Cardiologists are an inventive bunch — smart money says that they’ll be able to transplant an entire heart via an intravascular approach by 2020, and the more aggressive centers will do lungs at the same time via a radial approach. You laugh, but think about the first time you heard about transcatheter aortic-valve replacement. Avoiding the pain and risk of open heart surgery by using an intravascular approach sounded a lot like mad science to me, but it sure seems to be panning out pretty well. The SURTAVI trial by Reardon and colleagues is the latest multicenter RCT to compare transcatheter aortic valve replacement (TAVR) with a surgical approach (SAVR).
I’m not a cardiologist, though I can be pretty bro-y sometimes, so I understand the confusion. Still, I’ve had patients (and relatives, for that matter) go through the tough decision of which approach to take for severe symptomatic aortic stenosis, and when your cousin’s wife’s grandfather calls you up because you’re the doctor in the family, read on so you’ll have something to tell them.
If You’re Only Going to Read One Paragraph
This study demonstrates that TAVR is noninferior to SAVR for patients at intermediate surgical risk with regard to mortality and stroke at 2 years. This is consistent with the results of a previous trial, and patients at higher risk have previously been shown to have superior outcomes with TAVR. However, the outcomes in patients at low-risk (ATS 30-day mortality risk < 3%) have not been meaningfully evaluated to this point. Given the uncertain outcomes and higher expense associated with TAVR, off-label use should be limited pending availability of this data.
Patients, Intervention, Comparator, and Outcomes
This multicenter, multinational trial recruited patients with severe aortic stenosis (aortic valve area ≤ 1.0 cm² and mean gradient > 40 mm Hg) whose symptoms placed them in NYHA functional class II or greater. Patients had to be at intermediate surgical risk (predicted 30-day operative mortality ≥ 3% and < 15%) according to the Society of Thoracic Surgeon’s model of perioperative risk. A plug-and-chug calculator created using stepwise logistic regression from extensive clinical data, this model has been repeatedly validated and undergoes regular revision. Long story short, it’s a good tool and widely used in clinical practice. If you really love risk models or really hate yourself, feel free to read a little more about its development.
There was some room for clinical judgment in selecting patients, with a multinational adjudication committee able to exclude patients whom they did not feel were reasonable candidates for SAVR based on factors unmeasured by the STS calculation, or to include patients felt to be at intermediate risk despite calculated risk < 3% (most often for frailty or other comorbidities). There are also a whopping 42 prespecified exclusion criteria, which are mostly a list of things that would make one or both of the procedures a very bad idea (thrombocytopenia, active GI bleeding, inability to tolerate any antiplatelet therapy, and certain anatomical features that might make one of the approaches unfeasible). Need for another valve surgery was grounds for exclusion as well, though need for simultaneous revascularization was not, as it is standard practice to perform PCI or CABG along with AVR when indicated. Obviously patients had to agree to undergo either procedure, after which they were randomized to either a transcatheter or surgical approach. Multiple devices were used for TAVR, but all of them were manufactured by the trial sponsor.
The primary outcome was a composite of all-cause mortality or disabling stroke within 24 months post-procedure. It takes a pretty deep dive to find the definition of disabling stroke, which comes from consensus guidelines developed specifically for study of TAVR, which in turn refers to disabling stroke as a score of 2 or more the modified Rankin scale 90 days after the initial neurologic event. Anticlimactically, this translates to a minimum of “Slight disability; unable to carry out all previous activities, but able to look after own affairs without assistance.” So effectively, a disabling stroke is a stroke that limits at least one instrumental activity of daily living, and I can’t believe I just spent 10 minutes tracking that down to realize that disabling stroke means stroke with residual disability. The main secondary endpoints were major adverse cardiac events and stroke of any kind without regard to disability.
A Bit of Context
Smarter men than I could write a novel here, and those with more than a passing interest in this topic would do well to peruse the (fairly concise) Background section of the paper itself. However, a brief history of TAVR:
First they came for the patients nobody wanted to operate on anyway. The PARTNER trial enrolled patients with severe aortic stenosis who were not candidates for surgery and randomized them to TAVR vs standard care. There was a mortality and symptomatic benefit, though a higher risk of strokes (5.0% vs. 1.1%, P=0.06). A slew of trials, most of them funded by specific device manufacturers, then rushed to assess patients at high but not probitive risk for surgery (>15% but < 50% in a representative trial). TAVR consistently had reduced all-cause mortality in this group, and actually fewer strokes as well, despite concerns raised by the initial trial.
You will notice that we have not yet discussed patients at intermediate risk or below; registry data tells us that people TAVRed them anyway. For reasons of advanced age or patient preference, physicians regularly performed TAVR on patients not represented in the above trials. Fortunately, the PARTNER-2 study, a trial similar to the one we’re currently considering, did find noninferiority for patients with STS risk between 4% and 8%. What’s the difference between that trial and SURTAVI, our current undertaking? They’re funded by different device companies. And SURTAVI is newer — it’s HOT Evidence and Trials, guys, not year-old evidence and trials. Try to keep up.
1746 patients were randomized, and 1660 of them successfully underwent either TAVR or SAVR. There were low rates of crossover, with only 3 patients receiving the intervention intended for the other group; however, a fairly high number of patients (71) in the surgery group dropped out before receiving valve replacement. 20% of patients in each group were recommended for concomitant revascularization. Mean age was 79.8 years, ~56% of patients were male, and race was not reported. Rates of comorbidities were unsurprisingly high, with ~34% of patients having diabetes, ~63% having CAD, and > 90% with hypertension. Mean STS risk scores were 4.5% +/- 1.6%. About 15.5% of patients actually had STS scores < 3%, but had frailty or other comorbidity that qualified them as intermediate risk. All of these were similar in both groups.
TAVR was noninferior to SAVR with regard to the primary outcome.
Death or disabling stroke within 24 months occurred in 12.6% of the TAVR group and 14.0% of the surgery group, which met criteria for noninferiority. In terms of procedural complications, early AKI was more common in the surgery group than TAVR (4.4% vs 1.7%), as was cardiogenic shock and post-op A fib. However, TAVR patients were more likely to have major vascular complications (aneurysm or dissection), as well as to need a permanent pacemaker. The as-treated analysis had no significant differences from the intention-to-treat population, and there was no significant variability by clinical site. Echocardiographic endpoints, including AV area and mean gradient, favored TAVR:
Reardon et al., N Engl J Med 2017; 376:1321-1331
So TAVR looks great, or at any rate just as good. But let’s take a second to unpack the postprocedural outcomes as well as the overall trial results. The interesting thing here is that patients undergoing SAVR actually did better than expected, with an observed-to-expected mortality rate of 0.38, which is somewhere between incredible and unbelievable. For comparison, the analogous PARTNER-2 trial the need for pacemaker was actually equal between the two groups, and SAVR patients were more than four times as likely to have life threatening bleeding (43.4% vs 10.4%), a difference not seen between the two groups in this trial. SAVR is worlds better in SURTAVI than it was in PARTNER-2. Use of outdated SAVR techniques was actually one major criticism of the PARTNER-2 trial, so it is good to see SAVR patients with better outcomes in SURTAVI, and reassuring to see the noninferiority of TAVR hold up in spite of this.
If a minimally-invasive approach is noninferior, it’s easy to imagine its becoming the default procedure. However, it’s important to remember that while an expected mortality of ~4.5% may not sound incredibly morbid, the patients in this trial are still in the upper 20% of AVR patients in terms of surgical risk. In the 80% of patients at lower risk than those included here, it’s feasible that surgery could still be a superior approach — surgical outcomes should be better and TAVR outcomes may or may not be. And while yes, TAVR is getting better all the time (the prostheses used in this trial didn’t exist a few years ago, and before long they too will be replaced by something bigger and better), the surgeons are trying new and different techniques at the same time. The 0.38 observed-to-expected mortality ratio is proof of that.
So there’s equipoise on TAVR for intermediate risk patients, and it’s superior in those at high or prohibitive surgical risk. But let’s hold on the off-label TAVR usage for lower-risk patients until the data is in place. Given the money to be made by device manufacturers, I’m sure there will be no shortage of trials forthcoming. After all, TAVR is significantly more expensive than SAVR, in case you need one more reason not to rush the adoption of new technology without a clear benefit.
Read the study here.