This article is more than 5 years old. Information may no longer be current.
GARY: 1-year mortality higher with TAVR vs. surgery in real-world population
NEW ORLEANS — In a real-world population of patients with severe aortic stenosis, mortality at 1 year was higher after transcatheter aortic valve replacement compared with surgical aortic valve replacement.
The difference persisted even after propensity matching and was “most probably caused by additional confounders,” Nicolas Werner, MD, from Medical Clinic B, Klinikum Ludwigshafen, Germany, said during a press conference at the American Heart Association Scientific Sessions. Werner also noted that the TAVR population was generally sicker than the surgery population.
The GERMANY investigators analyzed patients at intermediate surgical risk, defined as a logistic EuroSCORE of 10 to 20, who were enrolled in the German Aortic Valve Registry (GARY). After exclusion criteria were applied, Werner and colleagues analyzed 5,997 patients who underwent TAVR or surgical AVR from 2011 to 2013.
The primary outcome of the GERMANY study was 1-year mortality. All-cause mortality at 1 year was 16.6% in the TAVR group vs. 8.9% in the surgery group (P < .001), Werner reported during a press conference. Patients who underwent TAVR were more likely to require a permanent pacemaker (19.1% vs. 5.3%), have more vascular complications (7.7% vs. 1.1%) and have more aortic regurgitation of grade 2 or higher (4.7% vs. 0.4%; P < .001 for all). In contrast, patients who underwent surgery had higher rates of bleeding complications with the need for transfusion (25% vs. 51.5%; P < .001), bleeding with the need for re-intervention (1.3% vs. 4.5%, P < .001) or the need for temporary dialysis after the procedure (2.3% vs. 3.6%, P = .024).
The researchers then calculated a propensity score based on a logistic regression model of baseline characteristics. Patients were stratified into quintiles based on their propensity score. In the overall propensity-matched analysis, the TAVR group continued to have a higher rate of 1-year mortality (15.52% vs. 10.89%; difference, 4.63%; 95% CI, 1.75-7.52; P = .002), which persisted even when the analysis was restricted to those who underwent transfemoral TAVR (difference, 3.48%; P = .021), Werner said.
Patients who underwent TAVR also had a lower BMI (28.2 kg/m2 vs. 27.2 kg/m2; P < .001), higher rates of prior MI (P = .003), atrial fibrillation (P < .001) and moderate to severe concomitant mitral regurgitation (P < .001).
When the groups differed in baseline comorbidities, only in prior cardiac surgery (P < .001) did the difference not favor the surgical AVR group, Werner said. Among other factors, those undergoing TAVR had more prior MI (P = .003), atrial fibrillation (P < .001) and severe mitral regurgitation (P < .001).
When TAVR was indicated, in 90.8% of cases it was based on a decision from the heart team, in 77.2% of cases age was a factor and in 47.3% of cases frailty was a factor, Werner said.
The 88 sites where patients were treated varied wildly in percentage treated with TAVR vs. surgery, ranging from 0% to 100%, he said.
Werner and colleagues determined 13 independent predictors of TAVR, including age (OR per year = 1.23; 95% CI, 1.21-1.25), low to moderate calcium score of the aortic valve (OR = 2, 95% CI, 1.798-2.38), concomitant CAD (OR = 2; 95% CI, 1.7-2.35), pulmonary hypertension (OR = 1.9; 95% CI, 1.614-2.24) and female sex (OR = 1.25; 95% CI, 1.09-1.44).
Limitations of this study include a substantial risk for unmeasured confounding, risk stratification based only on the logistic EuroSCORE, lack of risk scores for TAVR patients, inability to adjust for the medical opinion by the Heart Team, variability in treatment choices among sites and not all clinical variables of inoperability being recorded, he said.
“There was a marked selection bias in clinical reality, with [TAVR] patients being at higher risk,” Werner said. “Intermediate-surgical risk patients undergoing isolated [TAVR] in a real-world scenario have a relatively low in-hospital mortality rate, less than 4%. Even after propensity-score analysis, a significant difference in 1-year mortality rate persisted between [surgical AVR and TAVR], most probably caused by unmeasured confounders. PARTNER II showed noninferiority of [TAVR] compared to [surgical AVR] in a selected intermediate-risk population and GARY showed clinical reality and a reasonable selection of patients in everyday practice.”
Craig R. Smith, MD, chairman of surgery at NewYork-Presbyterian/Columbia University Medical Center, who discussed the results during the press conference, said “one of the things we have to ponder is which of these experiences, the one here today or [in] SAPIEN 3i, which was an intermediate-risk selected population, represents the future in intermediate risk? And, does a mortality gap between surgery and TAVR persist at lower-risk populations? Both of those may pivot on whether we’re talking about centers of excellence or an all-comers situation like this one.” – by Erik Swain
Perspective
Back to Top
The findings from the GARY registry are important to put into perspective comparing with randomized control trial data published earlier this year from the PARTNER IIA study which showed that TAVR was noninferior to surgery.
The researchers from GARY should be congratulated for such a large real-world registry study of all-comers of almost 50,000 patients enrolled between January 2011 and December 2013. The researchers constructed an intermediate-risk cohort of patients who received surgical AVR and TAVR. Their main conclusion is that in this intermediate-risk cohort, TAVR appears to be inferior to surgery although they point out that this could be due to confounders. A deeper dive is therefore necessary to understand this better.
This is a robust registry of all-comers; however, the treatment decision was based on a Heart Team approach rather than through patient randomization. As such, several intrinsic biases come into play. In particular, there was a selection bias toward favoring the high-risk patients to be treated with TAVR, which can definitely affect the interpretation of the results. Despite best statistical methods using a propensity-matching strategy, the results cannot substitute an adequately done and powered randomized control trial. There were several baseline differences between the surgical AVR and TAVR groups, with a higher burden of comorbidities in the latter group including pulmonary hypertension, AF and concomitant mitral regurgitation, all of which might have affected the outcomes reported. The investigators should be congratulated for a very low rate of stroke for both surgical AVR and TAVR, remarkably lower than what was seen in other randomized studies, which could be due to both their expertise and adequate patient selection.
To me, the message from GARY is not that TAVR is bad and should not be considered for intermediate-risk patients with symptomatic severe aortic stenosis, but rather that there are intrinsic limitations of nonrandomized registry data such as those presented. Another message to me is that surgical AVR still remains a very good option in these patients, in particular given the presence of concomitant mitral and tricuspid valvular disease. Just because TAVR can be done, it does not mean that it should be the default option for these patients where the competing valvular risks achieved might confer suboptimal TAVR results. We are doing better in understanding what confers increased risk in these patients and need to continue to monitor the performance of TAVR in the intermediate- and ongoing low-risk populations via randomized controlled trials.
I don’t think GARY will affect the current practice given that we have an intermediate-risk cohort randomized controlled trial with the PARTNER IIA study, but more studies are needed. I look forward to the results of the SURTAVI trial, which will hopefully enlighten us on this subject using a different self-expanding valve platform (CoreValve, Medtronic).
The message is not that TAVR is bad for intermediate-risk patients, but that we need to have a better understanding of the granularity of the data. If we compare apples to apples, perhaps we will not see worse outcomes with TAVR. I would also like to see the data on operator volume by site, because there were some centers only doing TAVR and others doing very little. We know that as you perform more TAVR procedures, the better you become. Would we see better TAVR outcomes if we selected centers that performed at least 100 cases per year? We also need to understand better the impact of concomitant mitral valve disease. This is a trial that is needed. We don’t know what to do with those patients.
This trial does not tell us that we should not do TAVR in intermediate-risk patients. I look forward to the results of the SURTAVI trial, which will be presented in 2017, which will enlighten us on this subject with a different platform, a self-expanding valve (CoreValve, Medtronic).