Introduction

Aortic stenosis (AS) is a slowly progressive disease and survival during the asymptomatic phase of AS is similar to that of age-matched controls with a low risk of sudden death when patients are followed prospectively and promptly report symptom onset (1–3). The potential benefits of aortic valve replacement (AVR) in asymptomatic patients with severe AS have not been thought to outweigh the operative mortality of AVR (4,5). Thus, current guidelines generally recommend a strategy of watchful waiting until symptoms emerge for AVR in asymptomatic patients with severe AS (6).

However, the recommendation is based on previous small single-center studies that sought to evaluate symptoms and/or AVR, rather than mortality as a primary outcome measure (1–3,7). A multicenter study design seemed to be crucial for extrapolating results into real clinical practice, because the quality of echocardiographic examination, the manner of patient follow-up, and the operative mortality of AVR might vary between centers. Furthermore, there is no large-scale study comparing an initial AVR strategy with the conservative strategy in asymptomatic patients with severe AS except for 1 small single-center observational study in patients with very severe AS (8). Also, patient demographics, age, and operative mortality of AVR in contemporary clinical practice may be different from those reported in previous studies (1–3,9–13). There is an obvious clinical need to evaluate the balance between risks and benefits of AVR in asymptomatic patients with severe AS in contemporary clinical practice.

Therefore, we sought to compare an initial AVR strategy with the conservative strategy to assess long-term outcomes of asymptomatic patients enrolled in a large Japanese multicenter registry of consecutive patients with severe AS.

Methods

Study population

The CURRENT AS (Contemporary outcomes after sURgery and medical tREatmeNT in patients with severe Aortic Stenosis) registry is a retrospective, multicenter registry enrolling consecutive patients with severe AS among 27 centers (of which 20 had an on-site surgical facility) in Japan between January 2003 and December 2011. We searched the hospital database of transthoracic echocardiography, and enrolled consecutive patients who met the definition of severe AS (peak aortic jet velocity [V_max] >4.0 m/s, mean aortic pressure gradient [PG] >40 mm Hg, or aortic valve area [AVA] <1.0 cm²) for the first time during the study period (6). We excluded patients with a history of aortic valve repair/replacement/plasty or percutaneous aortic balloon valvuloplasty. The institutional review boards in all 27 participating centers (Online Appendix) approved the protocol. Written informed consent from each patient was waived in this retrospective study, because we used clinical information obtained in routine clinical practice, and no patients refused to participate in the study when contacted for follow-up.

Among 3,815 patients enrolled in the registry, there were 2,005 patients with and 1,808 patients without AS-related symptoms at the time of index echocardiography, excluding 2 patients whose symptomatic status was not available. In this primary report from the CURRENT AS registry, 1,808 asymptomatic patients were subdivided into the initial AVR group (n = 291) and the conservative group (n = 1,517) according to treatment strategy selected after the index echocardiography (Figure 1). Baseline characteristics and 5-year clinical outcomes were compared between the initial AVR and conservative groups. Because the selection of initial AVR was determined by physicians, and the characteristics were highly different between groups, we developed a propensity score–matched cohort of 582 patients (evenly divided between initial AVR and the conservative strategy) as the main analysis set for the current report (Figure 1, Online Figure 1). We also analyzed the entire cohort of asymptomatic AS patients to explore the robustness of our analyses. Initial AVR and initial conservative strategies were compared by intention-to-treat analysis regardless of the actual performance of AVR. Follow-up commenced on the day of index echocardiography.

All patients underwent a comprehensive 2-dimensional and Doppler echocardiographic evaluation in each participating center. Maximal velocity (V_max) and mean aortic PG were obtained with the use of the simplified Bernoulli equation. AVA was calculated using the standard continuity equation, and indexed to body surface area (14).

Data collections and definitions

Collection of baseline clinical information was conducted through hospital chart or database review. Angina, syncope, or heart failure (HF) including dyspnea were regarded as AS-related symptoms. Follow-up data were mainly collected through review of hospital charts or collected through contact with patients, relatives, and/or referring physicians using mail with questions regarding survival, symptoms, and subsequent hospitalizations.

For the current analysis, primary outcome measures were all-cause death and HF hospitalization. Cause of death was classified according to VARC (Valve Academic Research Consortium) definitions, and adjudicated by a clinical event committee (Online Appendix) (15,16). Sudden death was defined as unexplained death in previously stable patients. Aortic valve–related death included aortic procedure–related death, sudden death, and death due to HF possibly related to aortic valve. HF hospitalization was defined as hospitalization due to worsening HF requiring intravenous drug therapy. Other clinical event definitions are described in the Online Appendix.

Statistical analysis

Categorical variables are presented as numbers and percentages, and compared using the chi-square test or the Fisher exact test. Continuous variables were expressed as the mean ± SD or median (interquartile range ([IQR]). Based on their distributions, continuous variables were compared using the Student t test or Wilcoxon rank sum test. We used the Kaplan-Meier method to estimate cumulative incidence and assessed the differences with the log-rank test.

A logistic regression model was used to develop propensity-score for the choice of initial AVR with 15 independent variables relevant to the choice of initial AVR listed in Table 1. The C statistic was 0.783 and the coefficients of the independent variables are shown in Online Table 1. Propensity score was calculated by summing up all coefficients multiplies corresponding variables. To create the propensity score–matched cohort, patients in the conservative group were matched to those in the initial AVR group using a 1:1 greedy matching technique (17). We then calculated the cumulative incidence using the propensity score–matched cohort. Not all relevant variables were well matched, probably due to the many patients in the conservative group. Therefore, we conducted an adjusted analysis using Cox proportional hazard models with the risk-adjusting variables of age, dyslipidemia, malignancy currently under treatment, EuroSCORE II, and Society of Thoracic Surgeons (STS) Predicted Risk of Mortality (PROM), as a sensitivity analysis in the propensity score–matched cohort. For the secondary analysis among the entire cohort of 1,808 asymptomatic patients, the 21 clinically relevant factors listed in Table 1 were included in the Cox proportional hazards models as the risk adjusting variables and the centers were incorporated as the stratification variable. With the exception of age, continuous variables were dichotomized by median values or clinically meaningful reference values. Because the difference in age between the 2 groups was too large to allow the dichotomous approach, we treated age as a continuous variable in the Cox proportional hazards models. Proportional hazards assumptions for the risk-adjusting variables including the categorized age in quartiles, were assessed on the plots of log (time) versus log [-log (survival)] stratified by the variable, and verified to be acceptable. The risks of an initial AVR strategy relative to conservative strategy for the clinical endpoints were expressed as hazard ratios (HRs) and their 95% confidence intervals (CIs).

All statistical analyses were conducted by a physician (T.T.) and a statistician (T.M.) with the use of JMP 10.0.2 or SAS 9.4 (both SAS Institute Inc., Cary, North Carolina). All reported p values were 2-tailed, and p values <0.05 were considered statistically significant.

Results

Baseline characteristics were significantly different between the initial AVR and conservative groups before matching (Table 1). Patients in the conservative group were much older (77.8 ± 9.4 years vs. 71.6 ± 8.7 years, p < 0.001) and more often had prior symptomatic stroke, atrial fibrillation or flutter, malignancy currently under treatment, chronic lung disease, and coronary artery disease. Patients in the initial AVR group, on the other hand, had greater AS severity than those in the conservative group. There were 247 patients who were regarded as ineligible for AVR by the attending physicians, although the decision regarding the ineligibility for AVR was not uniform in this retrospective study. Physicians had 1 or more reasons for ineligibility: extreme old age in 170 patients; reduced cognitive function in 53 patients; serious comorbid conditions that limit life expectancy in 43 patients; renal failure in 26 patients; muscle weakness in 26 patients; anatomic factors that precluded or increased the risk of cardiac surgery, such as a porcelain aorta, prior radiation, or an arterial bypass graft in 22 patients; reduced pulmonary function in 14 patients; malnutrition in 10 patients; and severe liver cirrhosis in 2 patients.

Among 291 patients referred for AVR despite absence of symptoms related to AS, 184 (63%) patients had 1 or more formal indications for AVR; very severe AS (V_max ≥5.0 m/s or mean aortic PG ≥60 mm Hg) in 118 patients (41%), left ventricular dysfunction (defined as left ventricular ejection fraction [LVEF] <50%) in 19 patients (7%), candidates for other cardiac surgery in 24 patients (8%), rapid hemodynamic progression in 32 patients (11%), and active infective endocarditis in 1 patient (0.3%).

Baseline characteristics of the initial AVR and conservative groups in the propensity score–matched cohort were much more comparable than those in the entire cohort except for the slightly younger age, a slightly lower STS score, and the greater AS severity in the initial AVR group (Table 1). Patients in the matched cohort had a mean age in the early 70s, and relatively low STS PROM score. Regarding classification of severe AS, 43 patients (15%) in the initial AVR group and 111 patients (38%) in the conservative group were included on the basis of AVA <1.0 cm² alone (LVEF ≥50%) with less severe V_max and mean aortic PG (Table 1).

Clinical outcomes in the propensity score–matched cohort

In the initial AVR group, surgical AVR was actually performed in 286 (98%) patients. One patient underwent transcatheter AVR with a median interval of 44 days from the index echocardiography (Table 2, Central Illustration). In the remaining 4 patients, 2 patients died suddenly, another died of respiratory failure while awaiting AVR, and the fourth was lost to follow-up on day 15. The 30-day mortality rate after AVR was 1.4% in the initial AVR group.

Among 291 patients in the conservative group, AVR was performed in 118 patients (41%) during follow-up with a median interval of 780 days from index echocardiography (Central Illustration). Among 116 patients (40%) with emerging symptoms related to AS during follow-up in the conservative group, AVR was performed in 80 patients (69%) with median interval of 72 (IQR: 42 to 121) days after symptom onset. AVR was performed in 30 patients (67%) of 45 who presented with New York Heart Association functional class III or IV HF (Online Table 2).

The cumulative 5-year incidence of all-cause death was significantly lower in the initial AVR group than in the conservative group (15.4% vs. 26.4%, p = 0.009) (Table 2, Central Illustration). The cumulative 5-year incidences of cardiovascular death and aortic valve–related death were also significantly lower in the initial AVR group than in the conservative group (9.9% vs. 18.6%, p = 0.01; and 5.3% vs. 13.5%, p = 0.003, respectively). The cumulative 5-year incidence of sudden death trended lower in the initial AVR group than in the conservative group (3.6% vs. 5.8%, p = 0.06). Among 46 patients with cardiovascular death in the conservative group, HF (9 patients who did not undergo AVR despite symptoms) and sudden death (8 patients who did not undergo AVR despite symptoms, and 10 patients without symptoms) were the dominant causes of cardiovascular death (Online Table 3). The initial AVR strategy as compared with the conservative strategy was also associated with markedly lower cumulative 5-year incidences of emerging symptoms related to AS and HF hospitalization (3.2% vs. 46.3%, p < 0.001; and 3.8% vs. 19.9%, p < 0.001, respectively) (Table 2, Central Illustration, Figure 2). The results from the adjusted analysis conducted as a sensitivity analysis were fully consistent with those from the unadjusted analysis (Table 2).

Crude clinical outcomes in the entire cohort

Among 1,808 asymptomatic patients with severe AS, median follow-up interval after the index echocardiography was 1,361 (IQR: 1,055 to 1,697) days with 90% follow-up rate at 2-year. Among 1,517 patients in the conservative group, follow-up information was collected from the hospital charts of the participating centers in 1,311 patients (86.4%). AVR was performed in 392 patients (26%) during follow-up with a median interval of 788 days (Table 3, Online Figure 2). The 30-day mortality rate after AVR was 2.6% in the conservative group, which tended to be higher than that in the initial AVR group, but without significant difference (p = 0.29). Among 492 patients with emerging symptoms related to AS during follow-up in the conservative group, AVR was performed in 239 patients (49%) with median interval of 70 (IQR: 41 to 131) days after onset of symptoms. AVR was performed in only 74 of 201 (37%) patients presenting with New York Heart Association functional class III or IV HF. Among 127 patients who presented with New York Heart Association functional class III or IV HF, but who did not undergo AVR, 96 (76%) patients died with a median interval of 95 (IQR: 11 to 467) days after symptom onset (Online Table 2). Among 679 patients who underwent AVR in the present study, AVR after symptom development during follow-up (n = 247, including 8 patients who became symptomatic at time of AVR) was associated with higher 30-day operative mortality than AVR while asymptomatic (n = 432; 3.7% vs. 1.2%, p = 0.03).

Cumulative 5-year incidence of all-cause death was significantly lower in the initial AVR group than in the conservative group (Table 3, Online Figure 2). Among 582 (32%) patients who died during follow-up, HF and sudden death were the predominant causes of death in the conservative group (101 and 82 patients, respectively), while those were uncommon causes of death in the initial AVR group (1 and 8 patients, respectively) (Online Table 4). The cumulative 5-year incidence of HF hospitalization was also significantly lower in the initial AVR group (Table 3, Online Figure 2). The cumulative 5-year incidence of sudden death was 7.6% (1.5%/year) in the conservative group versus 3.6% (0.7%/year) in the initial AVR group. Among 82 patients experiencing sudden death in the conservative group, 57 patients (70%) died abruptly without preceding symptoms and 32 (56%) of these sudden deaths occurred within 3 months of the last clinical follow-up visit.

The lower cumulative incidences of all-cause death and HF hospitalization in the initial AVR group compared to the conservative group were consistently seen in the 2 subgroups of patients with or without current recommendations for AVR such as very severe AS at low surgical risk or severe AS with left ventricular dysfunction (Online Figure 3).

Adjusted clinical outcomes in the entire cohort

The favorable effect of an initial AVR strategy for the clinical outcomes was similarly seen in the adjusted analysis of the entire cohort as well as in the propensity score–matched analysis, although the effect size was smaller in the propensity score–matched cohort than in the entire cohort (Table 3). The lower risks of an initial AVR strategy relative to the conservative strategy for all-cause death and HF hospitalization were consistently seen in the 2 subgroups of patients with or without current recommendations for AVR (Online Figure 3).

Discussion

The main finding of this study was that compared to a conservative “watchful waiting” approach, an initial AVR strategy was associated with lower risk for all-cause death and HF hospitalization in asymptomatic patients with severe AS in a propensity score–matched analysis (Central Illustration).

Although AVR is strongly recommended in symptomatic patients with severe AS who are candidates for surgery, management of asymptomatic patients with severe AS remains controversial. Current guidelines recommend a strategy of watchful waiting until symptoms emerge for AVR in asymptomatic patients with severe AS except for patients with left ventricular dysfunction, very severe AS, need for other cardiac surgery, or an abnormal exercise test (6,18). However, the current large-scale multicenter propensity score–matched analysis clearly demonstrated the benefits of an initial AVR strategy in reducing mortality and HF hospitalization as compared with a conservative strategy. The extent of benefit appeared to be similar regardless of the current indications for AVR such as left ventricular dysfunction or very severe AS.

Several important issues should be considered regarding the clinical relevance of the watchful waiting strategy for AVR. First, and most importantly, the current recommendations for AVR are mainly dependent on the patient symptoms. However, many patients with severe AS who could potentially benefit from AVR may not complain any symptoms because of their sedentary lifestyle. It is often difficult to distinguish the nonspecific symptoms such as fatigue and dyspnea on exertion from the true symptoms of AS. In the asymptomatic patients with severe AS, an exercise stress test is recommended to confirm both their asymptomatic status and hemodynamic response to exercise (6,19). However, exercise testing is not commonly performed in real clinical practice due to safety concerns or cannot be performed in many patients because of advanced age, limited exercise capacities, and comorbidities.

Additionally, prompt detection of symptoms during follow-up is not always possible in clinical practice (6). Patients may not always be compliant to close clinical follow-up (20). Indeed, in the current study, severe HF was the initial symptom in a sizable proportion of patients in the conservative group, for whom AVR was less frequently performed than in patients without severe HF, and mortality was extremely high if AVR was not performed. It is noteworthy that an initial AVR strategy was associated with markedly lower risk for HF hospitalization than the conservative strategy; HF hospitalization should be regarded as a very serious clinical event in patients with severe AS. Also, the present study and a prior large scale surgical report suggested that AVR after symptom development carries higher operative risk than AVR during the asymptomatic phase, although Brown et al. (21) reported similar AVR operative mortality between symptomatic and asymptomatic patients from a single center study (4).

The annual rate of sudden death in the conservative group (1.5% in the current study) suggests that the rate of sudden death during the asymptomatic phase might be higher than previously reported (<1.0%/year) (2,3,9). Finally, in the present study, 41% of patients managed conservatively required AVR within a median follow up of 2 years, suggesting little gain from waiting. Balancing the risks of watchful waiting with the improvement in operative mortality, AVR during the asymptomatic phase may be a viable treatment option in severe AS patients at low-risk for AVR.

Study limitations

In that this was a retrospective study, we could not exclude the possibility of ascertainment bias for symptoms related to AS at baseline. However, we thoroughly reviewed all patient charts and referred to the hospital database to evaluate symptomatic status. Also, we cannot deny residual confounding and selection bias in the comparison between initial AVR and conservative strategies, although the characteristics of the 2 groups were largely comparable after propensity-score matching. Actually in the present analysis, propensity-score matching did not completely eliminate the impact of differences in age as well as EuroSCORE and STS score in the 2 populations. However, results from the adjusted analysis conducted as a sensitivity analysis were fully consistent with those from the unadjusted analysis.

Some patients also were included in this study on the basis of AVA <1.0 cm² alone (LVEF ≥50%) with less severe V_max and mean aortic PG. The imprecision of assessing AVA by echocardiography with overestimation of AS severity in some patients is a possible concern. However, the proportion of those patients included on the basis of AVA <1.0 cm² alone was greater in the conservative group than in the initial AVR group. Clinical outcomes were worse in the conservative group than in the initial AVR group despite more frequent inclusion of somewhat less severe AS in the former group.

Finally, patient follow-up in this multicenter study might have been less close than in the previous single-center studies, therefore, emerging symptoms related to AS might have been underestimated. However, follow-up information in the conservative group was collected from the hospital chart of participating centers for 86.4% of patients, suggesting that the majority of patients were followed by cardiologists. It is important to note that even with follow-up by cardiologists, presentation with severe HF during follow-up was not uncommon in clinical practice.

Conclusions

The long-term outcome of asymptomatic patients with severe AS was dismal when managed conservatively in real-world clinical practice and might be substantially improved by an initial AVR strategy.

Footnotes