Stark Mortality Differences Seen in GARY Registry Tracking TAVR and Surgical Replacement

The results don’t mean surgery is the better choice, said several surgeons, just that real-world choices aren’t reflected in randomized trials.

Stark Mortality Differences Seen in GARY Registry Tracking TAVR and Surgical Replacement

New Orleans, LA—New findings from Germany show that intermediate-risk patients with aortic stenosis enrolled in a national registry there faced significantly higher 1-year mortality if they were treated with transcatheter aortic valve replacement (TAVR) rather than with surgical valve replacement (SAVR).

But according to multiple experts gathered here at the American Heart Association Scientific Sessions 2016, the results reflect the “real-world” realities of the patients currently being selected for each procedure and no amount of analyses can level the playing field in terms of the many factors that could contribute to the survival difference.

Nicolas Werner, MD (Medical Clinic B, Klinikum Ludwigshafen, Germany), who presented the German Aortic Valve Registry (GARY), noted that the survival difference favoring surgery persisted even after propensity-score matching but said additional confounders were almost certainly at play.

“This registry tells us that clinical data and retrospective analyses are something totally different from randomized trials, and we should be very careful in interpreting the data of these retrospective analyses,” Werner told TCTMD.

Almost 50,000 patients were enrolled in the GARY registry between January 2011 and December 2013, of whom 22.7% were deemed to be at intermediate surgical risk (logistic EuroSCORE I of 10-20%). After exclusions, 5,997 patients with isolated TAVR or SAVR were included in the analysis.

Baseline differences between groups were striking, with TAVR-treated patients showing higher-risk characteristics across the board, including greater age, higher surgical risk scores, worse heart failure, more prior MI, more atrial fibrillation, and many others. In an analysis looking at reasons underlying the choice of TAVR over surgery, the most frequent reason was patient age, followed by frailty.

Importantly there were also major differences in the proportion of patients treated with either approach according to the hospital site, with some sites performing no TAVR cases and others 100% TAVR.

Many of these things likely played a role in the significant differences in all-cause mortality seen in-hospital, at 30 days, and at 1 year.

All-Cause Mortality in GARY: 2011-2013




P Value





At 30 Days




At 1 Year



< 0.001

Next the authors did a propensity-score matched analysis, stratifying patients according to their level of risk by quintiles. Here again, however, a significant overall difference was seen in rates of all-cause mortality between the surgery-treated and TAVR-treated patients (10.89% vs 15.52%; difference 4.63%; 95% CI: 1.75-7.52%). Looking only at transfemoral TAVR patients, who made up 75% of this group, the differences was smaller but still statistically significant.

In his remarks to the media, Werner stressed that a number of confounding factors simply cannot be accounted for. Frailty, in particular, was not recorded for surgical patients, but it was for TAVR patients. The decision to choose TAVR over surgery was also made by a heart team in over 90% of patients and “it is impossible to adjust for the ‘medical opinion’ of a heart team,” he said. Finally, given the clear preferences for one type of procedure over another, treatment site likely also affected patient outcomes at some locations.

Craig Smith, MD (Columbia University Medical Center, New York, NY), zeroed in on this last point in his discussion of the study and, like Werner, stressed the many ways in which retrospective registry results cannot compare with clinical trial evidence. Smith was co-principal investigator for PARTNER 2, the only randomized controlled trial to date in intermediate-risk patients.  

“A propensity-score system like the one they used is the best way to attempt to [compensate for a nonrandomized analysis],” he observed. “But I must say that, however this came out, whatever direction the propensity score drove this, it’s hard to imagine that a propensity score could balance all these marked baseline differences, because these are very different groups.”

Martin Leon, MD (Columbia University Medical Center, New York, NY), the other co-principal investigator for PARTNER 2, also commented on the study, saying he was “not surprised” by these findings.

“GARY has consistently shown outcomes favoring SAVR versus TAVR which are different than the randomized trials in other populations,” he told TCTMD by email. “The designation of intermediate risk was a logistic EuroSCORE I of 10-20%, and the comorbidities outlined certainly suggest multiple differences in patient characteristics between SAVR versus TAVR, favoring a higher-risk cohort in the TAVR patients. This many confounders cannot be ‘erased’ with the propensity-score analysis methodology. Not to mention different devices, different generations of the same device, low- versus high-volume operators, and multiple other factors which make this sort of comparison retrospective analysis nonscientific and clinically without rigor when compared to carefully performed prospective randomized trials.”

In his remarks, however, Smith showed “hot off the press data” from the New York State cardiac surgery database showing that the annual mortality rate for aortic valve replacement across all patients and all surgeons there is now at 1.76%. “That is the kind of performance standard that TAVR will have to remain equivalent to in lower-risk populations,” Smith said.

Keeping a Door Open

Another surgeon, Frank Sellke, MD (Brown University, Providence, RI), had similar comments, also highlighting the problems of “optimally matching the groups.” But he does see the data as a reminder that surgery should not be overlooked as an important option for lower-risk patients with aortic stenosis, as the less-invasive strategy moves into progressively lower-risk and younger patients.

“I think we need to have a little bit more equipoise in recommending that TAVR can be performed in intermediate-risk patients [but that] surgery should also have a lot of consideration,” he told TCTMD. Citing the SAPIEN 3 intermediate-risk study, Sellke continued, “There’s one study suggesting that TAVR has an advantage, but we should not make changes in guidelines based on a single study. I don’t think we should completely change our practice until we have the results from multiple studies.”

Smith, too, pointed to the different findings from SAPIEN 3, another trial that used propensity-score analysis to match intermediate-risk TAVR patients with similar patients from an earlier cohort who were treated with SAVR. “What do we worry about looking ahead, or what can we conclude?” he asked. “One of the things that we need to ponder is which of the experiences—the one presented today or SAPIEN 3—really represents the future of intermediate risk. And will a morality gap between TAVR and surgery persist in low-risk populations? Both of these may pivot importantly on centers of excellence and real-world experience such as this one and of course, durability remains the big unknown in the long term, particularly in lower-risk patients.”

Discussing the results with the press, Werner pointed out that these procedures were conducted prior to the release of the PARTNER 2 intermediate-risk TAVR results. “So these patients, who were treated with TAVI at this time, from 2011 to 2013—they must have had a very good reason not to be operated, and that’s the bias we have in our results,” he said.

Following up on that, press conference moderator and cardiovascular surgeon Timothy Gardner, MD (Christiana Hospital, Newark, Delaware), pointed out that while these were “intermediate-risk patients” by STS and EuroSCORE I, “some of these patients would not have gone to surgery, even though they are intermediate-risk” according to available risk prediction tools.

One more surgeon weighed in, echoing the observations of his colleagues on the hazards of comparing these two groups. “At the end of the day you can have a bowl of apples and a bowl of oranges,” said David Taggart, MD (University of Oxford, England). “And you can do as much statistical analysis and remodeling and adjustment as you want, but at the end of the day you have a bowl of apples and a bowl of oranges.” 

Shelley Wood is the Editor-in-Chief of TCTMD and the Editorial Director at CRF. She did her undergraduate degree at McGill…

Read Full Bio
  • Werner N. Patients at intermediate surgical risk undergoing isolated interventional or surgical aortic valve replacement for severe symptomatic aortic valve stenosis. One year results from the German Aortic Valve Registry (GARY). Presented at: American Heart Association Scientific Sessions 2016. November 13, 2016. New Orleans, LA.

  • Werner reports having no conflicts.