Studies were included, if they fulfilled the following criteria. a)

Randomized controlled trials (RCTs) in patients with severe native AS undergoing TAVR.

We searched PubMed, EMBASE, Clinicaltrials.gov, Scopus, and Web of science for all studies on patients with severe aortic stenosis undergoing TAVR (since inception to April 17th 2020) without any language restriction. We used multiple posting suffix (.mp) to improve sensitivity of our search. In addition, we looked for cross-references in the screened studies, review articles, and meta-analyses to identify other potential studies to be included. Our detailed search strategy is provided in the Supplementary Material. The study protocol is registered with PROSPERO, International prospective register of systematic reviews (CRD42020181190).

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement was followed during the development of this systematic review and meta-analysis (6). After eliminating duplicates, screening of manuscripts was done based on title and abstracts to remove irrelevant articles by two independent authors. Full text assessment of relevant, identified articles were scrutinized again by the above authors. Risk of bias assessment was done using Cochrane risk-of-bias tool for randomized trials version 2 (RoB 2) (7). Assessment of risk of bias, inconsistency, indirectness, imprecision, publication bias and effect size were assessed to calculate “certainty of evidence” using the GRADE (Grading of Recommendations, Assessment, Development and Evaluations) approach (8). Furthermore, the GRADEpro guideline development tool was used to create a “Summary of findings” table and a GRADE “Evidence profile”. Screening, full-text assessment, data extraction and validity assessment were independently performed by two authors (NBS and HB). Discrepancy was resolved by the third author (BC). We extracted baseline characteristics of patients, procedural details and clinical outcomes from included studies.

The primary outcome of our study was all-cause mortality at 30 days. Several endpoints on early safety, clinical efficacy and device success as defined by Valve Academic Research Consortium-2 (VARC-2) criteria were examined as secondary outcomes (9), including cardiovascular (CV) mortality, all stroke (disabling and non-disabling), life-threatening bleeding, major vascular complications, major bleeding, acute kidney injury (AKI) Stage 2 or 3 (including renal replacement therapy), myocardial infarction (MI), coronary artery obstruction requiring intervention, valve-related dysfunction requiring repeat procedure, moderate to severe aortic regurgitation (AR)/para-valvular leak (PVL), atrial fibrillation, rehospitalizations for valve-related symptoms or worsening congestive heart failure, permanent pacemaker implantation, prosthetic valve endocarditis, valve thrombosis, NYHA class III or IV, early valve-related dysfunction, implantation of more than one valve per procedure, valve malposition, and annular rupture. Whenever outcomes reported were too few in numbers (<5 events) or reported by a single study only, they were not included in the final analysis.

Data extracted from the studies were imported into Review Manager Version 5.3 (The Nordic Cochrane Center, The Cochrane Collaboration Copenhagen, Denmark) for analysis. Pairwise meta-analysis was performed for overall analysis. We used DerSimonian and Laird random effects model for our analysis to calculate pooled risk ratio (RR) and 95% confidence interval (CI) for all outcomes. We calculated between-study heterogeneity by using the Higgins I2 statistic. We defined low and high heterogeneity as I² < 25% and >75% respectively. Publication bias was assessed visually by asymmetry in funnel plots. We performed sensitivity analyses utilizing various factors whenever statistically significant heterogeneity was found. This included an analysis after excluding studies one-by one that were considered an outlier based on methodological or interventional heterogeneity. A leave-one-out sensitivity analysis to remove the effect of one study at a time on our results was also performed. All tests were 2-tailed with a p value of <0.05 considered significant.

Six studies were included for final qualitative synthesis, of which four were RCTs (10–13), two were post hoc analysis of RCTs (14, 15) and one was a pre-specified device or valve type analysis of an RCT investigating the different peri-procedural antithrombotic strategies in patients undergoing TAVR (15) (Table 1). The post-hoc study belonged to the Portico-IDE (investigational device exemption) randomized trial, presented at international conferences (14, 15) but was not yet published in a peer-reviewed at the time of the databases search. Risk of-bias as assessed by RoB-2, showed low risk for 4 studies, some concern for 1 study and high risk for one study (Supplementary Table S2). BEV (Old generation-Sapien-XT and new generation- Sapien-3) was studied against four different types of SEVs (old generation SEV CoreValve and new generation SEVs that included Evolut R, Evolut PRO, Acurate neo, and Portico) and included a total of 3,141 patients (Figure 2). Out of 2,935 patients, 1,439 patients (49.0%) received BEV and 1,496 patients (51%) received SEV. The mean age of the patient population was 81.9 years and 53.3% were female. Baseline characteristics of included studies are summarized in the Supplementary Table S3.

Data regarding all-cause mortality was available from all six studies. Out of 1,439 patients who received BEV, 32 patients (2.2%) died at 30-day follow-up. In the SEV group, 67 patients (4.5%) died out of 1,496 patients. BEV was associated with significantly lower risk of death at 30 days compared with SEV [(2.2% vs. 4.5%; RR: 0.51; 95% CI: 0.31–0.82; p < 0.006, I² = 6%); High level of GRADE Evidence); (Central Illustration, Figure 2)]. Sub-group analysis was performed showed no effect of new generation SEV (Evolut-R, Acurate-Neo and Portico) vs. old generation SEV (CoreValve) on the result (Supplementary Figure S1). A sensitivity analysis using Mantel-Haenszel methods using a fixed effect model showed similar result at 30 days (p < 0.0001) (Supplementary Figure S2). Further sensitivity analyses were performed by leave-one out statistical analyses showed similar results (Supplementary Figure S3). Additional analyses based on the role of study type (RCTs comparing BEV vs. SEV head to head with post-hoc and pre-specified analyses of RCTs), and the role of recapturable valves revealed no significant subgroup effects (Figure 3 and Supplementary Figure S4, respectively). There was no publication bias as assessed by funnel plot (Supplementary Figure S5).

CV mortality at 30 days was available in 5 studies. BEV was associated with lower 30-day CV mortality compared with SEV [(2.5% vs. 4.3%; RR: 0.54; 95% CI: 0.32–0.90; p = 0.02; I² = 0%; Very low level of GRADE Evidence) (Figure 4A)]. Implantation of more than one valve per procedure (0.8% vs. 5.1%; RR: 0.15; 95% CI: 0.07–0.31); p < 0.00001; I² = 0%; moderate level of GRADE Evidence), and moderate/severe AR/PVL (2.5% vs. 9.0%; RR: 0.29; 95% CI: 0.17–0.48); p < 0.00001; I² = 0%; high level of GRADE Evidence) were also lower in the BEV arm (Figures 5A,B). A trend of reduced usage of pacemaker was observed with BEV as compared with SEV [(13.8% vs. 18.3%; RR: 0.73; 95% CI: 0.52–1.02; p = 0.06; I² = 42%; very low level of GRADE Evidence) Figure 5C].

All stroke (disabling and non-disabling) data were available in 6 studies. No significant difference in all stroke at 30 days was observed (Figure 4B). Similarly, no significant difference was noted between the 2 groups for early safety outcome at 30 days (Figure 4C), device success (Figure 5D), life-threatening bleeding, major vascular complications, AKI and atrial fibrillation (Figures 6A–D). In addition, no significant difference was found between the study arms for MI, major bleeding, rehospitalizations for valve-related symptoms or worsening congestive heart failure, valve related dysfunction requiring repeat procedure, valve malposition, and clinical efficacy (Supplementary Figures S6–S11). Sensitivity analysis using fixed effect model showed no difference in all the above parameters except AKI which was significantly lower in patients receiving BEV, with moderate heterogeneity (RR: 0.64; 95% CI: 0.40–1.01); p = 0.06; I² = 48%) (Supplementary Table S4). Outcomes on cardiac tamponade, annular rupture, NYHA class improvement, NYHA status ≥class 3, and conversion to open heart surgery and valve related dysfunction were not analyzed as their reported numbers were very low.

Four out of six studies had only low risk of bias assessment as assessed by Cochrane risk-of-bias tool for randomized trials version 2 (RoB 2). Two studies had some concern and one study was deemed to have risk of bias. GRADE system-based quality assessment was done for individual outcomes and a “Summary of findings” and GRADE “Evidence profile” are discussed in the Supplementary Tables S5, S6 respectively.

Our limitations include being a study-level meta-analysis and not a patient-level meta-analysis. Comparison of all forms of SEV together might be a limitation in our study given different design characteristics, but all the studied SEV are made of nitinol. Two of them were recapturable (Portico and Evolut R). Sensitivity analyses investigating the impact of recapturable SEVs showed no significant effect on outcomes (Supplementary Figure S5) (25). Inclusion of post-hoc analyses and RCTs with pre-specified endpoints are possible limitations but sensitivity analysis showed no effect on our primary outcome. We limited our analysis to 30-day outcomes due to absence of data across trials. As the outcomes of intention to treat analyses were not available for the post hoc studies, we collected data from the utilized modified-as-treated analyses as reported by the authors. Some studies also excluded patients with heavy calcification in the aortic annulus, left ventricular outflow tract (LVOT) or sinotubular junction, limiting the interpretation of these findings to those subgroups. Being a study-level metanalysis, we could not calculate the outcomes based on recently published VARC-3 crtieria (26).

Balloon-expandable TAVR is associated with reduced all-cause mortality (High level of GRADE evidence), CV mortality (Very low level of GRADE evidence) at 30 days compared with self- expanding TAVR in high risk patients undergoing TF-TAVR. Longer-term data are necessary to determine if the difference in outcomes persist and if the same effects are seen in lower risk patients.