In superhero films, the term “nemesis” refers to a hero’s principal and recurring enemy, that is, the villain representing the greatest challenge and the one who repeatedly confronts that hero. A nemesis is not just any villain, but rather an adversary with a special relationship to the hero because it represents a moral or psychological opposite, possesses similar or comparable abilities, and therefore makes the confrontation more evenly matched, while a deep personal history or rivalry often exists between them. For all these reasons, the nemesis becomes the hero’s greatest obstacle.
No definition could better capture what the crossing of survival curves is already beginning to represent for TAVI. The study discussed in this blog chapter is specifically a systematic review and Bayesian meta-analysis of randomised controlled trials (RCTs) comparing 5-year outcomes after transcatheter aortic valve implantation (TAVI) and surgical aortic valve replacement (SAVR) in low- and intermediate-risk populations, and it appropriately addresses this troublesome crossing of the curves head-on.
The inclusion criteria required RCTs comparing TAVI and SAVR in low- to intermediate-risk patients with a minimum follow-up of 5 years. Searches were conducted in MEDLINE, PubMed Central, Embase, and the Cochrane Library through June 4, 2025. Studies with shorter follow-up (<5 years) were excluded. The primary endpoint was all-cause mortality, whereas stroke and the composite of death and stroke were secondary endpoints. In contrast to traditional methods, a Bayesian model was applied, allowing the generation of probability distributions and a direct quantification of the likelihood of clinically relevant benefit or harm. Because the risk of death changes over time and the proportional hazards assumption is therefore violated, the authors consider it highly inappropriate to report hazard ratios (HRs), which may be misleading when curves cross. Accordingly, differences in the occurrence of the primary and secondary endpoints were assessed using pooled relative risk (RR) based on event rates specifically reported at the 5-year time point as a milestone analysis, together with calculation of 95% credible intervals (CrIs). Sensitivity and subgroup analyses were also performed through stratified analyses according to surgical risk (low vs intermediate) and meta-regression to assess whether valve type (earlier vs current-generation models) or expansion mechanism (balloon-expandable vs self-expanding) influenced outcomes.
Six clinical trials were included (PARTNER 2, PARTNER 3, SURTAVI, EVOLUT-LR, NOTION, and UK-TAVI), comprising a total of 7249 patients (3704 assigned to TAVI and 3545 to SAVR). Regarding 5-year all-cause mortality, mortality was 29.7% with TAVI versus 27.6% with SAVR, with a median RR of 1.12 (95% CrI = 1.02–1.22), representing a 12% increase in the risk of death with TAVI. In the posterior probability analysis, there was a 99.3% probability that surgery was superior to TAVI for 5-year survival. The probability that the difference exceeded an RR of 1.05 and 1.10 was 91.9% and 65%, respectively. Survival curves showed an early advantage for TAVI, but the curves crossed between years 3 and 4, after which surgery began to show superior outcomes.
As for stroke incidence, risk was also higher in the TAVI group, with a median RR of 1.13 (95% CrI = 0.93–1.39). The posterior probability that SAVR outperformed TAVI in stroke prevention was 88.0%. Because only 4 trials provided separate Kaplan-Meier curves for stroke, cumulative incidence curves were not generated for this endpoint. Regarding the composite endpoint of death and stroke, cumulative incidence was 31.4% for TAVI and 29.6% for SAVR, with a 99.5% probability in favour of surgical superiority (median RR of 1.12; 95% CrI = 1.03–1.22).
In the risk-stratified analysis and meta-regression, intermediate-risk patients had a higher risk of death with TAVI (RR = 1.14; probability of SAVR benefit 97.7%). Among low-risk patients, although the signal was less pronounced, surgery again showed a favourable trend (RR = 1.05; probability of SAVR benefit 71.2%). No statistical evidence was found that surgical risk level modified the overall superiority of surgery over TAVI (p = .470). With regard to the valve-type analysis, including both model generation and TAVI platform type, balloon-expandable versus self-expanding, meta-regression confirmed that the results were not affected by the use of different valve models.
In summary, this comprehensive analysis confirms that in patients with a life expectancy beyond 5 years, surgery provides a meaningful advantage in survival and in the reduction of neurological complications when compared with the transcatheter approach.
This work provides fresh evidence that SAVR is clearly superior at 5 years compared with the transcatheter procedure in low- and intermediate-risk patients and represents an important departure from earlier studies suggesting equipoise or non-inferiority between TAVI and SAVR. The absolute mortality difference, ranging from 2.1% to 3.3%, is clinically relevant, with a number needed to treat of 30 to 48 patients in favour of surgery. In addition, the data show that the advantages of surgery only become evident after longer follow-up, as the survival curves cross between the third and fourth year.
Although the study does not analyse the exact mechanisms leading to this finding, several explanations are proposed for the higher mortality and stroke risk seen with TAVI:
- the long-term impact of the cumulative burden of higher permanent pacemaker implantation rates, paravalvular leakage reported across all studies, procedure-related conduction disturbances such as bundle branch block, restricted coronary access in the event of future revascularisation, and the risks associated with redo surgery after structural degeneration of the transcatheter prosthesis;
- a greater susceptibility to leaflet thrombosis due to asymmetric or incomplete deployment, with subsequent thickening and reduced leaflet motion and flexibility, together with flow disturbances created when TAVI is deployed within the native calcified valve, which may generate a prothrombotic milieu and increase the risk of thromboembolic events.
The more pronounced disadvantage of the TAVI group among intermediate-risk patients suggests that these individuals, older and more comorbid, may be more vulnerable to the downstream impact of procedure-related complications such as pacemaker requirement or paravalvular regurgitation. Therefore, the authors caution that expanding TAVI indications toward younger patients is questionable, since its negative consequences seem to emerge precisely beyond 5 years, although TAVI remains a valid option for patients whose life expectancy is below 5 years regardless of their initial surgical risk.
The discrepancy between these findings and those of other recent meta-analyses favouring TAVI can be readily explained by temporal bias, whereby greater weight is assigned to studies with very short follow-up, typically 1 year, thereby masking later complications. Moreover, as already noted, the use of hazard ratios is inappropriate when survival curves cross. In this setting, the present study is undoubtedly statistically more robust.
The authors acknowledge several limitations that should be considered. There was marked heterogeneity in device technology because different valve types were included, balloon-expandable and self-expanding, although sensitivity analyses suggest that this did not alter the overall result. Loss to follow-up was greater in the SAVR group than in the TAVI group in nearly all trials analysed. With respect to valve morphology, most included patients had tricuspid valves, so the results are not necessarily generalisable to patients with bicuspid valves.
In conclusion, the authors argue that these findings challenge the ongoing expansion of TAVI into younger and lower-risk populations and support surgery as the preferred option for patients with longer life expectancy.
COMMENTARY:
In our view, this study represents a turning point in the interpretation of the many studies assessing mid-term outcomes of TAVI versus SAVR in the treatment of calcific aortic stenosis in low- and intermediate-risk patients, a field in which clinical practice guidelines have ventured into a major and highly questionable relaxation of TAVI indications.
First, because of the methodology used. Unlike conventional statistical approaches, this meta-analytic comparison is presented using a Bayesian framework. In Bayesian statistics, probability is interpreted as what we believe to be most likely on the basis of the information available. Starting from an initial assumption about a given event, a prior distribution is established. Once study data are collected, new information becomes available and the initial belief is updated, yielding what is known as the posterior distribution, which therefore describes which values are now considered more likely after observing the data. The posterior distribution is a curve showing which values are most plausible according to the observed data, and from it one can derive the credible interval, that is, the range encompassing the most likely values within the posterior distribution. The difference between a credible interval and the frequentist confidence interval is that a 95% credible interval genuinely implies a 95% probability that the event frequency lies within that interval. By contrast, a 95% confidence interval means that if the experiment were repeated many times, 95% of the calculated intervals would contain the true value, but we cannot say that the real parameter has more than a 95% probability of lying within that specific interval. This way of analysing the results neutralises the inappropriateness of reporting risk ratios based on proportional hazards when the event survival curves cross because the risk of event occurrence increases over time. And as is being shown across all studies comparing these 2 therapeutic strategies, this pattern has become a constant.
The results of this study should prompt an urgent reconsideration of the new guideline recommendations supporting TAVI in low- and intermediate-risk patients expected to live more than 5 years, because the probability that harm is being induced appears very high. In addition, as already suggested in the discussion of the study itself, age is not always a good surrogate for life expectancy. In fact, according to data from the Statistical Office of the European Union, once patients reach 80 years of age, life expectancy is 9.2 years for men and 11.2 years for women, so estimating TAVI indication on the basis of a fixed age such as 75 years, or even more aggressively 70 years as proposed in recent guidelines, seems to us an absolute recklessness if that parameter is considered in isolation to drive the indication. On the other hand, the even greater vulnerability reported in this study among intermediate-risk patients after that duration of follow-up should necessarily reopen the debate in this subgroup, which, due to its clinical profile, is being referred massively toward the percutaneous strategy. It should not be overlooked either that this new evidence adds to the 5-year PARTNER 2 results, where these patients were shown to require more reinterventions than surgical patients, but because of the burden of comorbidities inherent to moderate risk, this issue has largely been overlooked in most Heart Team decisions. At present, the discussion is often confined to low-risk patients alone, but as can be seen, moderate-risk patients with longer survival prospects should also merit surgical consideration, since the impact of TAVI-related complications translates into worse mortality and stroke outcomes earlier than in less comorbid patients.
One limitation of this study that deserves mention, and was appropriately highlighted by the authors, is the difference in complete follow-up rates between both groups. Except in SURTAVI, where follow-up completion was higher in the SAVR arm than in the TAVI arm (95.5% vs 93.7%), patients in the TAVI groups had higher follow-up percentages than those in the surgical groups, and the differences were striking, greater than 5%, in both PARTNER trials, specifically 9.6% and 6.3% in PARTNER 2 and PARTNER 3, respectively. The authors note that they are conducting a subanalysis addressing the characteristics and outcomes of these “lost” patients. The most plausible explanations for such follow-up losses may be, on the one hand, mortality, and on the other hand, paradoxically, the fact that patients with more complications and greater dependence on medical care tend to remain more engaged in study follow-up than those who are completely asymptomatic.
We should also strongly emphasise the uncertainty surrounding how representative the analysed RCTs truly are when compared with real-world data. In the meta-analysis by Tomsic et al, 19 studies were analysed, 9 randomised trials and 10 routine-practice studies, including 74797 patients who underwent SAVR. Early mortality was similar between patients enrolled in clinical trials and those treated in routine practice. However, in real-world practice, lower rates of periprocedural complications such as stroke, pacemaker implantation, and myocardial infarction were observed. At 5 years, overall survival was significantly better in patients treated in daily clinical practice than in those included in the SAVR arms of randomised trials, both in low-risk and intermediate-risk patients. All of this raises legitimate concerns about the external validity of comparative SAVR versus TAVI trials.
Finally, all studies analysed except UK-TAVI (n = 913) included only patients with tricuspid aortic valves, and in that trial the proportion of patients with bicuspid anatomy was not reported. NOTION 2 found that patients with bicuspid anatomy had worse outcomes in terms of mortality, stroke, and rehospitalisation compared with SAVR as early as 1 year after the procedure, with composite event rates of 14.3% versus 3.9% and an HR of 3.8 (95% CI = 0.8–18.5), so long-term differences will likely be even more striking once those results are fully analysed.
In conclusion, this new evidence reinforces doubts regarding the appropriateness of expanding TAVI indications over surgery in both low- and intermediate-risk subgroups with a life expectancy beyond 5 years and strengthens the unavoidable and well-established responsibility of Heart Teams to assess the most appropriate technique carefully for each patient. Quite clearly, the information provided to the patient during preprocedural counselling should include the possibility that despite lower invasiveness and probably lower peri-procedural mortality and stroke risk, the risk of mortality and stroke at mid-term follow-up, already at 5 years, is lower with the surgical option.
REFERENCE:
Marin-Cuartas M, Kawczynski MJ, de Waha S, Kiefer P, Falk V, Siepe M, et al. Updated 5-year outcomes of transcatheter versus surgical aortic valve replacement in patients with severe aortic stenosis at low- to intermediate-surgical risk. Heart. 2026 Feb 11:heartjnl-2025-327092. doi: 10.1136/heartjnl-2025-327092.
Reddy RK, Howard JP, Mack MJ, Reardon MJ, Jørgensen TH, Hørsted Thyregod HG, et al. Transcatheter vs Surgical Aortic Valve Replacement in Lower-Risk Patients: An Updated Meta-Analysis of Randomized Controlled Trials. J Am Coll Cardiol. 2025 Mar 11;85(9):926-940. doi: 10.1016/j.jacc.2024.12.031.
Tomsic A, Velders BJJ, Kawczynski MJ, Schoones JW, Klautz RJM, Palmen M, et al. Representativeness of surgical controls in aortic valve replacement trials: comparison with routine surgical cohorts. Heart. 2026 Feb 12;112(5):239-245. doi: 10.1136/heartjnl-2025-326162.
