Although widely known, the PARTNER 2A trial compared transcatheter aortic valve implantation using transfemoral (TF) or transthoracic transapical or transaortic (TAp or TAo) access with the SAPIEN XT prosthesis versus surgical aortic valve replacement (SAVR) in patients with severe symptomatic aortic stenosis (AS) and intermediate surgical risk. Patients were treated at 57 centers in the United States and Canada between 2011 and 2013. This publication provides the first 10-year follow-up from a large randomized trial comparing TAVI and surgery in patients at intermediate surgical risk.
Although the trial was originally designed with a 5-year follow-up period, the FDA required its extension to 10 years. Consequently, new patient consent had to be obtained to continue in-person clinical and echocardiographic assessments at 7 and 10 years, in addition to telephone follow-up. To complete mortality data for patients who did not renew their consent, withdrew from the trial, or were lost to follow-up, a Vital Status Sweep (VSS) was conducted using medical records, public databases, obituaries, social media, and telephone calls.
The primary endpoint was all-cause mortality at 10 years, including deaths identified through the VSS. Other outcomes included mortality according to vascular access route, aortic valve reintervention, quality of life assessed using the Kansas City Cardiomyopathy Questionnaire overall summary score (KCCQ-OSS), and prosthetic valve hemodynamic performance evaluated by echocardiography. An important feature of the statistical analysis was that the between-group comparison of valve reintervention accounted for death as a competing risk, given the high mortality among elderly patients with multiple comorbidities. Many patients who survived to 5 years did not renew their consent for continued follow-up to 10 years. Therefore, without the VSS, 10-year mortality data would have been available for only 61.9% of the TAVI group and 54.1% of the SAVR group. The VSS increased mortality data availability to 90.5% in the TAVI group and 89.5% in the surgical group. Baseline clinical and echocardiographic characteristics were similar in both groups. The mean age was 81.6 years, approximately 45% of patients were women, and the mean STS surgical risk score was 5.8%.
At 10 years, all-cause mortality, including deaths identified through the VSS, was 86.1% after TAVI, significantly higher than the 82.8% observed after surgery (HR: 1.13; 95% CI: 1.02-1.25; p = .02). However, mortality after TF TAVI was similar to that after surgery (83.9% with TAVI vs. 82.1% with surgery; HR: 1.07; 95% CI: 0.95-1.20; p = .27). By contrast, mortality was higher with TAVI in the TAp/TAo access group (93.2% vs. 85.1%; HR: 1.36; 95% CI: 1.11-1.67; p < .01; p for interaction = .03).
The cumulative incidence of aortic valve reintervention, accounting for the competing risk of death, was 6.3% in the TAVI group and 1.6% in the surgical group (p < .001). Kaplan-Meier estimates of aortic valve reintervention at 10 years were 16.1% after TAVI and 2.7% after surgery (HR: 4.26; 95% CI: 1.97-9.18; p < .01). Between years 5 and 10, reinterventions consisted primarily of transcatheter valve-in-valve procedures. Prosthetic valve restenosis and aortic regurgitation were the most frequent indications for reintervention in the TAVI group.
As in other randomized trials comparing these therapies, although the study was theoretically designed to compare isolated aortic valve replacement and severe unrevascularized coronary artery disease was an exclusion criterion, significantly more concomitant procedures were performed in the SAVR group. While 3.9% of patients underwent TAVI plus percutaneous coronary intervention, as many as 14.5% of patients in the surgical group required coronary revascularization. Ten-year mortality was similar among patients undergoing isolated surgery and those receiving additional surgical procedures.
Echocardiographic data at 10 years were available for only a negligible proportion of patients—2.4% of those initially randomized to TAVI and 3.6% of those randomized to SAVR. Furthermore, patients requiring valve reintervention were obviously excluded from this analysis, resulting in a highly selected sample. Therefore, although the authors report that mean aortic valve gradients remained stable and similar in both groups, I do not believe that the hemodynamic findings warrant further discussion in this review.
Quality of life improved in both groups during the first year and subsequently remained relatively stable and similar between groups from years 5 to 10.
In the discussion, the authors identify the main findings as follows: overall mortality was higher after TAVI than after surgery, although it was similar between TF TAVI and surgery and clearly higher after TAp or TAo TAVI. Valve reinterventions were more frequent after TAVI, as were aortic and paravalvular regurgitation (PVR). The higher reintervention rate may be related to the limitations of the SAPIEN XT prosthesis, a second-generation valve. Newer-generation prostheses, such as SAPIEN 3, incorporate design improvements and have been associated with lower PVR rates and better clinical outcomes, as observed in the 10-year report from the intermediate-risk PARTNER 2 SAPIEN 3 registry. In addition, computed tomography-based procedural planning was not routinely performed, valve sizing was less accurate, PVR was more frequent, and the range of available valve sizes was more limited. The 29 mm SAPIEN valve only became available after 42.5% of patients had already been randomized. The authors also caution that considering reintervention alone may underestimate the true extent of structural valve deterioration because many elderly, frail patients with multiple comorbidities may not have been considered candidates for a further procedure despite severe structural deterioration. Conversely, they also acknowledge that 15% of patients in the surgical group received Sorin Mitroflow or St. Jude prostheses, many of the latter probably Trifecta valves, both of which have now been withdrawn from the market.
The main limitation of the study was the loss of patients during follow-up. The requirement to obtain new consent at 5 years resulted in a substantial amount of missing data. Although the VSS provided additional mortality information, it could not recover data on other clinical events or consistently distinguish cardiovascular from noncardiovascular causes of death. Furthermore, loss to follow-up was more frequent in the surgical group, and patients who remained in the study may have represented a healthier population.
The authors conclude that TAVI with the SAPIEN XT prosthesis was associated with higher mortality than surgery, mainly driven by the TAp/TAo access cohort, and that aortic valve reintervention was also more frequent after TAVI. They emphasize, however, that these results should be interpreted in the context of the challenges inherent to long-term follow-up, including trial participant attrition, differences in follow-up between groups, and the competing risk of mortality.
COMMENTARY:
This study confirms what had already become apparent following publication of the 5-year PARTNER 2 results: long-term mortality after TAVI is comparable to surgery only in the transfemoral cohorts, whereas the transthoracic cohort consistently shows poorer outcomes across all risk profiles. It also provides compelling confirmation that second-generation transcatheter valves require valve reintervention more frequently than surgery, even in patient subgroups in whom the high mortality associated with advanced age and intermediate-risk comorbidities reduces both the likelihood of structural deterioration becoming clinically apparent and the probability that another procedure will ultimately be indicated when deterioration occurs and affects functional status.
In the PARTNER 2 trial, 2032 intermediate-risk patients with severe AS were randomly assigned to undergo TAVI with the Edwards SAPIEN XT prosthesis or SAVR. The incidence of death from any cause or disabling stroke was similar in both groups, with Kaplan-Meier event rates of 19.3% in the TAVI group and 21.1% in the SAVR group. In the transfemoral-access cohort, TAVI resulted in a lower event rate than SAVR (HR: 0.79; 95% CI: 0.62-1.00), although the difference was not statistically significant. In the transthoracic-access cohort, outcomes were similar between the TAVI and SAVR groups (HR: 1.21; 95% CI: 0.79-1.65). SAVR was associated with a lower incidence of moderate or severe PVR (0.6% vs. 3.7% at 30 days), and patients presenting with this complication had higher mortality during the 2-year follow-up than those with no, trace, or mild aortic regurgitation.
The 5-year follow-up results were subsequently published. In that analysis, the composite endpoint of all-cause mortality and stroke did not differ significantly between TF TAVI and SAVR. Consequently, the prespecified noninferiority criterion, defined as an HR of 1.2, was no longer met. Furthermore, the Kaplan-Meier curves showed the now familiar crossover after 2 years, and a subanalysis of outcomes between years 2 and 5 demonstrated significant differences favoring SAVR. Rehospitalization was significantly more frequent in the TAVI group, largely because of a higher incidence of decompensated heart failure, which may have been related to the long-term impact of PVR. Indeed, by 5 years, 4 times as many patients in the TAVI group required a further procedure to treat a dysfunctional prosthesis compared with the surgical group, despite a mean age of 81 years and a 5-year mortality rate exceeding 40%, with the associated competing risk.
A secondary intention-to-treat analysis of the PARTNER 2 results showed that all-cause mortality was significantly higher after TAVI (OR: 1.32; p = .002), with a loss to follow-up rate of 16%. Had these losses been assumed to represent patient deaths, the OR would have been 2.12 (p < .0001), even though fewer patients were lost to follow-up in the TAVI group than in the surgical group (11.7% vs. 20.4%).
This 10-year follow-up confirms that, had second-generation devices remained in use, TAVI would not have improved patient survival and would have exposed patients to a higher incidence of further procedures and, probably, impaired quality of life. A patient requiring reintervention for valve deterioration may remain in a poor functional class for months before the indication is established, the procedure is performed, and the patient—usually frail and in their eighties—recovers from another hospitalization.
Nevertheless, and in a manner that is difficult to explain, clinical practice guidelines have long tended to recommend these procedures prematurely, as has occurred with several other interventional therapies in Cardiology. TAVI therefore continues to be regarded as the first-line treatment for these patients because of their higher “surgical risk,” even when relatively unfavorable anatomical factors are present. It is true that advances in procedural planning using systematic computed tomography assessment, increasingly sophisticated and accurate software, and the availability of newer valve generations and sizes have reduced PVR rates and are expected to decrease structural valve deterioration. However, this remains to be demonstrated.
In the same issue of JACC, the “SAPIEN 3 PARTNER 2” report was strategically published. This study compared 10-year outcomes after propensity-score matching between a registry of intermediate-risk patients treated with the SAPIEN 3 prosthesis and the surgical arm of PARTNER 2. At 10 years, all-cause mortality was similar between groups—83.4% after TAVI and 82.3% after surgery (HR: 1.01; 95% CI: 0.91-1.13; p = .82), as was the rate of reintervention (2.0% after TAVI vs. 1.9% after surgery; p = .47).
Although the need for reintervention was numerically and substantially reduced in the TAVI group, the design of this study inevitably introduces major limitations. Event reporting in the SAPIEN 3 registry was center dependent, potentially introducing bias into event adjudication and decisions regarding further procedures in elderly patients with multiple comorbidities, particularly because the study was conducted outside the framework and oversight of the randomized trial itself.
In addition, the comparison involved noncontemporary cohorts, while only the prosthetic valve implanted in one study arm—the transcatheter arm—had been upgraded. In the surgical arm, almost 15% of the implanted prostheses were St. Jude valves, probably including many Trifecta prostheses, or Sorin Mitroflow valves. Both models have now been abandoned because of their high rates of structural valve degeneration. Conversely, the introduction of the Edwards INSPIRIS RESILIA prosthesis may result in a meaningful improvement in structural valve degeneration rates among surgical patients. Ultimately, the findings of this study should be considered hypothesis generating and cannot justify the assumption that SAPIEN 3 has definitively resolved the differences in long-term adverse-event rates identified in PARTNER 2.
Another consistent finding across these studies is the significant adverse effect of transthoracic access on medium- and long-term outcomes. These techniques are now being replaced by transaxillary, transcarotid, and other alternative access approaches. The equivalence between surgical and transfemoral outcomes over such an extended follow-up period should prompt us to reconsider the excellent therapeutic option that surgery represents for these patients despite their surgical risk.
Assuming that alternative nontransthoracic approaches will provide better outcomes, when none has been evaluated in long-term randomized trials, is equally hazardous. Surgery should not be reserved exclusively for patients “in whom a transcatheter valve cannot be implanted by any access route,” as frequently occurs during Heart Team discussions involving this population.
It should also be emphasized that, despite the perceived “invasiveness” of surgery, patient quality of life measured using the KCCQ-OSS was identical in both groups at every follow-up time point. In other words, the less invasive approach did not provide superior quality of life in the short, medium, or long term. Patients should be made aware of this when choosing between the two treatment alternatives.
In conclusion, these results and the considerations discussed above should lead us to recognize that, despite their risk profile, surgery must remain a fully valid option for this subgroup of patients. At present, their surgical risk often leads Heart Teams to regard the transcatheter approach as the first-line strategy, sometimes forcing the indication even in the presence of borderline anatomical features. These include bicuspid valves, severe left ventricular outflow tract calcification that increases the risk of PVR and atrioventricular block, and low coronary ostia requiring additional procedures without established long-term outcomes, such as chimney stenting.
As demonstrated, surgery provides these patients with survival and freedom from further procedures that are at least comparable to those achieved with TAVI. Moreover, it remains unproven that newer transcatheter valve models will not continue to produce worse long-term outcomes than surgery performed using prostheses with established durability.
REFERENCE:
Thourani VH, von Stein P, Mack MJ, Nazif TM, Babaliaros V, Alkhouli M, et al; PARTNER 2 Investigators. 10-Year Randomized Outcomes of Transcatheter or Surgical Aortic Valve Replacement in Intermediate-Risk Aortic Stenosis. J Am Coll Cardiol. 2026 Jun 16;87(23):3309-3321. doi: 10.1016/j.jacc.2026.03.169.
