Pablo Lamelas –
At ACC 2019 meeting two large randomized trials of patients at low-risk of Surgical Aortic Valve Replacement (SAVR): PARTNER 3 and EVOLUT low-risk. Here we summarize the trials, strengths, limitations, quick meta-analysis and research in context.
This trial tested the efficacy of balloon-expandable TAVR vs SAVR in patients with severe aortic stenosis at low-risk (STS <4%), in terms of death-stroke-rehospitalization. TAVR was associated with superior 1-year death-stroke-rehospitalization compared to SAVR. Also detected no difference in pacemaker insertions or moderate-severe para-valvular leak, while gradients remained similar or slightly superior in the SAVR arm.
High-risk anatomies for TAVR excluded
Patients with bicuspid aortic stenosis or severe LVOT calcium were excluded (38%, which was the most common exclusion criteria). In terms of external validity, recommending TAVR in these excluded patients increases indirectness (extrapolation) of its results.
Both LVOT calcium and bicuspid valves TAVR have been associated with worse outcomes compared to TAVR without those characteristics, therefore this is a risk factor. However, is unknown if these factors are effect modifiers: that in the presence of these effect modifiers patient more likely to benefit from SAVR than TAVR. Why? because bicuspids for instance is also associated with poor prognosis in SAVR. So, the question wether these factors are solely risk factors or also modify the efficacy of TAVR vs SAVR is basically: unknown. Therefore, indicating TAVR in these situations requires some degree of judgement.
On the other hand, what happens with transfemoral-only inclusion criteria of this study? Transapical TAVR has been associated with higher risk of adverse events compared to transfemoral access, so is a risk factor. Now, is an effect modifier? Although not conclusive, the evidence suggest that in intermediate risk-patients transapical approach was associated with less effectiveness of TAVR vs SAVR. So, for a low-risk patients likely better an AVR than a trans-apical TAVR.
What is unknown is if other non-transapical alternative access share this aspect. For instance, axilary-subclavian access (likely the most popular alternative access) can be done percutaneously, which resembles more the degree of invasiveness of a transfemoral rather than transapical. If outcomes of this alternative access share similar efficacy/safety to extrapolate the results from trans-femoral patients is unknown. This uncertainty does not suggest that a SAVR is better, is juts uncertain.
Surgical aortic valve sizes
In this trial surgeons did a great work with their prosthesis sizes, in which 80% where #23 or larger. This aspect also needs to be considered when applying these results to your centre in terms of external validity. Why? Because hemodynamic durability and risk of patient-prosthesis mismatch is closely related to the prosthesis size used.
Heard many critics about this outcome, that “the study was not powered for detecting 30-day stroke difference”, so we “should not trust the stroke results”. Stroke results at 30-days were 0.6% (TAVR) vs. 2.4% (SAVR) (HR 0.25; 95% CI 0.07 – 0.88; P = 0.02), at one year HR 0.38 (95% CI 0.15 – 1.00). Should we trust these results?
First we need to ask one question: What do we mean that “was not powered for”? Clinical trials are usually powered for a specific analysis, in this case 1-year stroke-death-rehospitalization. Therefore, if the study is “adequately powered” for testing other outcomes depend on different factors including: A) The pre-defined effect size (in this case, the expected magnitude of stroke reduction, and pre-defined assuming we are talking about forward sample size calculation, not post-hoc) and B) the absolute number of events (the larger, the better). Just looking the composite primary outcome that includes stroke, then taking out the other outcomes will be underpowered for stroke (alone) since absolute number of events will be inherently lower. So, yes, this study was not powered for this outcome.
But, then we move to the second question: Should we trust this result? Well, if we were doing “trial-based medicine” then its complicated, because this issue of imprecision and statistical fragility leave us with more uncertainties than facts. But, we are lucky to be practicing Evidence-Based Medicine, in which no single study is enough to guide clinical decisions, but the whole body of evidence. Results from randomized trials enrolling intermediate risk patients also showed reduction of strokes. Therefore, this signal of lower strokes in PARTNER-3 likely real, since is consistent with prior literature. Should we trust this result? Qualitatively yes (TAVR likely causes with lower strokes) but careful with its magnitude, which is unlikely to be 70% risk reduction. The magnitude question will be better answered by meta-analysis.
Another critic I heard about the stroke signal is: “the stroke benefit is early, only”. TAVR is a less invasive procedure that happens once, therefore the “safety” outcomes likely have an early impact and then no reasons to keep seeing the survival curves separating over time (like happens with drug trials in which the effect is being delivered over time). Therefore, early benefit with parallel curves thereafter is the expected behaviour of this outcome rather than a surprise to makes us think this is not real.
Rehospitalization as composite outcome
Using composite outcomes has strengths and limitations. They are very good at increasing study power and dealing with competing risks, but many times not easy to interpret. The optimal composite outcome should include outcomes that the pre-specified direction is shared with the rest (this was ok here) but also of similar importance for the patients.
One may argue that rehospitalization is not surrogate, indeed is a patient-important outcome, but is likely out of proportion compared to death-stroke. Indeed, the composite outcome component with the impacting the most in this result is rehospitalization, the least important of the three.
Should we trust this composite outcome then? Again, like in the stroke section, we are lucky to practice Evidence-Based Medicine in which we guide clinical decisions integrating the body of the evidence, and PARTNER-3 is not alone. Meta-analysis from randomized trials of TAVR vs SAVR not only confirmed the non-inferiority but also suggested superiority of TAVR for all-cause mortality and strokes up to two years.
But let’s do this exercise: What if PARTNER-3 was the only study ever done on this issue? Then there will be uncertainty if TAVR is really associated with better outcomes in terms of death-stroke, but still meeting criteria for non-inferiority, which is the actual question we are asking for. TAVR does not necessarily need to show superiority given its less invasiveness and better quality of life (which some parameters are even maintained benefitial at 1-year like 6-minute walk test).
This study tested the impact of self-expandable TAVR vs SAVR in patients with severe aortic stenosis at low-risk for SAVR, in terms of death-stroke. Results showed that TAVR was not inferior than SAVR for death-stroke up to 2-years. However, there was an increase in permanent pacemaker use and moderate-severe aortic regurgitation. In favour of TAVR there was better hemodynamic function of the valve vs SAVR.
Research in context
There were some prosthesis differences. Balloon-expandable TAVR had lower pacemakers and para-valvular leak, when self-expandable TAVR better systolic gradients and valve areas. These differences should be taken with some caution, since is and indirect comparison. For instance. EVOLUT low-risk did not exclude 1 in 3 patients severe calcium in the LVOT as PARTNER-3 did, and this may impact the para-valvular leak results.
Recent meta-analysis including all the trials (independent of risk) demonstrated that TAVR was associated with lower mortality and stroke up to 2 years. It was also associated with lower atrial fibrillation, major bleeding, and acute kidney injury, while SAVR was associated with lower pacemaker rates. Although the increased risk of pacemaker was driven by self-expanding valves, even balloon-expandable prosthesis demonstrated a small (20% relative risk) increase.
What has been demonstrated in that meta-analysis, is 1) the absence of interaction by surgical risk: patients benefited similarly regardless if they were high- moderate- low-SAVR risk; and 2) the mortality benefit driven by the transfemoral approach.
SAVR likely remain the standard for low-risk (STS) and higher anatomical risk (Bicuspids, high-risk of coronary obstruction, non-transfemoral, etc). We have excellent 5+ years durability data. If longer term (10+ years) durability is needed or not before doing SAVR in young non-high risk patients is a present debate. So far, a cut-off of 75 years for considering trans-femoral TAVR as a first option seems reasonable. This age-limit may be affected by durability data coming in the next few years.