Introduction
Surgical aortic valve replacement (SAVR) had been the only effective therapy for patients with aortic stenosis (AS) until the introduction into clinical practice of transcatheter aortic valve implantation (TAVI). The remarkable advances in bioengineering technology and interventional cardiology techniques over the years have benefited from the following issues: (i) a drastic reduction in mortality rates, (ii) a significant reduction of complications, due to better patient selection and preprocedural computerized tomography, and (iii) greater operator experience. The robust evidence in favor of TAVI resulting from randomized controlled trials (RCTs) were summarized in the international guidelines [1, 2] that strongly recommend TAVI in inoperable, high- or intermediate-risk patients. Recently, on the basis of RCTs and registries, TAVI was successfully reported in patients with intermediate and low surgical risk with comparable or even better results than SAVR [3–14]. Moreover in 2019, PARTNER 3 and EVOLUT LOW RISK trials, performed on patients with severe AS at low risk of death with surgery, demonstrated benefits of TAVI over surgery [9, 11]. As a consequence, recently the United States Food and Drug Administration (FDA) first approved an expanded indications for several transcatheter heart valves (the Edwards Lifesciences’s Sapien 3 and Sapien 3 Ultra, and self-expanding Medtronic CoreValve Evolut R and CoreValve Evolut PRO) including patients who are at low surgical risk for death or major complications associated with open-heart surgery (https://www.fda.gov/news-events/press-announcements/fda-expands-indication-several-transcatheter-heart-valves-patients-low-risk-death-or-major. Accessed May 14, 2020). Indeed, on the basis of the results of PARTNER 3 trial, Edwards Lifesciences announced that SAPIEN 3 valve cleared for use in low-risk patients in Europe (https://www.edwards.com/ns20191106. Accessed May 14, 2020). As encouraging data continue to emerge, TAVI seems destined to replace SAVR as the gold standard therapy of AS [15, 16].
Based on the previous evidence, the aim of our meta-analysis of RCTs and observational studies (Obs) was to compare TAVI vs. SAVR in patients with AS at low surgical risk.
Methods
The present meta-analysis was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. The literature search was performed through PubMed and Cochrane computerized database and at the site https://clinicaltrials.gov, in order to include all studies published between January 2014 to March 31st, 2020 reporting on TAVI vs. trans-vascular SAVR in patients with severe AS at low surgical risk. The low-risk population was defined by STS score < 4% (Society of Thoracic Surgeons Predicted Risk of Mortality [STS--PROM]) or Logistic (European System of Cardiac Operative Risk Evaluation [LES]) Euroscore < 10%) [17–19]. The reference lists of retrieved full-text articles were also examined to identify potentially relevant studies not selected by the electronic search.
Two investigators independently performed the studies selection with the aim to include only studies that reported 30-day and/or at least one of the safety endpoints under evaluation. Conflicts were resolved by consulting a third investigator.
Studies published in languages other than English, conference abstracts or proceedings, TAVI performed using transapical approach, SAVR performed with sutureless prostheses and duplicate studies, were excluded from the meta-analysis.
Outcomes
The primary endpoints were all-cause 30-day mortality and 1-year mortality. The 30-day safety endpoints were: the need for new permanent pacemaker (PM) implantation, major bleeding (including major, life threatening, or disabling bleeding), acute kidney injury stage 2 or 3, stroke, moderate/ /severe paravalvular leak. Endpoint criteria were selected according to the standardized definitions of the Valve Academic Research Consortium (VARC)-2 [20].
Statistical analysis
The meta-analysis was performed using Review Manager (RevMan) (computer program) version 5.3. (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014) [21]. The relative risks (RR) with the 95% confidence intervals (CI) were computed for each individual study, and RRs were combined using the Mantel--Haenszel random-effect model to take into account possible heterogeneity among studies rather than chance. A Forest plot was used for a graphical presentation of the results (reporting the effect estimates for the individual studies together with the overall measure of effect) and the selected studies were examined to assess the homogeneity/heterogeneity of the results by visually inspecting the overlap of the CIs of the risk estimates in the different studies and by computing the Cochran Q test and I2 statistics. The meta-analysis was performed taking into account RCTs and Obs subgroups, using the test for subgroup differences to evaluate the agreement/disagreement of the results between RCTs and Obs. In case of heterogeneity greater than moderate into each subgroup (i.e., I2 > 60%) [22] a funnel plot together with the 95% confidence limits around the summary treatment effect (i.e.: the expected distribution of studies in the absence of heterogeneity or of selection biases) [23] was drawn and a sensitivity analysis was performed by excluding the studies falling outside the 95% CI at the visual inspection of the Funnel plot.
All statistical tests were two sided and alpha (α) error of ≤ 0.05 was defined as statistically significant.
Results
After detailed review, 9 studies related to 4 RCTs [4, 5, 8–11, 13] and 5 Obs [14, 24–28] out of 5946 articles were selected (Fig. 1). The main characteristics of the selected studies are reported in Table 1.
Study |
Recruitment |
Country |
Center (n) |
Type |
Design |
THVs |
TAVI |
SAVR |
Patients included in the meta-analysis |
Follow-up (years) |
|||||
No. of patients |
STS score |
No. of patients |
STS score |
||||||||||||
RCT |
Obs |
RCT |
Obs |
||||||||||||
Tyregod et al., |
12/2009–04/2013 |
Europe |
Multicenter (3) |
RCT |
Superiority |
SE |
CoreValve (Medtronic) |
ITT: 145 |
– |
2.9 ± 1.6 |
ITT: 135 |
– |
3.1 ± 1.7 |
276 |
6 |
Serruys et al., |
06/2012–06/2016 |
Canada, |
Multicenter (87) |
RCT |
Non-inferiority |
SE |
Medtronic: CoreValve Evolut R |
mITT: 131 |
– |
2.3 ± 0.5 |
mITT:123 |
– |
2.3 ± 0.5 |
254 |
1 |
Mack et al., 2019 [9, 10] PARTNER 3 NCT02675114 |
03/2016–10/2017 |
USA, Australia, Canada, Japan, New Zealand |
Multicenter (71) |
RCT |
Non-inferiority |
BE |
SAPIEN 3 (Edwards Lifesciences) |
ITT: 503 |
– |
NA |
ITT: 497 |
– |
NA |
950 |
1 |
Popma et al., 2019 [11] |
03/2016–11/2018 |
Australia, |
Multicenter (86) |
RCT |
Non-inferiority |
SE |
Medtronic: CoreValve (3.6%), Evolut R (74.1%) Evolut PRO (22.3%) |
ITT: 734 |
– |
1.9 ± 0.7 |
ITT: 734 |
– |
1.9 ± 0.7 1.9 ± 0.7 |
1403 |
2 |
Waksman et al., |
TAVI: 01/2016–01/2018 |
USA |
Multicenter (11)
|
Obs |
Prospective |
SE, BE |
CoreValve, Evolut R or Evolut PRO (Medtronic) SAPIEN 3 (Edwards Lifesciences) |
– |
200 |
1.8 ± 0.5 |
– |
719 |
1.6 ± 0.6 |
919 |
1 (only |
Bekeredjian et al., |
2014–2015 |
Europe |
Multicenter (78) |
Obs |
Prospective registry |
NA |
NA |
– |
5113 |
< 4 |
– |
14487 |
1.8 ± 0.9 |
19600 |
1 |
Oh et al., |
01/2010–08/2018 |
Asia |
Two-center |
Obs |
Cohort study |
NA |
NA |
– |
168 |
< 4% |
– |
93 |
< 4% |
261 |
2 |
Schaefer et al, |
TAVI: 2008–2016 |
Europe |
Single-center |
Obs |
Prospective |
SE, BE |
Only first generation devices were used |
– |
431 |
< 4 |
– |
341 |
< 4 |
772 |
Median: |
Virtanen et al., |
01/2008–11/2017 |
Europe |
Multicenter (5) |
Obs |
Retrospective study |
BE, SE |
Evolut R, SAPIEN 3, Acurate neo, Lotus |
– |
325 |
2.1 ± 0.5 |
– |
2516 |
1.8 ± 0.6 |
2841 |
3 |
Primary endpoints
30-day mortality was reported in 8 of the 9 selected studies and were assessed in 26,989 patients (2629 from 3 RCTs, 24,360 from 5 Obs) [5, 9, 11, 14, 24–27], occurred in 1.6% of TAVI compared to 2.7% of SAVR. Indeed, the study by Serruys et al. [13] on SURTAVI subgroup with STS < 3 reports only 1-year mortality. The overall analysis showed a non-significant risk reduction in TAVI compared to SAVR (RR = –36%, p = 0.11); the analysis by subgroups showed in RCTs a significant risk reduction in favor of TAVI (RR = –56%, p = 0.04) while, in Obs the reduction of deaths in favor of TAVI was not significant (RR = –25%, p = 0.51) (Fig. 2). Indeed, the RCTs showed homogeneous results (I2 = 0%), whereas Obs were affected by high heterogeneity (I2 = 67%) (Fig. 2A). The visual inspection of the Forest plot detected in the study by Schaefer et al. [25] indicated the potential source of bias: it was the only study with a RR significantly in favor of SAVR (RR = 3.56, with a 95% CI ranging from 1.22 to 10.42; Fig. 2A). At the Funnel plot, the larger studies were plotted at the central top of the graph, demonstrating a convergence in the risk estimation with the increase of the sample size, whereas the smaller studies were scattered at the bottom of the graph. Again, the study of Schaefer et al. [25] was the only one falling outside the 95% CI. In the sensitivity analysis, by excluding the study by Schaefer et al. [25], the reduction in 30-day mortality in TAVI became significant also in the overall analysis (RR = –45%, p < 0.00001) and in Obs (RR = –44%, p < 0.00001) in absence of heterogeneity (I2 = 0%) (Fig. 2B). The test for subgroup difference showed agreement between RCTs and Obs (Chi2 = 0.31, df = 1, p = 0.58, I2 = 0%; Fig. 2B).
One year mortality, was assessed in 22,701 patients (2883 from 4 RCTs, 19818 from 2 Obs) [5, 9, 11, 13, 14, 25]. The overall analysis showed a non-significant risk reduction in TAVI compared to SAVR (RR = –24%, p = 0.16); the analysis by subgroups showed in RCTs a significant risk reduction in favor of TAVI (RR = –38%, p = 0.04) whereas in Obs non-significant risk increases in TAVI were observed (RR = +2%, p = 0.93; Fig. 3). The results among the 4 RCTs were homogeneous (I2 = 0%); a slight heterogeneity affected the 2 Obs (I2 = 25%). On the contrary, a high heterogeneity between RCTs and Obs was demonstrated in the test for subgroup differences (Chi2 = 2.87, df = 1, p = 0.09, I2 = 65.1%; Fig. 3).
Safety endpoints
Permanent pacemaker. The overall analysis showed a significantly increased risk of new PM implantation for TAVI compared to SAVR (p < 0.0001). The risk was increased both in RCTs (p = 0.007) and in Obs (p = 0.02) for TAVI (Fig. 4A). However, the comparisons were affected by high heterogeneity both in RCTs (I2 = 84%) and in Obs (I2 = 88%) (Fig. 4A). At the Funnel plot 2 RCTs [5, 9] and 2 Obs [25, 27] fell outside the 95% CI (Fig. 4B). By excluding these studies, the sensitivity analysis confirmed a significant increase of the risk for new PM (p < 0.00001) with homogeneous results (Fig. 4A).
Major, life threatening or disabling bleeding. Definitions of bleeding for each included study are reported in Table 2. VARC criteria were adopted by all the included studies with the only exclusion of Virtanen et al. [26]. This study, even though it was included in the initial analysis, was excluded in the sensitivity analysis because it was a source of heterogeneity (Fig. 4A).
Included studies |
Bleeding criteria |
Excluded at the |
Randomized controlled trial |
||
Tyregod et al., 2015 |
VARC 2 |
No |
Mack et al., 2019 |
VARC 2 |
Yes |
Popma et al., 2019 |
VARC 2 |
No |
Observational study |
||
Waksman et al., 2018 |
VARC 2 |
No |
Oh et al., 2019 |
VARC 2 |
No |
Schaefer et al., 2019 |
VARC 2 |
Yes |
Virtanen et al., 2019 |
“In this study, the Valvular Academic Research Consortium-2 definition of major and life-threatening bleeding was not applied because, unlike patients undergoing TAVR, a significant decrease of hemoglobin level is |
Yes |
The overall analysis showed a significant reduction of bleeding in TAVI compared to SAVR (RR = –65%, p = 0.008); the analysis by subgroups showed in RCTs, was a significant risk reduction in favor of TAVI (RR = –71%, p = 0.002) whereas in Obs the reduction of bleeding in favor of TAVI was not significant (RR = –60%, p = 0.18; Fig. 4A). Indeed, the overall heterogeneity was extremely high (I2 = 92%) both in the RCTs (I2 = 84%) and in Obs (I2 = 92%) (Fig. 4A). After the sensitivity analysis, the comparisons were performed between more homogeneous populations (RCTs: I2 = 36%; Obs: I2 = 0%) and showed a significant reduction of major bleeding for TAVI both in RCTs (p = 0.0004) and in Obs (p = 0.001) (Fig. 4A). Indeed, Obs and RCT had the same trend in the test for subgroup differences (Fig. 4A).
Acute kidney injury stage 2 or 3. The overall analysis showed a non-significant reduction of acute kidney injury in TAVI compared to SAVR (RR = –40%, p = 0.51; Fig. 4A). In the analysis by subgroups, in RCTs (p = 0.0003), but not in Obs (p = 0.63), the risk of acute kidney injury was significantly reduced for TAVI (Fig. 4A). Indeed, RCTs were homogeneous (I2 = 0%), but the Obs were not (I2 = 96%), due to the study by Schaefer et al. [25] falling outside the 95% CI at the Funnel plot (Fig. 4B). By excluding the study by Schaefer et al. [25], a significant reduction of acute kidney injury was observed for TAVI (Fig. 4A).
Stroke. The overall analysis showed a non--significant reduction of stroke in TAVI compared to SAVR (RR = –26%, p = 0.21; Fig. 5A). In the subgroups, the results of RCTs did not substantially differ from those of Obs: the risk of stroke was lower for TAVI, without reaching any statistically significant difference (Fig. 5A).
Paravalvular leak. A significant increase of moderate/severe paravalvular leak for TAVI in the overall analysis was observed (p < 0.00001) both in RCTs (p = 0.0005) and in Obs (p = 0.001) (Fig. 5B). The results of the analysis of RCTs were in accordance with those of Obs (test for subgroup differences: Chi² = 0.01, df = 1, p = 0.92, I2 = 0%) (Fig. 5B).
Discussion
The treatment of AS with TAVI in all patients in whom aortic valve surgery is indicated, irrespective of the surgical risk, is a useful goal to achieve because the interventional cardiology procedure is less invasive compared to cardiac surgery. Obviously, in order to extend the indications to TAVI in all patients with indications to SAVR, mostly in the low surgical risk, there must be conditions that allow it, and, in particular, the results of TAVI must be superimposable if not superior to those of surgical intervention [29, 30]. A major boost in this direction was given by the results of PARTNER 3 and EVOLUT LOW RISK trials, performed on patients with severe AS at low risk of death with surgery, which demonstrated the benefits of TAVI over surgery [9, 11]. Recently on the basis of RCTs [3–13] and registries [14, 24–28], TAVI was successfully reported in patients with moderate and low surgical risk with comparable or even better results than SAVR. Due to the limited evidence coming from RCTs, the present meta-analysis also included available evidence coming from Obs with the aim of increasing the power of analysis. However, the meta-analysis was performed by the two subgroups in order to take into account separately the results coming from the two types of studies. The test for difference between RCTs and Obs was also computed to evaluate their agreement/ /disagreement. Indeed, as stated by Briere et al. [31], “the inclusion of real-world evidence in meta--analyses may facilitate the confirmation of conclusions drawn from randomized controlled trials and, thus, reassure decision-makers that findings can be extrapolated to real-world populations.” [31]. In the comparisons evaluating 30-day mortality, after the exclusion of the study by Schaefer et al. [25] according to the sensitivity analysis, the results of the meta-analysis with both RCTs and Obs were homogeneous (I2 = 0%) and superimposable and highlighted a significant 45% reduction in mortality in the patients undergoing TAVI (Fig. 2B). Indeed, the study by Schaefer et al. [25] was a source of heterogeneity not only in 30-day mortality but also in the analysis of many safety endpoints. As affirmed by Schaefer et al. [25]: “An important limitation of this study is that only first-generation devices were used in TAVI patients” and despite: “Baseline, procedural, and follow-up data were prospectively collected from dedicated databases” data were “retrospectively analyzed”, having all the limitations of a retrospective study design [25].
In the present meta-analysis, at 1 year follow up, only the analysis of RCTs showed a significantly lower mortality in TAVI compared SAVR (p = 0.04), in disagreement with the results of the overall analysis, which nevertheless was affected by high heterogeneity in the test for subgroup differences (I2 = 65.1%) (Fig. 3).
As for the safety endpoints, the analysis did not demonstrate a different incidence of stroke between TAVI and SAVR (Fig. 5A). Furthermore, in TAVI a significantly increased risk of paravalvular leak (Fig. 5B) and new PM implantation (Fig. 4A) was observed (Fig. 5B). However, the analysis on new PM implantation was affected by high heterogeneity both in the overall effect (I2 = 83%) and into each subgroup (RCTs, I2 = 84%; Obs, I2 = 88%); Fig. 5A). As reported in the literature, this heterogeneity could be attributable to the inclusion of different types of prostheses in the analysis (Table 1) [32, 33].
Limitations of the study
The analysis of Obs could overestimate the effect of the treatment, due to the lack of randomization [34, 35]. However, the results of Obs were in agreement with RCTs for most comparisons, except when including the study of Schaefer et al. [25]. The reason was probably related (i) to the procedures performed in different intervals of time and (ii) to exclusive implantation of first-generation devices in TAVI (Table 1).
Conclusions
On the basis of the results of this meta-analysis TAVI is not superimposable to SAVR in patients with severe AS at low surgical risk. Some differences have to be highlighted. In particular, if 30- -day and 1-year mortality, major/life threatening or disabling bleeding and acute kidney injury stage 2 or 3 were significantly lower for TAVI, the need of new PM implantation and perivalvular leak were significantly lower in SAVR. Indeed, these last two events do not always have an early prognostic impact, but their long-term implications have not yet been established.
Consequently, we suggest the need of more trials to evaluate the effectiveness of TAVI as routine therapeutic procedure in the treatment of patients with low surgical risk severe AS.