a low TVG. The area under the receiver operating characteristics curve was 0.86, indicating good discrimination between patients who received and did not receive AVR. After dividing the study group into quintiles, based on propensity scores, patients in propensity quintiles 2 to 4, where the probability of AVR ranged from 11% to 80%, had reasonable matching of propensity scores, variances of propensity scores and baseline characteristics. Further analyses were confined to these propensity-matched patients; 39 patients in the AVR group and 56 patients in the control group. Among patients in quintile 1, who were least likely to receive AVR, there were 31 patients in the control group and 0 patients in the AVR group. Among patients in quintile 5, who were most likely to receive AVR, there were 29 patients in the AVR group and 2 patients in the control group.
Group data were expressed as the mean value ± SD for continuous variables or as percent frequencies for categorical variables. Clinical, echocardiographic, cardiac catheterization and surgical data were compared between patients by using the ttest, Wilcoxon rank-sum test or chi-square test, as appropriate. The McNemar test statistic was determined to compare preoperative and postoperative functional classes. Propensity analysis (26)was used to calculate a propensity score for each patient; this score represented the probability of receiving AVR. Survival was analyzed by constructing Kaplan-Meier curves (28)among patients who did or did not receive AVR, and was expressed as the mean value ± SEM. The Cox proportional hazards model (29)was used to assess the association between AVR and time to death, with the proportional hazards assumption confirmed by testing the time-dependent covariates. The relationship between preoperative variables and postoperative LVEF was assessed by simple and multiple linear regression analyses. All statistical analyses were performed using the SAS system (version 8.1, SAS Inc., Cary, North Carolina), except for the relationship between preoperative variables and postoperative LVEF (SPSS version 10.0, SPSS Inc., Chicago, Illinois).
Among patients in the AVR and control groups, 22 (32%) and 74 (83%) deaths occurred over 2.7 ± 2.3 years and 1.0 ± 1.3 years of follow-up, respectively. The baseline characteristics of the total study group are summarized in Tables 1 and 2, ⇓⇓and the survival curves of the total study group are shown in Figure 1. The baseline characteristics of the propensity-matched patients are shown in Table 3.
Among patients in the AVR group, 24 patients (35%) received a prosthesis that was <22 mm in size, whereas 44 patients (65%) received a prosthesis >22 mm. Among patients in the AVR group, a sub-group of 24 (35%) patients received a prosthesis that was <22 mm while a sub-group of 44 (65%) patients received a prosthesis >22 mm. There was no difference in the percentage that received bioprostheses in either sub-group (79% vs. 77%, respectively; p = NS). All four patients who died perioperatively received a 21-mm prosthesis, three of which were bioprostheses.
Hospital mortality among propensity-matched patients.
among the 39 patients in the avr group, there were three in-hospital deaths (8%) that occurred on postoperative days 2, 44 and 56. among the 56 patients in the control group, there were eight in-hospital deaths (14%).
Medium-term survival among propensity-matched patients
Kaplan-Meier analysis of survival of all propensity-matched patients is shown in Figure 2. The median follow-up among in-hospital survivors in the AVR group was 2.13 years (25th to 75th percentile: 0.83 to 4.78 years). In addition to the three perioperative deaths, there were 11 deaths at late follow-up. One- and four-year survival rates were 82 ± 6% and 78 ± 7%, respectively. The median follow-up period among in-hospital survivors in the control group was 0.75 years (25th to 75th percentile: 0.16 to 1.66 years). In addition to the eight in-hospital deaths, there were 39 deaths at late follow-up. One- and four-year survival rates were 41 ± 7% and 15 ± 5%, respectively.
Multivariable predictors of mortality among propensity-matched patients
Among the 95 propensity-matched subjects in quintiles 2 to 4, there were 61 deaths. A series of Cox regression models relating receiving or not receiving AVR to the risk of death, adjusted for propensity score and six other covariates (i.e., age, serum creatinine level, LVEF, gender, AVA, body surface area), found that receiving AVR remained a strong and independent predictor of survival (adjusted risk ratio (RR) 0.19, 95% confidence interval [CI] 0.09 to 0.39, p < 0.0001), with an 81% decrease in the risk of death. Other variables that predicted mortality were elevated serum creatinine (≥1.5 mg/dl; RR 1.5, 95% CI 1.2 to 1.9, p = 0.0005) and increased age per one-year increment (RR 1.05, 95% CI 1.02 to 1.07, p = 0.002).
When the reason for not receiving AVR was determined as being due to patient refusal, the decision of the physician assessing the patient or significant comorbidities, there were no differences in the survival of each of these patients. Similarly, among those patients in the control group who received or did not receive cardiac catheterization, survival was equally poor. As shown in Table 4, for patients who had AVR, there was no preoperative mean TVG below which survival deteriorated (either perioperative or late deaths). No interactions between receiving AVR, older age, renal dysfunction and LVEF were noted for the prediction of mortality.
Change in functional class in patients who had AVR
Among 44 of 46 long-term survivors, the incidence of NYHA functional class III or IV symptoms decreased from 68% to 18% (Fig. 3), whereas CCS class III or IV symptoms decreased from 23% to 2% at follow-up (p < 0.001).
Doppler hemodynamic data and LVEF in patients who had AVR
The LVEF was assessed by echocardiography at a mean interval of 21 months after AVR in 53 (83%) of 64 in-hospital survivors. The overall change in LVEF was from 21 ± 7% to 30 ± 12% (p = 0.001). In the patients in whom LVEF was assessed after AVR, 35 (66%) of 53 showed an increase of 15 ± 9% in LVEF. After multiple linear regression analysis, the presence of syncope (p = 0.02) and the absence of hypertension remained the only independent predictors of a postoperative increase in LVEF (p = 0.04).
Postoperative TVGs after AVR were recorded among 52 (81%) of 64 in-hospital survivors. The mean TVG decreased from 25 ± 4 mm Hg to 13 ± 5 mm Hg, and the peak TVG decreased from 42 ± 7 mm Hg to 23 ± 9 mm Hg (p < 0.001). Among the 42 patients with a bioprosthesis, the postoperative mean TVG was less than that of the 10 patients with a mechanical valve (12 ± 4 mm Hg vs. 17 ± 8 mm Hg, p = 0.007).
We report the first study comparing survival among patients with well-defined, low TVGs with severe AS and severe LV systolic dysfunction who did and did not receive AVR. Despite similar propensity scores and baseline characteristics among patients in the AVR and control groups, patients who did not have AVR were at increased risk of death over the short to medium term.
The benefit of AVR for severe AS, clinical heart failure and LV dysfunction was first demonstrated in the 1970s in a series of 19 patients (4). The benefit of AVR, despite the presence of LV dysfunction, was confirmed in a larger series of 154 patients with an elevated TVG who underwent AVR (6). Few series have assessed the outcome exclusively among patients with severe AS with a low mean TVG (9,12)and with severe LV dysfunction (12). Connolly et al. presented the results of 52 patients who underwent AVR from 1985 to 1995, with an AVA of 0.7 ± 0.2 cm2(range 0.3 to 1.2 cm2), mean aortic valve gradient of 23 ± 4 mm Hg and LVEF of 26 ± 8%. Similar to an earlier study that consisted of 18 patients exclusively in NYHA functional class III or IV with low-gradient, severe AS (9), there was considerable perioperative mortality (21%). The overall three-year survival rate was 62%, but in the absence of CAD, there was a much better outcome, with three- and five-year survival rates of 71%. Multivariable analysis identified a small prosthesis size as the only predictor of perioperative mortality. In both these studies (9,12), there was an improvement in functional class.
In comparison, there have been no studies that have specifically assessed survival in patients with severe AS with severe LV dysfunction and a low TVG who did not receive AVR. In a nonrandomized study of symptomatic patients with severe AS with high gradients and preserved LVEF, survival at three years was poor (21%) in those patients who refused AVR, as compared with 87% in those who underwent AVR (3).
Survival benefit among propensity-matched patients who had AVR
Because the decision to perform AVR was not based on a randomized assignment, we used propensity analysis (26)to account for presumed confounding and selection biases. This method, including its advantages and limitations, has been discussed in detail elsewhere (30,31)and has already been used to assess a number of cardiovascular interventions (32,33). Among the 95 propensity-matched patients, survival among patients who did not receive AVR remained poor. In contrast, receiving AVR was the strongest predictor of survival, decreasing all-cause mortality by 81%.
Possible mechanisms of perioperative survival
The perioperative survival among patients in the AVR group was better than that of previously reported series (9,12). This relates, in part, to the more recent period of our study, as there have been advances in surgical techniques, improvements in valve prostheses and related hemodynamic variables, anesthetic monitoring and the use of new inotropes, such as phosphodiesterase inhibitors (34).
There was considerable improvement in valvular hemodynamic data among those in-hospital survivors who received postoperative echocardiography. The use of bovine-pericardial tissue bioprostheses that have a low TVG, especially with smaller sized prostheses (35,36), may have contributed to this improvement in postoperative hemodynamic data.
It is also possible the patients selected for AVR consisted of a less sick cohort of patients, as compared with the patients in other studies. At baseline in our study, the prevalence of NYHA functional class III or IV symptoms was 65%, compared with 85% reported in earlier series of patients with severe AS and severe LV dysfunction (11,12). This indicates that some patients were referred for surgery owing to the presence of severe LV dysfunction, before developing incapacitating symptoms. In addition, the improvement in late survival, as well as NYHA functional class, of patients in the AVR group may have been partly due to current advances in the medical management of patients with LV dysfunction (37,38). We did not find a mean TVG below which survival deteriorated; however, as seen in Table 4, the number of patients with a TVG ≤25 mm Hg was too small to draw any conclusions.
Change in postoperative LVEF
Among patients with severe AS, elevated TVG and LV dysfunction, AVR is associated with improved postoperative LVEF (4,6,12,39). This improvement reflects myocardial reserve, with the relief of afterload mismatch.
After multiple linear regression analysis, the only predictors of an increase in LVEF after AVR were the presence of syncope and the absence of hypertension at baseline. Although the occurrence of syncope is strongly associated with severe AS, it had a low prevalence and was found only among 8 (15%) of 53 patients. Structural remodeling of the myocardium, with an abnormal accumulation of collagen (40), can result in myocardial fibrosis (41)in patients with hypertension. This may have contributed to persistent myocardial dysfunction after the operation in those patients in the AVR group who did not have improvement in postoperative LVEF. Hypertension is also known to be a risk factor for the development CAD; however, neither the absence of CAD nor multivessel CAD predicted an improvement in postoperative LVEF in our study. Other studies have found that AVA and female gender (6,42)predicted increased postoperative LVEF.
This study was observational in design, and not all the factors that influenced the decision to refer or not to refer patients for surgery are known. It is likely that the patients in the AVR group were expected to benefit most, because of the presence of anatomically severe stenosis and, likely, less severely detrimental comorbidities, as compared with patients in the control group. Although we undertook a detailed propensity analysis to correct for these possible biases, it is possible there were factors not captured among patients in the control group that affected survival. This may have led to an overestimation of the benefit of AVR surgery.
Dobutamine stress echocardiography aids in predicting the benefit of AVR among this high-risk cohort with or without CAD, by determining whether a fixed stenosis is present (43). However, preoperative dobutamine stress echocardiography was used in too few patients to assess its predictive accuracy.
This study shows that among patients with low TVG, severe AS and severe LV systolic dysfunction, select patients who receive AVR are associated with significantly improved survival, as compared with those who did not receive AVR. It also confirms previous reports (9,12)demonstrating that survival is accompanied by an improvement in functional status. Although these results are promising, they are still observational and derived from a small number of patients. Thus, they can only be regarded as hypothesis-generating. It will be important for them to be reproduced among a large population of patients and in other series, before AVR becomes an accepted treatment strategy for this high-risk cohort.
☆ This study was supported in part by grant no. NCC9-60 from the National Aeronautics and Space Administration, Houston, Texas. Dr. Lauer receives funding from the American Heart Association (Established Investigator grant no. 0040244N) and from the National Heart, Lung and Blood Institute (grant no. HL66004-01), National Institutes of Health, Bethesda, Maryland.
- aortic stenosis
- aortic valve replacement
- aortic valve area
- coronary artery bypass graft surgery
- coronary artery disease
- Canadian Cardiovascular Society
- confidence interval
- left ventricular
- left ventricular ejection fraction
- New York Heart Association
- risk ratio
- transvalvular gradient
- Received August 28, 2000.
- Revision received November 28, 2001.
- Accepted January 30, 2002.
- American College of Cardiology Foundation
- Schwarz F.,
- Baumann P.,
- Manthey J.,
- et al.
- Smith N.,
- McAnulty J.H.,
- Rahimtoola S.H.
- Carabello B.A.,
- Green L.H.,
- Grossman W.,
- Cohn L.H.,
- Koster J.K.,
- Collins J.J. Jr.
- Connolly H.M.,
- Oh J.K.,
- Orszulak T.A.,
- et al.
- Lund O.
- Brogan W.C.,
- Grayburn P.A.,
- Lange R.A.,
- Hillis L.D.
- Connolly H.M.,
- Oh J.K.,
- Schaff H.V.,
- et al.
- Boyle C.A.,
- Decoufle P.
- Curb J.D.,
- Ford C.E.,
- Pressel S.,
- Palmer M.,
- Babcock C.,
- Hawkins C.M.
- Enriquez-Sarano M.,
- Tajik A.J.,
- Schaff H.V.,
- Orszulak T.A.,
- Bailey K.R.,
- Frye R.L.
- Schiller N.B.,
- Shah P.M.,
- Crawford M.,
- et al.
- Helmcke F.,
- Nanda N.C.,
- Hsiung M.C.,
- et al.
- Perry G.J.,
- Helmcke F.,
- Nanda N.C.,
- Byard C.,
- Soto B.
- Hosmer D.,
- Lemeshow S.
- Cox D.
- Joffe M.M.,
- Rosenbaum P.R.
- Yamani M.H.,
- McCarthy P.M.,
- Starling R.C.
- Kennedy J.W.,
- Doces J.,
- Stewart D.K.
- Querejeta R.,
- Varo N.,
- Lopez B.,
- et al.
- Pardo Mindan F.J.,
- Panizo A.