Should Transcatheter Aortic Valve Replacement Be the Procedure of Choice for Intermediate-Risk Patients With Severe Aortic Stenosis?

Paul N. Fiorilli, MD; Jay Giri, MD, MPH


Circulation. 2019;138(23):2624-2626. 

Transcatheter aortic valve replacement (TAVR) is considered a reasonable alternative to surgical aortic valve replacement (SAVR) in intermediate-risk patients with severe aortic stenosis (AS) as noted in the 2017 American Heart Association/American College of Cardiology focused update to the guidelines on valvular heart disease[1] (Class IIa; Level of Evidence B-R). This recommendation is based on substantial randomized evidence from the PARTNER II trial (Placement of Aortic Transcatheter Valve) and SURTAVI trial (Surgical Replacement and Transcatheter Aortic Valve Implantation).[2,3] Although high-quality randomized controlled data exist to guide decision making, the extent to which TAVR and SAVR are used in a real-world, intermediate-risk population is not well defined. In this issue of Circulation, Werner et al[4] aim to shed light on the real-world use of TAVR and SAVR in intermediate-risk patients and to describe the outcomes of patients undergoing each approach.

The authors analyzed data from the prospective, all-comers German Aortic Valve Registry (GARY) on 7613 patients deemed to be at intermediate surgical risk on the basis of the Society of Thoracic Surgeon's (STS) risk score undergoing either TAVR or SAVR at 92 sites in Germany between 2012 and 2014. In this national observational study, the average STS score was 5.6±1.1 for the TAVR group versus 5.2±1.0 for the SAVR group. Although this difference was statistically significant because of the large number of patients examined (P<0.001), the clinical significance of a 0.4% average difference in STS score is negligible. These STS scores closely mirror those of patients randomized in the PARTNER II trial, where the average STS score was 5.8%. Superficially, then, it would appear that the present analysis serves as a fair real-world confirmation of the results of this pivotal randomized trial. However, a closer look at the patients enrolled and their specific clinical circumstances casts some doubt on this assumption.

First, patients were enrolled from the years 2012 to 2014, a time period before any prospective randomized data were available on the safety and efficacy of TAVR in an intermediate-risk population. It is likely that the extensive use of TAVR in this population (85% of the 7613 patients studied received TAVR) represented "risk creep" that has been seen in other TAVR analyses,[5] a phenomenon in which real-world populations treated with novel technologies are a bit healthier than those enrolled in pivotal trials. Indeed, a remarkable 25 of the 92 sites in the study used exclusively TAVR to treat isolated intermediate-risk AS during the study period. The primary presented analysis does not properly account for these treatment patterns or for the potential interaction of site volumes of TAVR/SAVR on outcomes. The authors attempted to address this through sensitivity analyses excluding sites that performed exclusively TAVR with additional adjustment for TAVR site volume (results were similar to their primary analysis). Although somewhat reassuring, more optimal methods to account for these issues would have been preferable, including the use of hierarchical models. Hierarchical modeling allows the acknowledgment that practices and outcomes within a site are more likely to be similar than those between sites.

Given the imbalance between SAVR and TAVR use, it is clear that sites and operators had preferences for the use of the 2 approaches in different circumstances. The authors had information on the reasons for selecting TAVR in this group of ostensibly intermediate-risk patients. In >50% of cases, frailty was listed as the rationale for TAVR selection. This is not surprising given the more advanced age of the TAVR cohort in the study (82.5 years versus 76.6 years). Frailty is an important factor not included in the STS score that drives surgical risk.[6] It is common practice for patients with intermediate STS scores and objective markers of frailty to be designated as high risk by heart teams, with patients essentially receiving an appropriate "bump up" in risk categorization as a result of frailty and surgeons looking to avoid SAVR in these cases. Other characteristics not accounted for in the STS score for which we do not have information in the present registry analysis include liver disease, the presence of anatomic contraindications to SAVR such as porcelain aorta and midline left internal mammary artery, and the presence of low stroke volume index. The last characteristic in particular is both strongly associated with outcomes and a common finding in TAVR populations.[7] Consequently, patients who exhibit frailty or these other criteria are much more likely to be treated with TAVR than SAVR. In addition, none of these characteristics were available for inclusion in the propensity model used for adjusted analyses. Thus, the present analysis is likely best viewed as a comparison between patients undergoing TAVR who were at high predicted risk for traditional surgical mortality and patients undergoing SAVR who were at intermediate predicted risk for this.

The results, examined in this light, are fairly remarkable. Despite the higher baseline risk profile of patients undergoing TAVR, unadjusted in-hospital mortality was equal at 3.6% in both groups. Not surprisingly, unadjusted 1-year mortality was higher in the TAVR group than the SAVR group (17.5% versus 10.8%; P<0.001), but this relationship was no longer significant after propensity score matching (17.1% versus 15.7%; P=0.59). Propensity matching attenuated the unadjusted 1-year mortality difference despite the inability to include known covariates associated with both treatment selection and outcomes in the model. It is likely that residual confounding affected the analysis and biased the results in favor of the SAVR cohort, but 1-year mortality was still equal between groups.

Despite the aforementioned limitations, the authors should be congratulated on this attempt to examine the real-world outcomes of TAVR in intermediate-risk patients with severe AS. Even in the presence of probable confounding, these findings are highly reassuring in that they demonstrate results with TAVR in an elderly intermediate- to high-risk population that are comparable to those seen in randomized trials. In addition, the registry confirms prior work indicating that traditional surgical risk scores do not appropriately capture patient risk for this procedure. Several risk scores designed to predict mortality after TAVR have been developed.[8–11] These models generally perform better than models derived from cardiac surgery, but in clinical practice, the surgical risk models are still routinely used. Having appropriate risk assessment for patients undergoing TAVR (especially in predicting long-term clinical risk) is critical, especially as the field progresses toward treating lower-risk patients. Just as important as developing the models will be incorporating their use into clinical care. This will require a culture change in which patients are not just considered for TAVR when they are deemed risky for SAVR. Rather, most patients with AS deserve specific individualized risk assessment for each of the procedures with treatment decisions predicated on these evaluations.

It is important to note that TAVR is a rapidly evolving field that has experienced substantial technological development in a short time frame. Since the study period, several key developments have occurred. First, valve technology has significantly improved, most notably typified by commercially available valves in the United States now designed specifically to minimize paravalvular leak.[12,13] Second, the procedure is performed more frequently via the transfemoral approach with fewer vascular complications given smaller device delivery profiles and improvements in the management of large-bore access. All available data, including that from the PARTNER II trial, have demonstrated superior outcomes with transfemoral access compared with transapical access.[2] In the present study, 21.1% had transapical access, but this rate was only 6.1% in US practice in 2015.[14]Third, a minimalist approach to the procedure, including conscious sedation, has been taken at many institutions, a practice that has been associated with lower in-hospital and 30-day mortality compared with general anesthesia.[15] In the present study, 62.2% of the procedures were performed under general anesthesia, whereas at our institution, >90% of TAVR procedures are performed under conscious sedation. Overall, TAVR today is significantly less invasive and the outcomes are better now compared with when this study was performed.

This begs the question: Is TAVR now the procedure of choice for intermediate-risk patients? A few outstanding issues still need adjudication in this population. First, we should think carefully about subsets of patients in whom the use of TAVR has not been well validated in randomized trials such as younger patients or those with bicuspid valves. Second, it is important to note that the long-term durability of transcatheter valves compared with surgical aortic valves is not well defined. This has significant implications for low- and intermediate-risk patients in whom we might expect long-term survival substantially greater than the 5-year follow-up of current TAVR randomized trials. That being said, the present analysis continues to add to the mountain of data that support a less invasive strategy for aortic valve intervention with a TAVR approach in the grand majority of intermediate-risk patients with calcific AS. This continues forward the seemingly inexorable march toward a coming future in which most patients with AS are treated with TAVR.