After taking it on the chin for previously reporting Bayesian estimates, actual 2-year data from the EVOLUT Low Risk trial confirm that transcatheter aortic valve replacement (TAVR) is noninferior to surgery for the primary endpoint of death or disabling stroke.
Among 1,414 as-treated patients, Kaplan-Meier rates for all-cause death or disabling stroke at 24 months were 4.3% with TAVR and 6.3% with surgery (P = .084).
There was also no difference in the individual components of all-cause death (3.5% vs. 4.4%; log-rank P = .366) and disabling stroke (1.5% vs. 2.7%; log-rank P = .119).
Recent low-risk TAVR studies have raised questions about whether there’s a possible catch-up for surgery between 12 and 24 months, given the early mortality benefit from the less-invasive transcatheter procedure, prompting a landmark analysis, John K. Forrest, MD, said during the virtual presentation at the 2021 Congress of European Association of Percutaneous Cardiovascular Interventions, EuroPCR 2021.
“Between 1 and 2 years, there was no convergence of the Kaplan-Meier curves for death or disabling stroke,” with an incidence of 1.9% for the TAVR group and 2.1% for the surgery group (log-rank P = .742), said Dr. Forrest, of Yale University, New Haven, Conn. “The lines were almost superimposed upon each other.”
Session moderator Bernard Prendergast, MD, observed that the Bayesian analysis, which was reported in 2019 and used 12-month follow-up to predict 2-year outcomes, generated questions and criticism over whether this was an appropriate method, compared with traditional Kaplan-Meier analysis. “Indeed, some people accused the investigators of gaming it with this form of statistical analysis.”
To act as a “fact checker,” Dr. Prendergast called in Christopher Cook, MRC, from the PCR Clinical Research Group and Imperial College London. The key methodologic question, Dr. Cook said, is whether Bayesian methods accurately predict actual clinical outcomes in this randomized clinical trial. “The simple answer to this for me, is yes.”
He pointed out that the Kaplan-Meier data for the primary outcome at 2 years were, in fact, numerically better than Bayesian estimates of 5.3% in the TAVR group and 6.7% in the surgery group.
“This validates the use of the original Bayesian methods to estimate patient outcomes in low-risk TAVI patients and, indeed, it may act as an example of where Bayesian methods can be safely applied in order to fast track potentially transformative procedures and technologies to our patients,” Dr. Cook said.
The rate of disabling stroke with TAVR was 1.5% in the new analysis, up from 1.1% in the Bayesian analysis, and 2.7% with surgery, down from 3.5% in the Bayesian analysis.
All-cause mortality, also noted earlier, was 3.5% with TAVR and 4.4% with surgery, whereas the Bayesian estimate was 4.5% for each group.
Dr. Prendergast of St. Thomas’ Hospital, London, said the actual 2-year data are reassuring regarding the statistical tools used and supplement those recently reported from low-risk patients in PARTNER 3.
But, he asked, “what does this mean for practice, what does it mean for guidelines, and how long do we need to wait until we are comfortable and reassured that we can apply TAVI in younger and low-risk patients with a durable outcome?”
Dr. Forrest said that clinicians can be reassured that these patients “are doing very well” but that long-term follow-up is critical.
“We need to be realistic here. We’re really going to be interested in 5- and 10-year outcomes and potentially even thereafter,” he said. “What happens to these valves when they eventually fail? Are superior hemodynamics going to give us longer valve durability in some way or are there going to be other unforeseen things that come up 10 years out? We don’t know those answers.”
TAVR with a supra-annular, self-expanding valve (CoreValve , Evolut R, or Evolut PRO) had superior hemodynamics in the new 2-year analysis with lower aortic valve gradients (9.0 vs. 11.7 mm Hg) and larger valve areas (2.2 vs. 2.0 cm2).
Prosthesis-patient mismatch also favored TAVR, with moderate or severe mismatch occurring in 7.2% and 2.1%, respectively, compared with 19.1% and 4.9%, respectively, with surgery. “We know that this has an impact on long-term outcomes, so it’s important to note that significant difference here,” Dr. Forrest said.
The chink in TAVR’s armor remains paravalvular leak, particularly mild leak, which was significantly higher at 26.6%, compared with only 2.6% with surgery. Moderate to severe leaks were seen in 1.7% and 0.4%, respectively, reflecting the improvement in TAVR with new iterations, he said.
Surgery was also superior to TAVR with regard to the need for permanent pacemaker implantation (7.9% vs. 21.1%). This compares with Bayesian estimates of 6.7% and 23.0%, respectively.
Rates of myocardial infarction remained constant in the two analyses for the TAVR (2.2%) and surgery (1.6%) groups, whereas heart failure hospitalizations improved slightly, from 5.4% versus 7.9%, respectively, in the Bayesian analysis to 5.3% versus 7.1%, respectively, in the new analysis.
Fellow discussant Marie-Claude Morice, MD, Institute Hospitalier Jacques Cartier, Massy, France, highlighted several meta-analyses in different risk patients showing “a lot of good news,” including decreased stroke and mortality rates and the combined outcome clearly favoring TAVR.
“The remaining question is the longevity of the valve, but with 5 years’ follow-up we have for comparison [in high-risk patients], it is the same,” she said. “What this illustrates is that the tidal wave of TAVR is continuing, and we can look optimistically to the future. Is it the nonsymptomatic patients?”
Medtronic funded the study. Dr. Forrest reported grant support from, serving on the advisory board, and proctoring for Edwards Lifesciences and Medtronic. Dr. Prendergast has received grants from Edwards Lifesciences; and speaker/consultancy fees from Abbott, Anteris, and Edwards.
A version of this article first appeared on Medscape.com.