Transcatheter aortic valve implantation (TAVI) has emerged as a popular treatment for patients with severe aortic stenosis, surpassing surgical procedures in some countries (1). We previously indicated a potential higher mortality associated with TAVI compared to surgical aortic valve replacement (SAVR) at 5-year follow-up (2), mainly in high risk patients (3–5). The longer-term performance of TAVI vs. SAVR in patients with lower risk remains uncertain. Additionally, the temporal changes in TAVI performance at different timepoints have yet to be determined. Given the expansion of TAVI to low-risk patients with increased life expectancy, this assessment holds critical clinical importance.

The 5-year follow-up data from nearly all registered comparative randomized controlled trials (RCTs) of TAVI vs. SAVR have recently been published (6–9). We therefore are able to assess the longer-term outcomes of TAVI and conduct a landmark analysis to identify the timepoint at which the performance of TAVI might diverge from SAVR, as indicated in some studies (5). The aim of our study was to evaluate the longer-term and temporal performances of TAVI compared to SAVR, both overall and within important subgroups.

We reported the meta-analysis in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline (Supplementary Table S1).

This present meta-analysis, including comprehensive data from all available trials comparing TAVI with SAVR, with >8,000 patients and longer-term follow-up data from nearly all trials, yields several important conclusions (Central Illustration). First, TAVI was associated with a higher risk of longer-term primary outcome compared to SAVR among participants with higher risk, but not among those with lower risk. However, a significant temporal interaction was detected in both risk profiles. Second, TAVI with BEV was associated with a higher risk of longer-term primary outcome compared to SAVR, whereas no statistical difference was found with SEV. There was a significant interaction between the two valve systems, and a temporal interaction was observed in both TAVI systems. Third, landmark analysis revealed a lower risk of primary outcome in TAVI compared to SAVR within the initial 30 days, comparable between 30 days and 2 years, and a significant higher risk beyond 2 years. Fourth, overall analysis showed that TAVI was associated with a higher longer-term risk of all-cause death, TIA, MVC, PPM, reintervention, rehospitalization, and moderate to severe PVL, a comparable risk of stroke and myocardial infarction, but a lower risk of major bleeding and new-onset atrial fibrillation.

We conducted a comprehensive search on PubMed to identify relevant meta-analyses comparing the longer-term outcomes of TAVI and SAVR. However, these meta-analyses included 3–4 trials with 5-year follow-up data, focusing exclusively on patients with higher risks (2, 24, 25). In contrast, our meta-analysis incorporated a larger dataset, comprising 7 trials with longer-term follow-up data, encompassing both higher- and lower-risk patients. It is important to note that our study utilized longer-term data from nearly all registered large RCTs. One of the identified meta-analyses employed a network meta-analysis approach but considered 1-to-2-year follow-up as long-term (26). Another meta-analysis included only 3 RCTs but supplemented them with 7 propensity-score matching observational studies, which were limited by inadequate adjustment for important confounding (27). We also performed several additional analyses. Firstly, we conducted a landmark analysis to assess the differences in TAVI outcomes within specific time intervals, revealing significant temporal variations in the effect of TAVI. Secondly, we conducted subgroup analyses based on TAVI systems and surgical risks, revealing noteworthy distinctions between subgroups.

None of the trials included were specifically designed to have sufficient statistical power to detect a significant reduction in all-cause death. However, our meta-analysis revealed a significant higher risk of longer-term mortality associated with TAVI. This finding aligns with the temporal trend observed in primary outcome. Further subgroup analysis indicated a significantly higher risk of all-cause death in TAVI among higher-risk patients and with BEV, but no significant difference was observed in lower-risk patients or with SEV. Importantly, the temporal trend was also only evident in the former two subgroups. A separate meta-analysis of 7 propensity matched studies corroborated our findings by showing a significantly higher risk of mortality at 5-year follow-up (27).

TAVI demonstrated initial superiority over SAVR within the first year but lost this advantage thereafter in lower risk patients. Given that lower risk patients typically have good life expectancy, this temporal interaction warrants intensive and close attention. In the PARTNER 3 trial, Kaplan–Meier event curves for the primary outcome crossed around the 2- to 3-year mark, thereafter favoring SAVR, while in the Evolut Low Risk trial, the curves remained parallel, favoring TAVI (8, 9). Although there were some differences, the pooled analysis of longer-term data from these lower-risk trials did not show substantial heterogeneity (I² = 20%). A large real-world registry including 42,586 patients who underwent isolated SAVR and meeting the inclusion and exclusion criteria for the PARTNER 3 and Evolut Low Risk trials, revealed excellent survival rates in low-risk patients following SAVR, with all-cause mortality of 7.1% at 5 years and 12.4% at 8 years (28). Similar findings were observed in other large registries (29). Determining whether TAVI can achieve such excellent long-term outcomes as SAVR will require robust evidence from follow-up periods exceeding 10 years. The recommendation of TAVI in these patients is pending this evidence.

We showed a higher longer-term risk of primary outcome and all-cause death in TAVI compared to SAVR among higher-risk patients. These observations seem a paradox, i.e., patients with a higher surgical risk actually had better longer-term outcomes when they underwent surgery instead of opting for TAVI. Notably, the short-term risk of all-cause death was not decreased in TAVI in higher risk patients. This observation was similar to several meta-analyses with higher-risk patients (2, 24). Unfortunately, no randomized trials in high-risk patients using newer-generation valves have been conducted thus far. There have been some propensity-matched studies that shed light on this topic. For instance, a study involving 72 pairs of high-risk patients, although utilizing mixed generations of TAVI valves, showed a lower in-hospital mortality rate but a higher risk of all-cause death at 5-year follow-up in the TAVI group (30). Another propensity-matched analysis of 783 pairs of intermediate-risk patients (mean age: 81.7 years, mean STS score: 5.5) using newer-generation SAPIEN 3 valves demonstrated a comparable risk of death or disabling stroke at 5 years compared to SAVR (31). Further studies are warranted to evaluate the performance of TAVI with newer-generation valves compared to SAVR in the context of higher-risk patients.

An interesting finding of our analysis was the significant interaction between BEV and SEV regarding the primary outcome and all-cause death at longer-term follow-up. A temporal interaction was observed in BEV for both the primary outcome and all-cause death, while in SEV, it was observed only for the primary outcome. These temporal trends closely align with those reported in the PARTNER 2A trial (5), which compared early-generation BEV TAVI with SAVR in higher surgical risk patients, and the PARTNER 3 trial (9), which compared newer-generation BEV TAVI with SAVR in lower surgical risk patients. Landmark analyses of clinical events between 2 and 5 years in both trials demonstrated higher rates of all-cause death and the primary outcome in TAVI compared to SAVR. Similarly, in another trial of BEV TAVI, the Kaplan–Meier event curves for all-cause death converged at 2 years (4). In contrast, trials comparing SEV TAVI to SAVR showed Kaplan–Meier event curves for the primary endpoint that remained parallel, favoring TAVI in the Evolut Low Risk trial (8), nearly overlapped in the SURTAVI (6) and NOTION (20) trials, and converged until the 5-year mark in the U.S. CoreValve trial (3). Longer-term data from head-to-head comparisons of BEV with SEV TAVI have been reported in only one RCT (32). In this trial, with 241 high-risk patients randomly assigned to early generation BEV and SEV, all-cause mortality (53.4% vs. 47.6%) and cardiovascular mortality (31.6% vs. 21.5%) at 5 years were numerically higher in the BEV group compared with the SEV group, consistent with our findings. These differences might be attributed to better forward flow hemodynamics and less structural valve deterioration in SEV compared to BEV (32). Several propensity-matched studies showed varied findings, but these conclusions were limited by residual confounders that could not be fully accounted for, such as patients’ anatomical suitability. It is likely that more patients with extensive outflow tract calcifications, low implanted coronary arteries, or complex and small femoral access received SEV (33). We found no significant difference between BEV and SEV at short-term follow-up, which is also consistent with findings from other RCTs (34, 35).

Our analysis had several strengths. Firstly, we incorporated the largest number of RCTs with longer-term follow-up outcomes, ensuring a comprehensive evaluation of the data. Additionally, the trials included in our analysis had nearly identical follow-up durations, enabling landmark analyses and mitigating the potential impact of variations in follow-up durations on the outcomes.

However, it is important to acknowledge some limitations. Firstly, our analysis was based on trial-level rather than patient-level data. Although we performed subgroup analyses based on clinically relevant subgroups, we were unable to conduct more detailed meta-regression analyses to account for potential confounding factors beyond the subgroup variables. Secondly, concomitant procedures were performed in both TAVI and SAVR groups in original trials, which could potentially influence the evaluation of isolated TAVI vs. isolated SAVR. Thirdly, our assessment of publication bias was limited by the relatively small number of trials, potentially affecting the ability to detect small-study effects.

TAVI was associated with a higher longer-term risk of primary outcome compared to SAVR in higher-risk patients and with balloon-expandable valves. However, a characteristic temporal interaction was documented in all subgroups. Long-term follow-up data from low-risk trials and large trials comparing TAVI with balloon-expandable and self-expanding valves are warranted to test these findings.