PI indicates prediction interval.
eAppendix 1. Details on data extraction
eAppendix 2. Details on Bayesian meta-analysis of prognostic indicators
eFigure. Flow-chart of systematic literature search and study inclusion
eTable 1. Subgroup analyses
eTable 2. Predictors from multivariable analyses
eTable 3. Sensitivity analysis of predictors of death or aortic valve intervention according to a half-normal prior distribution of τ
eTable 4. Additional information on studies comparing early surgery vs conservative treatment strategy
eTable 5. Ongoing randomized trials evaluating early intervention vs conservative treatment
Customize your JAMA Network experience by selecting one or more topics from the list below.
Gahl B, Çelik M, Head SJ, et al. Natural History of Asymptomatic Severe Aortic Stenosis and the Association of Early Intervention With Outcomes: A Systematic Review and Meta-analysis. JAMA Cardiol. 2020;5(10):1102–1112. doi:10.1001/jamacardio.2020.2497
What is the natural history of asymptomatic severe aortic stenosis, which variables predict prognosis, and can early intervention improve outcomes?
In this systematic review and meta-analysis of 29 studies with 4075 patients with 11 901 years of follow-up, the rate of all-cause death was 5 per 100 conservatively treated patients per year, of which 3 and 1 were of cardiac and sudden cause, respectively. Twenty per 100 patients per year developed an indication for intervention; early intervention was significantly associated with improved survival.
Patients with asymptomatic severe aortic stenosis may develop indication for intervention and have deaths that are mostly cardiac but not only sudden.
Whether intervention should be performed in patients with asymptomatic severe aortic stenosis (AS) remains debated.
To meta-analyze the natural history of asymptomatic severe AS and examine the association of early intervention with survival.
PubMed, Embase, and Cochrane databases were searched from inception to February 1, 2020.
Observational studies of adult patients with asymptomatic severe AS.
Data Extraction and Synthesis
Two investigators independently extracted study and patient characteristics, follow-up time, events, and prognostic indicators of events. Random-effects models were used to derive pooled estimates.
Main Outcomes and Measures
The meta-analysis on natural history was performed on the primary end point of all-cause death occurring during a conservative treatment period, with secondary end points consisting of cardiac death, death due to heart failure, sudden death, development of symptoms, development of an indication for aortic valve intervention, and aortic valve intervention. The primary end point for the meta-analysis of early intervention vs a conservative strategy was all-cause death during long-term follow-up. Finally, meta-analysis was performed on the association of prognostic indicators with the composite of death or aortic valve intervention found in multivariable models.
A total of 29 studies with 4075 patients with 11 901 years of follow-up were included. Pooled rates per 100 patients per year were 4.8 (95% CI, 3.6-6.4) for all-cause death, 3.0 (95% CI, 2.2-4.1) for cardiac death, 2.0 (95% CI, 1.3-3.1) for death due to heart failure, 1.1 (95% CI, 0.6-2.1) for sudden death, 18.1 (95% CI, 12.8-25.4) for an indication for aortic valve intervention, 18.5 (95% CI, 13.4-25.5) for development of symptoms, and 19.2 (95% CI, 15.5-23.8) for aortic valve intervention. Early intervention was associated with a significant reduction in long-term mortality (hazard ratio, 0.38; 95% CI, 0.25-0.58). Factors associated with worse prognosis were severity of AS, low-flow AS, left ventricular damage, and atherosclerotic risk factors.
Conclusions and Relevance
Data from observational studies and a recent randomized clinical trial suggest that many patients with asymptomatic severe AS develop an indication for aortic valve intervention, and their deaths are mostly cardiac but not only sudden. Other end points besides sudden death should be considered during the decision to perform early intervention that are associated with improved survival.
Patients with symptomatic severe aortic stenosis (AS) have an indication for surgical aortic valve replacement (SAVR) or transcatheter aortic valve replacement. The role of intervention is less clear in patients with asymptomatic severe AS. North American and European guidelines agree on a class I indication for SAVR in patients with a reduced left ventricular (LV) ejection fraction (<50%) but are inconsistent for patients with other disease or comorbid factors.1-3
Studies suggest that as many as 50% of patients with asymptomatic severe AS progress to a symptomatic status and require surgery within the first 2 years of follow-up4 and that this waiting period increases the risk of sudden cardiac death and congestive heart failure.5,6 In light of these results, the concept of early intervention has raised increasing interest.5,7 However, advocates of a conservative approach argue that the procedural risk does not balance against the potential benefits of early intervention and that many patients will never become symptomatic.8 Such arguments come mainly from single-center observational studies with few patients and based on events that occur infrequently.1
The natural history should be better quantified to improve our understanding of potential benefits and harms of intervention vs conservative treatment. Moreover, risk factors of poor prognosis should be identified to evaluate which patients are at highest risk and may particularly benefit from early intervention. Therefore, we have performed a systematic review and meta-analysis of studies evaluating the natural history of patients with asymptomatic severe AS and determined whether early intervention improves long-term survival.
The PubMed, Embase, and Cochrane databases were searched from their inception to February 1, 2020, for full-length, English-language, observational studies that reported on patients with asymptomatic severe AS who were initially treated conservatively. We searched among titles and abstracts using the keywords asymptomatic AND aortic AND stenosis. No search software was used. Authors were not contacted for studies that did not fulfill inclusion criteria or if data were unclear. This study complies with the Meta-analysis of Observational Studies in Epidemiology (MOOSE) reporting guideline.9
Two investigators (S.J.H. and M.Ç.) independently reviewed the search result in duplicate. In case of disagreement, consensus was reached through discussion. The title and abstract were reviewed during the first stage, after which the remaining articles were reviewed in depth during the second stage. Reference lists of potentially valid studies and review articles were checked to ensure no relevant studies were missed. Abstracts from meetings were not considered.
Studies were included if they fulfilled the following criteria: (1) the study included adult patients with severe AS quantified by at least an aortic valve area of less than 1.0 cm2 or an indexed aortic valve area less than 0.6 cm2/m2, a jet velocity of more than 4.0 m/s, or a mean gradient of more than 40 mm Hg; (2) patients were considered to be asymptomatic if reported as such, which was left to the discretion of the physicians and investigators of the individual studies and performance of exercise testing was not considered mandatory to confirm absence of symptoms; and (3) at least the event of death during follow-up and the mean/median duration of follow-up was reported. Studies with a combined inclusion of patients with moderate and severe AS were excluded unless results were separately reported for patients with severe AS. In case there was overlap in the patient populations in different studies from the same center, we included only the study with the longest follow-up or largest patient cohort. A list of excluded articles is available on request.
Two investigators (S.J.H. and M.Ç.) independently extracted and crosschecked clinically relevant data and data necessary for study inclusion and meta-analysis (eAppendix 1 in the Supplement). Inconsistencies were resolved by discussion.
For the meta-analysis on the natural history, the primary end point was all-cause death. Secondary end points consisted of cardiac death, sudden death, death due to congestive heart failure, the development of an indication for aortic valve intervention, the development of symptoms, and aortic valve intervention by either SAVR or transcatheter aortic valve replacement. For the meta-analysis of early intervention vs conservative treatment, the primary end point was all-cause death. For the meta-analysis of predictors, the primary end point consisted of the composite of all-cause death and aortic valve intervention (or development of symptoms) but allowing for studies to include hospitalization or congestive heart failure as additional end point in the composite.
We calculated the log rate of events per 100 patients per year of observation time and the corresponding standard error within studies and then used a DerSimonian and Laird random-effects model to derive pooled estimates and corresponding limits of the 95% CI10 and back-transformed pooled estimates and limits of the 95% CI to rates per 100 patient-years throughout. If the total amount of follow-up time was not reported, this was calculated by multiplying the number of patients by the mean follow-up time. In case of 0 events, we derived the upper end of the 95% CI of the rate as described by Hanley and Lippman-Hand, adding a continuity correction of 0.01 to the numerator, and a continuity correction of 0.01 multiplied by the mean follow-up time to the denominator to derive rates.11 We explored heterogeneity across studies using the DerSimonian and Laird between-study variance τ2 statistic12 and calculated 95% prediction intervals for the pooled rates in addition to conventional CIs taking into account the between-study variance to reflect residual uncertainty.13 Our analysis on the natural history consisted of pooling the studies that reported events occurring only during a period of time in which patients were asymptomatic and no aortic valve intervention took place. Prespecified subgroup analyses were restricted to the 26 studies with follow-up until aortic valve intervention, investigating heterogeneity by study design (prospective vs retrospective), year of initiation of patient recruitment (before 1999 vs 1999 or later), number of patients included in the study (<100 vs ≥100 patients), length of mean follow-up time (<2 vs ≥2 years), length of accumulated follow-up patient-time (<200 vs ≥200 patient-years), and whether or not good LV ejection fraction (defined as ≥50%, ≥55%, or normal) was an inclusion criterion of the study. Subgroup analyses were accompanied by a test for interaction from random-effects meta-regression.
For the comparison of all-cause mortality following early intervention vs conservative treatment, we included studies that did not censor patients at the time of intervention and evaluated long-term mortality. We pooled studies using the study-level hazard ratios (HRs) in a random-effects model with Knapp-Hartung modification of the variance as the number of cohort studies that reported HRs for this comparison was low.
For the pooling of the effect of prognostic indicators on events, whenever 2 or more studies reported the HRs of the association between prognostic indicators and events during follow-up, we pooled them across studies using a random-effects bayesian meta-analysis. Details are provided in eAppendix 2 in the Supplement. Analyses were performed in Stata version 14.2 (StataCorp) and WinBUGS version 14 (Medical Research Council Biostatistics Unit).
The literature search yielded 2370 studies that were potentially relevant for inclusion in the meta-analysis, and 29 studies were included in the meta-analysis on the natural history (eFigure in the Supplement). All studies were observational. A total of 4075 patients with a median (interquartile range [IQR]) follow-up of 2.3 (1.6-3.3) years were included in the natural history analysis (Table 1). In addition, 9 studies were included in the meta-analysis comparing an early surgical treatment strategy with watchful waiting, of which 1 was a randomized clinical trial (eFigure in the Supplement). A total of 3904 patients with a median (IQR) follow-up of 5.0 (3.7-5.7) years were included in our analyses comparing an early surgical treatment strategy with watchful waiting (Table 1).
The rate of all-cause death was 4.8 (95% CI, 3.6-6.4) per 100 patients per year in 21 studies with 3041 patients with a median (IQR) follow-up of 2.3 (1.7-3.4) years (Figure 1A). Cardiac death occurred at a rate of 3.0 (95% CI, 2.2-4.1) per 100 patients per year in 18 studies with 2813 patients with a median (IQR) follow-up of 2.1 (1.4-2.9) years (Figure 1B). The rate of death due to congestive heart failure was 2.0 (95% CI, 1.3-3.1) per 100 patients per year in 11 studies with 1809 patients with a median (IQR) follow-up of 2.3 (1.9-2.9) years (Figure 1C). Sudden death occurred at a rate of 1.1 (95% CI, 0.6-2.1) per 100 patients per year in 12 studies with 1767 patients with a median (IQR) follow-up of 2.3 (1.7-3.1) years (Figure 1D).
An indication for aortic valve intervention was reported in 11 studies with 1754 patients with a median (IQR) follow-up of 2.3 (1.8-3.2) years and occurred in 18.1 (95% CI, 12.8-25.4) per 100 patients per year (Figure 2A). There were 16 studies with 2234 patients and median (IQR) follow-up of 1.9 (1.3-3.1) years that reported the number of patients that developed symptoms, with a pooled rate of 18.5 (95% CI, 13.4-25.5) per 100 patients per year (Figure 2B). Aortic valve intervention was performed in 19.2 (95% CI, 15.5-23.8) per 100 patients per year (21 studies with 3494 patients with a median [IQR] follow-up of 2.3 [1.7-3.0] years) (Figure 2C).
eTable 1 in the Supplement shows results of subgroup analyses. Studies with shorter total follow-up were associated with higher rates of all-cause death. Studies with shorter mean and total follow-up were associated with higher rates of symptom development and aortic valve interventions. Rates of an indication for aortic valve intervention (21.0 [95% CI, 15.8-28.0] vs 10.6 [95% CI, 9.6-11.6] per 100 patients per year; P = .02) and development of symptoms (21.2 [95% CI, 16.2-27.6] vs 8.7 [95% CI, 7.9-9.7] per 100 patients per year; P = .007) were markedly higher in prospective vs retrospective studies. There were no interactions with subgroups by LV ejection fraction.
Fifteen studies performed a multivariable analysis on the composite of death or aortic valve intervention. Outcomes were largely associated with measurements of the severity of AS and LV dysfunction, with clinical factors being limited to atherosclerotic risk factors (eTable 2 in the Supplement). There was inconsistency in how variables and cutoffs were used in multivariable models, but pooling consistent variables with 2 or more results in multivariable analyses resulted in a set of independent variables (Table 2). Heterogeneity was low for all pooled analyses. Results were consistent in sensitivity analyses using different assumptions for the prior distribution of τ (eTable 3 in the Supplement).
There were 9 studies that compared patients who underwent early intervention vs an initial conservative treatment strategy, which included a combined 3904 patients with a median (IQR) follow-up of 5.0 (3.7-5.7) years (eTable 4 in the Supplement).5,39,43,47-49 All but 1 randomized clinical trial used either propensity-score matching or multivariable models to adjust for differences in baseline characteristics between treatment groups. Intervention consisted of surgery in most cases. Our meta-analysis indicates that intervention was associated with a significant reduction in all-cause mortality during follow-up (HR, 0.38; 95% CI, 0.25-0.58), with moderate heterogeneity (τ2 = 0.21) (Figure 3).
In this systematic review and meta-analysis of 29 studies on the natural history of patients with asymptomatic severe AS, we found that there were overall 5 deaths per 100 patients per year during a conservative treatment strategy, with a high rate of progressing to a symptomatic state and developing an indication for aortic valve intervention. Particularly patients with more severe AS, abnormal LV characteristics, and atherosclerotic clinical factors were at a higher risk of death or an indication for intervention. Moreover, among another 9 studies that investigated performing early intervention, consisting of surgery in the majority of cases within these studies, early intervention was associated with a significant reduction in all-cause death during follow-up. While it has been argued that many patients do not develop an indication for intervention and that the risk of death is low during conservative treatment, the results of the current meta-analysis suggest otherwise. Indeed, most studies focus on sudden death, but this meta-analysis demonstrates that sudden death accounts for only part of cardiac deaths that occur in asymptomatic patients with severe AS and that the risk of death may therefore be underestimated. These data suggest that early intervention may need to be considered in a greater proportion of patients with asymptomatic severe AS.
Currently, the largest and only available randomized clinical trial on asymptomatic patients with severe AS analyzed 145 patients and found that initial surgery vs an initial conservative treatment significantly reduced the all-cause death and operative or cardiovascular death, even when 74% of patients in the conservative group required SAVR during follow-up.43 This study is pivotal in the debate on treating asymptomatic patients, but it only provides a perspective on patients with very severe AS, applying inclusion criteria of an aortic valve area of 0.75 cm2 or less with either a jet velocity of 4.5 m/s or more or a mean gradient of 50 mm Hg or more, while lacking evidence on the much broader patient population with asymptomatic AS. Further data from observational studies as summarized in the current meta-analysis provide these additional insights. The largest available observational study analyzed 291 propensity-matched pairs and found that early surgery vs an initial conservative treatment significantly reduced the 5-year rates of all-cause death and hospitalization for heart failure, even when 41% of patients in the conservative group required SAVR during follow-up.5 When pooling multiple studies on the effect of intervention on survival, we found that intervention vs conservative treatment was associated with significantly improved survival with an HR of 0.38. While this may be a true effect, considering the high rates of death and progression to an indication for aortic valve intervention (eg, symptoms or LV dysfunction) among conservatively treated patients in this meta-analysis, most of the observational studies may be biased because physicians could have opted for a conservative treatment strategy for patients owing to a high risk for surgery, as was often the case before the introduction of transcatheter aortic valve replacement, when most of these studies were performed.50 Moreover, not all studies specifically evaluated the effect of intervention within a short (eg, 3 months) period after the diagnosis of severe AS. Patients who went on to have intervention at a later follow-up time are inherently a selected group with a better prognosis because the highest-risk patients may have died within the early follow-up period. Indeed, Le Tourneau and coauthors39 found that the point estimate of the HR in favor of surgery was much larger if conservative treatment was compared with surgery being performed within 1 year of presentation as opposed to surgery at any time during follow-up (HR, 0.58 vs HR, 0.39). Data from the RECOVERY trial are consistent with that of these observational studies,43 but additional results from ongoing randomized clinical trials comparing an early interventional treatment strategy and a conservative strategy in asymptomatic patients with severe AS will add significant knowledge and provide important insight to substantiate the role of early intervention (eTable 5 in the Supplement).
The decision to undergo early intervention should depend on a critical assessment of symptoms and careful and individualized consideration of potential benefits and harms. Cardiac magnetic resonance to detect LV damage furthermore helps identify patients that may benefit from early intervention.51 Apart from LV dysfunction as an indication to perform SAVR in patients with asymptomatic severe AS, current clinical guidelines provide several additional recommendations to consider intervention in patients with asymptomatic severe AS.1 Our meta-analysis of variables associated with mortality-related outcomes indicates that prognosis is significantly worse if global longitudinal strain or valvulo-arterial impedance is present even with a preserved LV function,22,23,28,33 if AS is more severe as measured by higher valve gradient and lower valve area, and if atherosclerotic risk factors, such as dyslipidemia or diabetes, are present. These additional diseases and comorbid characteristics are not considered in current guidelines or are inconsistently recognized in North American and European guidelines. Therefore, we suggest that cardiologists and surgeons take these additional factors into account when deciding to perform early intervention or initiate a conservative treatment strategy. Of note, our subgroup analysis could not confirm that lower LV ejection fraction was associated with worse outcomes, which is most likely related to the criteria used in the individual articles; almost all studies included patients with preserved LV ejection fraction.
An important strength is that a large number of studies could be pooled in a random-effects model with moderate statistical heterogeneity, increasing the validity of the results. The included studies consisted exclusively of patients with asymptomatic severe AS, unlike many other studies and reviews that have not stratified results according to the severity of AS in asymptomatic patients.7,52 Lastly, using bayesian methods for meta-analyses of a low number of studies allowed a more reliable estimation of between-trial variance and its uncertainty to identify particular disease and patient factors that affect the prognosis of asymptomatic severe AS. This resulted in identifying several variables that are currently not included in clinical guidelines.
This is a meta-analysis of observational studies, which is dependent on the quality of the individual studies that were included. Many of the studies were single center and retrospective, and it may therefore have been difficult to adjudicate events related to the development of symptoms and indications for intervention during follow-up. Second, only a few studies routinely performed stress testing in patients with asymptomatic severe AS, and we were therefore not able to determine whether all patients in these studies were truly asymptomatic. In addition, studies mainly reported that patients with severe AS referred to their clinic were included but did not clarify whether patients already had severe AS a certain time before primarily being evaluated in the clinic (eg, prevalent cases) or had mild or moderate AS when primarily being evaluated and progressed to severe AS just before a later check (eg, incident cases). Nevertheless, there was considerable heterogeneity in our meta-analyses of event rates. Although subgroup analyses to detect heterogeneity within meta-analyses of observational studies should be interpreted with caution, our subgroup analyses revealed that the type of study (prospective vs retrospective) and the duration of follow-up (short vs long mean and total follow-up time) were associated with differences in event rates. This may have been the result of more closely monitoring patients who were prospectively followed, with earlier recognition of symptoms and timely referral for intervention, as opposed to a less strict follow-up regimen in retrospective studies. Moreover, the higher rates of symptom development, (an indication for) aortic valve intervention, all-cause death, and sudden death in studies with a shorter mean and total length of follow-up of a conservative strategy are most likely related to shorter follow-up due to the occurrence of these events, and publication bias may also play a role. Lastly, the effect of the associations between variables from multivariable analysis of several studies could not be pooled due to different definitions or cutoffs used in the models. Initiatives like the Valve Academic Research Consortium can further standardize studies to improve meta-analyses.53
In this meta-analysis, asymptomatic severe AS was associated with a high rate of developing an indication for aortic valve intervention, while all-cause, cardiac, and sudden death occurred in 4.8, 3.0, and 1.1, respectively, of 100 patients per year during a conservative strategy. Therefore, it is important to consider not only sudden death but also cardiac death due to heart failure or other causes. Patients with higher severity of AS, low-flow AS, evidence of LV damage, and atherosclerotic risk factors are at particular high risk of death or requiring intervention. Moreover, our meta-analysis suggested that surgery vs an initial conservative treatment strategy is associated with better long-term survival. Although existing guidelines provide some guidance on when to perform SAVR in patients with asymptomatic severe AS, this meta-analysis provides additional data to support a recommendation to consider early intervention in patients at high risk of adverse events. Further results from the ongoing randomized clinical trials are required to substantiate the role of early intervention in patients with asymptomatic severe AS.
Corresponding Author: Stuart J. Head, MD, PhD, Department of Cardiothoracic Surgery, Erasmus Medical Center, PO Box 2040, 3000 CA Rotterdam, the Netherlands (email@example.com).
Accepted for Publication: May 19, 2020.
Published Online: July 8, 2020. doi:10.1001/jamacardio.2020.2497
Author Contributions: Drs Head and Gahl had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Gahl and Çelik contributed equally as first authors; Drs Jüni and da Costa contributed equally as last authors.
Concept and design: Çelik, Head, Pibarot, Reardon, Van Mieghem, Kappetein, Jüni, da Costa.
Acquisition, analysis, or interpretation of data: Gahl, Çelik, Head, Vanoverschelde, Jüni, da Costa.
Drafting of the manuscript: Gahl, Çelik, Head, Vanoverschelde, Reardon, Jüni, da Costa.
Critical revision of the manuscript for important intellectual content: Çelik, Vanoverschelde, Pibarot, Reardon, Van Mieghem, Kappetein, Jüni, da Costa.
Statistical analysis: Gahl, Head, Kappetein, Jüni, da Costa.
Administrative, technical, or material support: Çelik.
Supervision: Head, Vanoverschelde, Reardon, Van Mieghem, Jüni, da Costa.
Conflict of Interest Disclosures: Dr Pibarot reports grants from Edwards Lifesciences and Medtronic outside the submitted work. Dr Reardon reports consulting for Medtronic. Dr Van Mieghem reports research grants support by Medtronic, Edwards Lifesciences, Boston Scientific, Abbott, and PulseCath and serves on the advisory board for PulseCath and Ancora. Dr Kappetein reports personal fees from Medtronic during the conduct of the study and outside the submitted work. Dr Juni reports grants from AstraZeneca, Biotronik, Biosensors, Eli Lilly and Company, and The Medicines Company outside the submitted work and serves as unpaid member of the steering group of trials funded by AstraZeneca, Biotronik, Biosensors, St. Jude Medical, and The Medicines Company. No other disclosures were reported.