Editor's Corner: Transcatheter Aortic Valve Choices: Observational Studies, Randomized Clinical Trials and Truth
In the world of transcatheter aortic valve replacement (TAVR) there are a number of valve choices available to TAVR centers particularly outside the U.S. In the "real world" of TAVR in the U.S. there are only two - the Edwards and Medtronic valves (with Boston Scientific’s Lotus valve the new contender). For the past 10 years, TAVR trials have been designed to tell us about safety and efficacy of the two major players. These randomized clinical trials (RCTs) pitted TAVR against surgery (save for PARTNER IB that predictably showed medical therapy for aortic stenosis (AS) could not match relief of the stenosis itself). We learned in surgical high-risk patients (by STS score) TAVR matched surgery, and even seemed superior to surgery in many ways. As RCTs enrolled patients with lower and lower risk, TAVR kept pace with constantly improving outcomes and in some trials even seemed to lessen stroke risk, for example.
RCTs are, for us, the current gold standard. They are “scientific,” and their outcomes are “truthful.” We believe them. At the annual ACC, AHA, TCT and ESC meetings, the late-breaking clinical trials symposia are repeatedly the highlights of new information that are eagerly anticipated. For the trials presented at such venues, expert panels evaluate the importance of the outcomes and weigh possible confounders that may have made outcomes less or more clear. Their analysis is always helpful, even when occasionally an unanticipated outcome appears, such as the high stroke rate in the surgical arm of SURTAVI, and is given over to “chance.”
RCTs, however, have drawbacks we tend to overlook. Here’s a reminder of some shortfalls:
- Completing an RCT is only possible for an intervention patients will undertake. If most patients do not join, the population is too selected and results applicable only to a small group.
- The results are frequently short-term (e.g., TAVR trials with 30-day outcomes).
- RCTs may use surrogate vs. long-term clinical outcomes, or combination outcomes wherein only one of them and perhaps the least clinically relevant “drives” statistical relevance. Participants may not be representative of a true clinical population.
- If the trial is complex or requires multiple tests/interventions in follow-up, patients may not return, which can compromise/minimize the dataset.
RCTs, so far, have told us little about which of the TAVR valves available in the U.S. we should choose. Enter the “observational study.” Despite critics, results of observational studies have always been part of our knowledge base. Interventional cardiologists are always looking for new therapies or procedures that are first described in observational studies. RCT zealots look at observational studies as second-class citizens (until reminded of the classic lack of a randomized trial confirming the efficacy of using a parachute when jumping out of an airplane). But where would we be if Andreas Gruentzig, MD, had not sent the description of his first angioplasties in his landmark letter to The Lancet in 1978?
Observational studies have great merit, but potential biases. Yet they are still “scientific.” Thus, they have become generators of new hypotheses, can determine need for RCTs and identify safety signals, and are vehicles to identify size of some effects and a means of identifying potential therapies for orphan/rare diseases. Most importantly, for finding the “truth” in results of RCTs, observational studies are perfect. They help us understand if results of an RCT, derived from a controlled, selected population can be extended into “real world” care. Without such information, our ability to “generalize” results of an RCT are limited.
A recently reported example is FORWARD – a real-world experience with a self-expanding transcatheter aortic valve prosthesis presented at EuroPCR. This prospective, single-arm, multinational observational study reported 30-day results of the Medtronic Evolut R valve implantation in 1,038 patients with severe AS. The Evolut R valve has unique features. It minimizes paravalvular leak, has a small profile and can be repositioned or retrieved. FORWARD supports the clinical safety and effectiveness of the Evolut R System.
FORWARD indeed seems to reflect the real world. The mean age of patients was 81.8 years, 64.9 percent were women and the mean STS score was 5.5 percent. The mean transvalvular aortic valve gradient post implantation was a laudable 8.5 ± 5.6 mm Hg. Post-implant moderate or severe aortic regurgitation, all-cause mortality and disabling stroke all occurred in <2 percent of patients. Most of these outcomes are quite similar with those reported from trials using the Edwards valve systems.
What do we learn from FORWARD? First, it expands our understanding of how Evolut R is used in clinical practice and how it performs in a large cohort. Mortality and stroke rates are surprisingly low, as perhaps one might expect from an observational cohort not rigorously studied by neurologists post TAVR, but are in keeping with previous RCTs of the Medtronic valve. Problems are the short patient follow-up (though longer term follow-up is promised) and the >15 percent rate of pacemaker need (left to discretion of attending physician without documentation of reasons).
Will this report convince TAVR operators to use the Evolut R valve over its competitor? Unlikely, especially because TAVR operators have their “favorites” and the trial was not designed to do so. And that – despite all the information we have from observational trials – is the crux of the TAVR issue. To understand differences between, for example, using the Medtronic or Edwards valves, we need to go back to an RCT.
In the TAVR world only one trial has compared two valve types. REPRISE III compared the Lotus valve and CoreValve (about half the CoreValve implants were Evolut R). For the primary effectiveness endpoint (all-cause death, disabling stroke and moderate or greater paravalvular aortic leak at one year) Lotus was superior to CoreValve (16.7 vs. 29.0 percent). But the endpoints were driven primarily by moderate or greater paravalvular leak (Lotus 2 percent vs. CoreValve 11 percent, the latter probably enhanced by use of an earlier generation Medtronic valve), and the disabling stroke rate with Lotus was 3.6 percent vs. 7.1 percent with CoreValve (a surprisingly high figure not reflected in the FORWARD “real world” report for Evolut R). All-cause mortality, cardiac death and stroke were non-inferior. A composite safety endpoint of all-cause mortality, stroke, life-threatening and major bleeding events, stage 2/3 acute kidney injury or major vascular complications through 30 days also showed Lotus was non-inferior to CoreValve.
That leaves a mind-numbing conglomeration of data – some easily understandable, some not. Why was the stroke rate so high in the Medtronic cohort in REPRISE III? What effect will the pacemaker rate in FORWARD have on patient decision making, etc? We process all this by hoping that clinical situations (e.g., previous pacemaker implants in some TAVR candidates) will help make decisions, but the answer to “Which valve is better?” for current use in the U.S. lies in the outcomes of an RCT that I doubt we will ever see. So, we are left with a series of observational trials of varying populations, with varying endpoints, technology and outcomes to make decisions that hopefully are best for each patient. It is comforting that clinical judgment is still in vogue.
Tweet this article: Tweet
|Click the cover image above to read the latest issue of Cardiology Interventions in e-pub format or click here to read it on the web!|
< Back to Listings