Chronological comparison of TAVI and SAVR stratified to surgical risk: a systematic review, meta-analysis, and meta-regression

Absract Background Transcatheter aortic valve implantation (TAVI) has been established as a reasonable alternative to surgical aortic valve replacement (SAVR) in patients with severe aortic stenosis. However, long-term outcomes including valve durability and the need for reintervention are unanswered, especially in younger patients who tend to be low surgical risk. We performed a meta-analysis comparing clinical outcomes after TAVI and SAVR over 5 years stratified to low, intermediate, and high surgical risks. Methods We identified propensity score-matched observational studies and randomised controlled trials comparing TAVI and SAVR. Primary outcomes, including all-cause mortality, moderate or severe aortic regurgitation, moderate or severe paravalvular regurgitation, pacemaker placement, and stroke, were extracted. Meta-analyses of outcomes after TAVI compared to SAVR were conducted for different periods of follow-up. Meta-regression was also performed to analyse the correlation of outcomes over time. Results A total of 36 studies consisting of 7 RCTs and 29 propensity score-matched studies were selected. TAVI was associated with higher all-cause mortality at 4–5 years in patients with low or intermediate surgical risk. Meta-regression time demonstrated an increasing trend in the risk of all-cause mortality after TAVI compared with SAVR. TAVI was generally associated with a higher risk of moderate or severe aortic regurgitation, moderate or severe paravalvular regurgitation, and pacemaker placement. Conclusions TAVI demonstrated an increasing trend of all-cause mortality compared with SAVR when evaluated over a long-term follow-up. More long-term data from recent studies using newer-generation valves and state-of-the-art techniques are needed to accurately assign risks. BRIEF SUMMARY Transcatheter aortic valve implantation (TAVI) was associated with increased all-cause mortality at longer periods of follow-up irrespective of surgical risk. Aortic regurgitation, paravalvular regurgitation, major vascular complications, and pacemaker placement favoured surgical aortic valve replacement (SAVR) over TAVI. TAVI remained superior to SAVR in major bleeding and renal failure events. Long-term data on newer generation valves and up-to-date implantation techniques may provide better durability and improved outcomes after TAVI. HIGHLIGHTS TAVI had higher mortality at longer follow up irrespective of surgical risk. Aortic regurgitation and paravalvular regurgitation favour SAVR over TAVI. Major vascular complications and pacemaker placement also favour SAVR over TAVI. TAVI remains superior to SAVR in major bleeding and renal failure events. Long-term data on newer generation valves and up-to-date techniques are needed.


Introduction
Aortic stenosis (AS) is caused by degenerative changes from atherosclerosis, calcification, and ossification, and affects nearly 12% of the population above the age of 75 years [1,2].The prevalence of severe AS is 3.4% in this elderly population, about three-quarter of whom complain of symptoms attributable to their valve disease [1].In 2002, Cribier et al. described the first percutaneous implantation of aortic valve prosthesis in a 57-year-old man with calcific AS, leading to a paradigm shift in the treatment strategy for AS [3].The U.S. Food and Drug Administration (FDA) first approved transcatheter aortic valve implantation (TAVI) in severe AS in patients with prohibitive surgical risk in 2011 and subsequently to intermediate and low surgical risk groups [4].These advancements were propelled by landmark randomised controlled trials (RCTs) which provided the grounds for formulating the American College of Cardiology/American heart association (ACC/AHA) valvular heart disease guidelines for the management of AS [5][6][7].Currently, the guidelines recommend surgical aortic valve replacement (SAVR) for severe symptomatic AS patients under 65, TAVI for those above 80, and consideration for either TAVI or SAVR for those in between these age groups irrespective of surgical risk [5].
In this guideline-driven expansion of TAVI, questions regarding the long-term durability of percutaneously implanted valves have been brought to light, including the need for reintervention and the risk of structural valve deterioration [8].The present metaanalysis aims to compare clinical outcomes after TAVI and SAVR over 5 years stratified to low, intermediate, and high surgical risks and subsequently perform meta-regression to examine the yearly trend of these outcomes.

Search strategy and inclusion criteria
The authors declare that all supporting data are available within the article.Two authors (D.P. and S.A.) searched the online libraries, PubMed, Medline, Embase, Cochrane Library, and relevant websites (www.acc.org,www.escardio.org,www.europer.com,and www.tctmd.com)from inception to 1 January 2022.Official abstract presentations of landmark RCTs in prominent conferences were included.For inclusion, propensity score-matched observational studies needed to differ in the years from which the patients were recruited or the follow-up period if they analysed an identical registry.If studies of the same registry reported outcomes from the same years, a study with a larger sample size was preferentially selected to prevent duplicative inclusion of the same patients.
The search terms applied were 'transcatheter aortic valve implantation,' 'transcatheter aortic valve replacement,' 'TAVI,' 'TAVR,' 'surgical aortic valve replacement,' and 'SAVR.'The inclusion criteria were as follows [1]: RCTs or propensity score-matched observational studies with TAVI in one arm and SAVR in another arm [2]; disclosure of mean surgical risk score of TAVI and SAVR groups using either the Society of Thoracic Surgeons (STS) score or logistic EuroSCORE; and [3] both TAVI and SAVR groups with mean surgical risk scores in the same category (low, intermediate, or high).Sensitivity analyses including only RCTs were conducted to validate the findings from analyses that included observational studies.The low surgical risk was defined by a mean STS score <4% or logistic EuroSCORE <10%; intermediate surgical risk by a mean STS score 4-8% or logistic EuroSCORE 10-20%; and high surgical risk by a mean STS score >8% or logistic EuroSCORE >20%.When both STS and logistic EuroSCORE were reported, the former score was used to categorise the surgical risk.Studies in which TAVI was predominantly performed via the transapical approach were excluded as transapical TAVI is no longer the standard of treatment.

Data extraction and quality assessment
Primary outcomes were set as all-cause mortality, moderate or severe aortic regurgitation, moderate or severe paravalvular regurgitation, pacemaker placement, and stroke.Secondary outcomes consisted of major bleeding, major vascular complications, myocardial infarction, and renal failure.For the current metaanalysis, outcomes were grouped into the following intervals: perioperative to 30 days, 31 days to 1 year, 2 to 3 years, and 4 to 5 years.If outcomes were reported at 2 or more periods of follow-up within the same interval, the later outcome was selected to avoid duplications within the intervals.For the meta-regression, outcomes were plotted at their specific point in time without grouping them into intervals.
For each selected study, author, source of data, published year, years TAVI or SAVR was performed, follow-up period and sample sizes were obtained (Table 1).Baseline demographics, surgical risk scores, echocardiographic features, procedural characteristics, and valves used in the studies were arranged into tables (Supplementary Tables 2, 3, and 4).Inclusion and exclusion criteria and the definition of outcomes in the selected studies were further organised in tables (Supplementary Tables 4 and 5).The present metaanalysis adhered to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines [9].The authors declare that all supporting data are available within the article.The risk of biases was assessed using Cochrane Collaboration's tool for RCTs and the Newcastle-Ottawa scale for observational studies (Supplementary Tables 6 and 7) [10,11].Publication biases were assessed using Begg-Mazumdar and Egger tests after the visualisation of funnel plots (Supplementary Table 8).The study was exempt from ethical approval as data from published studies were retrieved and synthesised.

Statistical analysis
Conventional frequentist meta-analysis based on DerSimonian and Laird random effects model was performed to calculate pooled risk ratio (RR) and 95% confidence intervals (CI) of TAVI compared to SAVR as a reference.Haldane-Ascombe corrections were applied to make zero-cell corrections.Higgins and Thompson's I 2 statistic and s 2 were generated to evaluate the heterogeneities among the selected studies.P values below 0.05 were considered statistically significant.For meta-regression, mixed-effects logistic regression was used to examine the association of outcomes with time.Beta coefficient with its corresponding P value, s 2 , I 2 , H 2 , and R 2 indexes was generated from all meta-regression models (Supplementary Table 9).All statistical analyses were performed using metagen and metareg functions within the meta package in R version 4.0.5 (R Foundation for Statistical Computing, Vienna, Austria).

Results
After searching the literature and relevant sources, 36 studies consisting of 7 RCTs and 29 propensity scorematched studies were selected (Figure 1).A total of 63,518 patients, consisting of 30,122 TAVI and 33,396 SAVR patients, were included.There were 7839 (12.3%) patients from RCTs compared with 55,679 (87.7%) patients from propensity score-matched studies.Seventeen studies were classified as low surgical risk, 11 studies as intermediate surgical risk, and 7 studies as high surgical risk.One study reported outcomes for each of the surgical risks [12].
The mean age across all studies was in the 7th or 8th decade, but the comorbidities were widely variable among the studies (Supplementary Table 1).Most studies predominantly adopted the transfemoral approach when performing TAVI, and percutaneous coronary intervention (PCI) was simultaneously performed on average 8.8% of TAVI in 20 studies that reported concomitant PCI (Supplementary Table 2).Open heart surgery via full sternotomy was the most common method of SAVR, and coronary artery bypass graft (CABG) was simultaneously performed on average in 15.8% of SAVR in 30 studies that reported concomitant CABG.Types and brands of valves used were widely heterogeneous (Supplementary Table 3).The inclusion and exclusion criteria of the included studies are summarised in Supplementary Table 4.Most studies adopted the Valve Academic Research Consortium-1 (VARC-1) and VARC-2 definitions when determining outcomes [13,14] (Supplementary Table 5).Risks of biases in the studies were moderate to low (Supplementary Tables 6 and 7).Visualisation by funnel plot and assessment by Begg-Mazumdar and Egger tests demonstrated that publication bias was not significant (Supplementary Table 8).

Primary outcomes
In patients with low surgical risk, TAVI was associated with lower risk of all-cause mortality at 30 days (RR 0.68, 95% CI 0.55-0.86,p < 0.01), but with higher risk at 1 year (RR 1.12, 95% CI 1.03-1.21,p < 0.01) and 4-5 years (RR 1.44, 95% CI 1.03-2.01,p ¼ 0.03).In those with intermediate surgical risk, there was no difference in all-cause mortality between the two groups at 30 days, 1 year or 2-3 years; however, TAVI was associated with a higher risk of all-cause mortality at 5 years (RR 1.21, 95% CI 1.00-1.47,p ¼ 0.05).In those with high surgical risk, TAVI was associated with a higher risk of all-cause mortality at 2-3 years (RR 1.49, 95% CI 1.04-2.15,p ¼ 0.03), but not at 30 days, 1 year, and 4-5 years (Figure 2).The complete forest plots of all the primary outcomes, stratified to the three surgical risk strata, at different periods of follow-up are shown in Supplementary Figures 1-28.Meta-regression to years of follow-up demonstrated a trend of increasing risk of all-cause mortality in all surgical risks combined (b ¼ 0.06, R 2 ¼13.9%, p < 0.01), and low surgical risk (b ¼ 0.09, R 2 ¼13.3%, p < 0.01), but not in intermediate (b ¼ 0.01, R 2 ¼0%, p ¼ 0.79) and high surgical risks (b ¼ 0.07, R 2 ¼9.1%, p ¼ 0.07) (Figure 3).Sensitivity analysis of meta-regression including only RCTs revealed an increasing trend of all-cause mortality in low surgical-risk TAVI that was marginally significant (b ¼ 0.07, R 2 ¼0%, p ¼ 0.06) (Figure 4).TAVI was generally associated with a higher risk of moderate or severe aortic regurgitation, moderate or severe paravalvular regurgitation, and pacemaker placement at 30 days, 1 year, 2-3 years, and 4-5 years across all surgical risks (Figure 2).Meta-regression did not reveal the significant association of the risks of these three outcomes with time (Supplementary Table 9).On the other hand, no significant difference in the risk of stroke was found between TAVI and SAVR over 5 years across all surgical risk groups.Additionally, meta-regression did not demonstrate a significant association of stroke risk with time.Graphs displaying the results of the complete meta-regression are shown in Supplementary Figures 29-37.

Secondary outcomes
TAVI was associated with a lower risk of major bleeding and renal failure at 30 days and 1 year across all surgical risks (Figure 5).On the contrary, TAVI was associated with a higher risk of major vascular complications at 30 days and 1 year across all surgical risks.Meta-regression did not show a significant association between the risks of major bleeding, renal failure, and major vascular complications with time (Supplementary Table 9).The risk of myocardial infarction was not significantly different between TAVI and SAVR over 5 years across all surgical risks, but meta-regression showed a significant increase in risk over 5 years in all surgical risk strata combined (b ¼ 0.10, R 2 ¼0%, p < 0.01) and in the intermediate surgical risk group (b ¼ 0.11, R 2 ¼0%, p ¼ 0.02).

Discussion
Our meta-analysis showed that TAVI was associated with higher all-cause mortality at 4-5 years in patients with low or intermediate surgical risk.Meta-regression also demonstrated a trend of increasing the risk of allcause mortality with time.Our findings are consistent with multiple previous meta-analyses that examined the risk of all-cause mortality after more than 2 years of follow-up [15][16][17][18].Barili et al. reported that TAVI became a risk factor for all-cause mortality after 24 months, Swift et al. noted a higher risk of cardiovascular mortality after TAVI at 5 years, and Zhang et al. remarked that the risk of all-cause mortality with TAVI seemed to increase over time [15,16,18].Previous meta-analyses that included outcomes within 2 years were not able to find this trend of increasing all-cause mortality over longer periods of follow-up [19][20][21][22][23].However, it should be noted that previous meta-analyses were limited to either one subset of surgical risk or one point in the follow-up period.A few meta-analyses have examined the collective outcomes over time; however, these studies combined all outcomes after 2 years into one time period and did not stratify into different surgical risk groups [15,16,18].
Although not consistent across all surgical risk strata probably because of interstudy heterogeneities and differences in the number of studies included, a higher risk of all-cause mortality was observed at various durations of follow-up in all surgical risks.A possible explanation for the time-dependent increasing risk of all-cause mortality in the TAVI group may be related to the substantial proportion of SAVR patients undergoing concomitant CABG (Supplementary Table 2).CABG provides longterm mortality benefits in multivessel coronary artery disease, even in stable coronary artery disease, which may have tipped the balance towards SAVR [24][25][26].Our results are predominantly driven by older observational studies, so the difference in all-cause mortality may curb with emerging long-term data from more recent studies using newer-generation valves and state-of-the-art techniques.Echocardiography and multidetector CT scans for assessment of calcification severity and location have assisted in appropriate sizing and valve deployment hence reducing the risk of paravalvular leak, an unfavourable post-TAVI outcome.Furthermore, balloon post-dilatation, second valve-in-valve TAVI, and use of percutaneous vascular plugs are available options to manage paravalvular leak, although additional studies are needed to understand long-term valve hemodynamics [27].We also suspect that sicker patients may have been preferentially included in the TAVI arms of observational studies as not all clinically significant factors are accounted for in propensity score matching.Other explanations include a greater proportion of aortic regurgitation, paravalvular leak, and pacemaker implantation after TAVI, each of which has been reported to be associated with a poorer prognosis [28,29].Moreover, given that sensitivity analysis of meta-regression including only low surgical-risk RCTs showed a trend towards increasing all-cause mortality over time (p ¼ 0.06), more studies are needed to strongly recommend TAVI to low surgical-risk patients especially since these patients tend to be younger and have longer life expectancy [28].TAVI was found to be associated with a higher risk of moderate or severe aortic regurgitation and moderate or severe paravalvular regurgitation over 5 years across all surgical risks.This is consistent with the findings of previous meta-analyses [19][20][21][22][23]. Unlike SAVR, direct visualisation of the aortic root and annulus is not possible in TAVI, and only indirect assessment using transesophageal echocardiography, cardiac computed tomographic angiography (CCTA), and angiography are possible [30].Discrepancies in the size and The four graphs show the results of the mixed-effects logistic regression.The top left graph includes all surgical risks combined, and low, high, and intermediate surgical risk in clockwise order, respectively.Log of the risk ratio of all-cause mortality in TAVI compared to SAVR (y-axis) has been graphed against period of follow-up (x-axis).Detailed results of the meta-regression are separately shown in Supplementary Table 9.The significance of the regression is shown as a P-value at the top right corner of each graph.
shape of the valve to the annulus and slight misplacement of the valve can easily lead to transcatheter aortic valve regurgitation [31].TAVI was also generally associated with a higher risk of pacemaker placement and major vascular complications over 5 years across all surgical risks.Compared to surgical valves, transcatheter valves have the potential to be deeply implanted into the left ventricular outflow tract, leading to injury to the AV node and left bundle branches [32].Despite the increased use of CCTA, higher rates of conduction abnormalities requiring pacemakers have been one of the major drawbacks of TAVI, especially with self-expandable valves [33].It remains to be seen if new techniques, such as minimising depth according to the membranous septum (MIDAS) approach and cusp overlap technique, will be able to reduce the need for pacemaker placement to a level comparable to SAVR [34,35].Vascular complications, predominantly secondary to failure of percutaneous device closure and access site haematomas, also occur more frequently after TAVI due to the large sheath system, especially with early TAVI technology.However, with the utilisation of smaller sheath sizes, flexible delivery systems, CT angiography assessment of the peripheral  vasculature, and increasing operator experience, the incidence of vascular complications has been decreasing in more recent studies [36,37].
As expected, TAVI was generally associated with a lower risk of major bleeding and renal failure over years of follow-up across all surgical risks.These  00 The four graphs show the risk ratio of TAVI compared to SAVR at different periods of follow-up stratified to low, intermediate, and high surgical risk strata for the following outcomes: major bleeding, myocardial infarction, major vascular complications, and renal failure.Risk ratio above 1 favours SAVR over TAVI.Each vertical line inside the blue box indicates the risk ratio while the perpendicular horizontal lines show the corresponding 95% confidence interval.The size of the blue box is indirectly proportional to the size of the confidence interval.
findings are not surprising as SAVR is more invasive, predominantly performed by full sternotomy and requires cardiopulmonary bypass, which is associated with coagulopathy [38].Bleeding after SAVR also tends to be more severe, resulting in acute kidney injury and dialysis [39].The difference in major bleeding and renal failure between TAVI and SAVR may widen in the future with more experienced operators, smaller TAVI devices, and low-risk populations [40].
Both stroke and myocardial infarction were not different between TAVI and SAVR over 5 years across all surgical risk groups.However, meta-regression revealed a trend of increasing myocardial infarction in TAVI compared to SAVR in all surgical risk strata combined and in the low surgical risk group.This can be explained by the high frequency of CABG performed together SAVR, providing long-term benefits in reducing myocardial infarction [41].Studies did not specify the type and duration of antiplatelet and anticoagulant used after valve replacement, whose differences may also impact the incidence of myocardial infarction.

Limitations
Our analysis is subject to several limitations.We did not have access to individual patient-level data which limited analysis of other specific outcomes of interest such as rates of prosthesis-patient mismatch, rehospitalization, reintervention Definition of outcomes was variable among studies (Supplementary Table 5), and bias may be inherent in outcomes reported by observational studies as there was neither standardisation nor adjudication.We also stratified the studies into low, intermediate, and high surgical risks at the studylevel and not at the patient-level.To create more comprehensively pooled outcomes with greater power, outcomes were divided into four intervals (perioperative to 30 days, 31 days to 1 year, 2 to 3 years, and 4 to 5 years), which may have led to underestimations.Many relatively older observational studies were included in our analysis.Patients in RCTs may have also been included in studies of registries.Although confounding factors were minimised by only including observational studies with propensity-score matched analysis, unmeasured cofounders may still be present, contributing to differences in observed outcomes.Given that sicker patients are more frequently offered TAVI, the TAVI groups in the observational studies may be subject to more unaccounted comorbidities and worse severity of clinical conditions.Propensityscore matched studies were included despite the limitations inherent in observational studies because the number of RCTs on the topic was insufficient to produce robust pooled outcomes especially when stratified into low, intermediate, and high surgical risk strata.We also determined that the inclusion of large registry-based studies can provide insights into realworld outcomes.In addition, a previously validated meta-analysis on this topic also included propensity score-matched studies [23].Only a few studies were included when generating pooled risk ratios for some outcomes, such as paravalvular regurgitation.

Conclusion
To our knowledge, our analysis is the largest metaanalysis chronologically comparing TAVI and SAVR over 5 years stratified to low, intermediate, and high surgical risks.We also included the 5-year outcomes of the PARTNER 2 A and SURTAVI trials and the 2-year outcomes of the PARTNER 3 trial [42][43][44].Although we found a trend of increasing risk of all-cause mortality over 5 years, there is significant weight given to older observation studies.Long-term data from recent RCTs using newer-generation valves and state-of-the-art techniques are needed to accurately assign risks, especially in patients with low surgical risk.Continued research and development are needed to improve outcomes on aortic regurgitation, paravalvular regurgitation, major vascular complications, and pacemaker placement after TAVI.

Ethical approval
This study was exempt from ethics approval as only data from previously published studies were retrieved and synthesised.

Clinical trial registration
This study is not a clinical trial.

Patient consent
Patient consent was not necessary as only publicly available study-level data from previously published papers were used.

Permission to reproduce material from other sources
Material from other sources has not been used in this study.References have all been appropriately cited.

Figure 1 .
Figure1.Flow diagram of the study selection process.The flow diagram shows the process of how the studies included in this meta-analysis were selected.All steps of this meta-analysis adhered to the PRISMA guidelines.

Figure 2 .
Figure2.Timeline of primary outcomes stratified to surgical risk.The five graphs show the risk ratio of TAVI compared to SAVR at different periods of follow-up stratified to low (blue), intermediate (yellow), and high (red) surgical risk strata for the following outcomes: all-cause mortality, moderate or severe aortic regurgitation, moderate or severe paravalvular regurgitation, stroke, and pacemaker placement.Risk ratio above 1 (represented by gray dotted lines) favours SAVR over TAVI.Red dotted lines represent hypothetical linear connections showing a potential trend between two points in time when the risk ratio at 2 to 3 years of follow-up was not present.

Figure 3 .
Figure 3. Meta-regression of all-cause mortality in TAVI vs SAVR by time.The four graphs show the results of the mixed-effects logistic regression.The top left graph includes all surgical risks combined, and low, high, and intermediate surgical risk in clockwise order, respectively.Log of the risk ratio of all-cause mortality in TAVI compared to SAVR (y-axis) has been graphed against period of follow-up (x-axis).Detailed results of the meta-regression are separately shown in Supplementary Table9.The significance of the regression is shown as a P-value at the top right corner of each graph.

Figure 4 .
Figure 4. Meta-regression of all-cause mortality including only randomised clinical trials.The four graphs show the results of the mixed-effects logistic regression that only included randomised controlled trials.The top left graph includes all surgical risks combined, and low, high, and intermediate surgical risk in clockwise order, respectively.Log of the risk ratio of all-cause mortality in TAVI compared to SAVR (y-axis) has been graphed against period of follow-up (x-axis).The significance of the regression is shown as a P-value at the top right corner of each graph.The beta coefficient (b) is written below each line of best fit displayed in red.

Table 1 .
General characteristics of selected studies.
AT: as treated; ITT: intention to treat.
Secondary outcomes by time stratified to surgical risk.