Feb 20 2026 This Week in Cardiology

Please note that the text below is not a full transcript and has not been copyedited. For more insight and commentary on these stories, subscribe to the This Week in Cardiology podcast, download the Medscape app or subscribe on Apple Podcasts, Spotify, or your preferred podcast provider. This podcast is intended for healthcare professionals only.

In This Week’s Podcast

For the week ending February 20, 2026, John Mandrola, MD, comments on the following topics: EVOLUT Low Risk data, a provocative meta-analysis, DNR orders, targeted hypothermia, good news in HFpEF evidence, and GLP-1s as AF drugs.

EVOLUT Low Risk 6-Year Results and a 5-Year Meta-Analysis of TAVR vs SAVR

There was big news in the transcatheter aortic valve replacement (TAVR) vs surgical aortic valve replacement (SAVR) in low-risk patients this week. JACC published 6-year outcomes. There were surprises.

Recall that EVOLUT Low Risk first published in the NEJM in 2019. It was 1468 patients with severe aortic stenosis (AS) randomized to the self-expanding Medtronic TAVR valve vs SAVR. It was a non-inferiority design with a primary endpoint of death or disabling stroke at 2 years, using Bayesian methods.

That trial found a >99% probability that TAVR was non-inferior to SAVR. Many trials have shown similar outcomes for TAVR vs SAVR at 2 years. Low-surgical risk patients want to know about longer duration outcomes, including valve reliability and re-intervention rate.

Before I say anything about the results, I have to say that both the balloon expandable and self-expanding trialists have been excellent about reporting longer-term outcomes. Other than PARTNER 3’s mysterious change of primary endpoint, the entire evidence landscape of TAVR has been exemplary.

Main Results

The primary composite endpoint continued to be death or disabling stoke. However, what is changed is that the trialists no longer analyze by Bayesian methods, and they no longer use non-inferiority. Now, the analysis is superiority using Kaplan-Meier estimates.

The composite of all-cause mortality or disabling stroke was 23.3% for TAVR vs 20.4% for surgery (difference: 2.8%). This large absolute risk increase did not come close to reaching statistical significance with a P-value of 0.43.

At 7 years the mortality rates were more divergent—27.7% TAVR vs 23.9% SAVR. Cardiovascular mortality was essentially identical between groups (11.1% vs. 11.0%).

The major point of emphasis came when looking at reintervention rates, which were similar at 5 years but divergent in years 6 and 7. Namely…cumulative reintervention rates were 5.5% vs. 3.3% at 6 years (P = .07), rising to 9.8% vs. 6.0% at 7 years (P = .02) in favor of surgery.

You could easily gloss over the medical jargon “reintervention” but I would not. Why? Because reintervention for TAVR can be quite serious if it is surgical (as it was in two-thirds of re-ops) because observational data have shown death rates as high as 13% when you have to operatively cut out a TAVR and replace it with a surgical valve. Reintervention death rates were only 3.3% in this trial, but these were low-risk patients and highly skilled operators. In other words, reintervention in TAVR is serious.

The causes of reintervention are important. Rates of intervention for aortic stenosis (AS) were similar in the two groups. But more patients in the TAVR are had reintervention for aortic insufficiency (AI) than those in the SAVR. That could be important because AS tends to be slow and AI can be acute.

Seemingly perplexed by the higher re-intervention rates the authors did an exploratory analysis looking at the use of post deployment balloon dilatation using a balloon larger than the valve waist—this, they termed off-guidance post-dilatation.

This analysis showed a just barely significantly higher rate of reintervention in patients who had off-guidance dilatation. Proponents could say…aha — just don’t do off-guidance dilatation and all is good. I would caution against this sort of thinking because a) the association could be noise, and b) doctors don’t do dilatation willy-nilly, the do it because they see a potential problem (non-ideal valve expansion or paravalvular leak) and therefore it could be reverse causality or confounding.

There is also a suggestion that older valve types may be causal. Newer iterations are better, but of course we don’t have 6-year data on that.

Comments From a Neutral Martian

The first thing to say is that we must take the view of the patient: even if TAVR is 1, or 2, or 3 percent worse at 10 years, many people will favor the substantially less invasive procedure. I mean…it takes months to recover from open heart surgery.

But many 70-year-old patients with AS would care about what happens when they get close to 80-years old. I am not sure I would, but I am not 70 yet.

The other thing to start with is that TAVRs may be similar to PFA in that not all TAVR’s are the same. The higher re-intervention rates with TAVR have not been seen with balloon expandable valves.

But...But... Not only is the reintervention worrisome, the death signal also looks worrisome to me.

I do not think we should ignore a nearly 3% higher rate of death in the TAVR arm. Dr. David Cohen pointed out to me that the P-value of 0.4 suggests it is not surprising given a null hypothesis of no difference, but I would counter that by saying a typical analysis of Kaplan-Meier curves requires proportional hazards over time. This is not the cause in these trials as the curves cross (early TAVR benefit changes to later hazard) and this means that there are nonproportional hazards. Now… I know this sounds like gibberish, but it is not. How do I know? Because the same journal — JACC — has published a review article discussing the problems (and analytical solutions) for nonproportional hazards.

The second issue I have the higher death rates in the TAVR arm is that the original EVOLUT trial used a non-inferiority design with a non-inferiority margin of 6%, which seems high. But starting with the 3-year analysis the authors changed to a superiority test. I could not find where that was prespecified.

Now at 7 years, the authors write in the discussion section that “no significant differences were observed between the treatment arms for all-cause mortality.” But listen to the numbers: 27.7% TAVR vs 23.9% surgery; difference in higher death rate: 3.7% [95% CI, −2.4% to 9.9%]; log-rank P = 0.29), but the upper bound is 9.9% worse.

Something is changing over time in these the mortality curves. If you combine reports of death from the 4-year paper to this paper you get:

TAVR mortality goes from 9% to 27.7%. That’s an increase of +18.7%
SAVR mortality goes from 12.1% to 23.9%. That’s an increase of +11.8%

Had the authors stayed with their Bayesian design, what would the probability of TAVR having a higher death rate at 6-7 years? I don’t know; I suspect it would be worrisome.

Now, you think I am being too harsh on TAVR. Well, what about the recently published BMJ Heart meta-analysis by Mateo Marin Cuartas et al? This was an updated 5-year outcomes of TAVR vs SAVR in low to intermediate risk patients. They meta-analyzed 6 trials and about 7400 patients.

The median relative risk (RR) for all-cause mortality was 1.12 (95% Credible Interval, 1.02–1.22; heterogeneity τ²=0), with a 99.3% posterior probability that SAVR is superior to TAVR for this endpoint.

For stroke, the median RR was 1.13 (95% Credible Interval, 0.93–1.39; heterogeneity τ²=0.04), resulting in an 88.0% posterior probability that SAVR outperformed TAVR for this outcome.

Proponents of TAVR may argue that a 2025 meta-analysis, first author Rohin Reddy, JACC found 5-year results better for TAVR. Yet that meta-analysis weighted trials with shorter results — specifically the DEDICATE and EVOLUT 4-year trials.

One final thought: UK-TAVI was a pragmatic randomized trial comparing TAVR to SAVR in low-risk patients in the UK. The 1-year results were favorable for TAVR and published in JAMA.

The 5-year results were presented during EuroPCR in May 2025. The results were far less favorable for TAVR:

This report found a statistically significant (and clinically relevant) increase in stroke in TAVR patients; 12.9% vs 7.3% (P = .01) And again there was a numerically increased mortality in TAVR patients. It’s now February of 2026. I wonder why this data is not published.

In sum, my comments on TAVR remain: It’s an amazing procedure for high risk patients with AS. It’s probably a good procedure for intermediate risk patients. But for low-risk patients, I have concerns over reliability, reintervention, and higher stroke rates. To the investigators’ credit, they will present this data, but the curve crossing is worrisome. A young patient with AS needs to know this information, because it’s relevant to shared decision-making.

DNR in the Hospital

JAMA published an interesting viewpoint on documentation of unilateral do-not-resuscitate (DNR) orders — which I did not know existed.

Normal practice or a standard DNR is requested and agreed on by the patient or their surrogate. Yet there are many cases when clinicians know that CPR is not only futile but also totally inappropriate clinically.

In this case, I learned from the editorial, from Drs Piscitello, DeMartino, and Parker, that a unilateral DNR can be placed. A unilateral DNR is a DNR order that does not require a patient or surrogate consent. This is indeed an area of medical ethics controversy.

The most obvious reason for such an order is that CPR — a medical intervention — is felt to be contraindicated, in the same way an ablation or heart catheterization or percutaneous coronary intervention (PCI) would be.

I also learned that clinicians often document unilateral DNR decisions by placing a standard DNR order in the electronic health record, rendering them indistinguishable from DNR orders entered when a patient or surrogate requests to not receive CPR.

The three authors argue that this is unethical because it obscures the unilateral aspect of the decision.

There is a nice figure in their paper about the types of DNR, specifically the unidirectional DNR where there is conflict from when the patient/surrogate wants CPR, and when the patient/surrogate assents more covertly.

When there is conflict, the authors argue this is clearly a unidirectional DNR. However, the authors note the more common assent unidirectional DNR is covert.

They write:

An assent unidirectional DNR decision occurs when a clinician makes a unidirectional DNR decision and the patient or surrogate disagrees with this decision but either does not express disagreement (eg, does not demand CPR) or has a preference to receive CPR but chooses to defer to clinician decision-making.

One of the main thrusts of this short editorial is that currently there is poor documentation of unilateral DNR orders. And these could contribute to inconsistencies in patient care.

For instance, the authors write:

lack of awareness that a DNR order was a unilateral clinician decision decreases the likelihood that clinicians will reevaluate the DNR order for patients whose clinical condition improves.

This has actually happened to me when a patient with VT storm improved and went from DNR to full resuscitation weeks later.

They call for hospital systems to create and implement a unidirectional DNR order unique from a standard DNR.

This unidirectional DNR order, they write...

should include specifications within the order that require clinicians to regularly reevaluate this code status if the patient clinically improves and designate whether patient or surrogate disagreement exists with this order.

One consequence, they write

of such a visible policy would be to increase visibility of these unidirectional DNR decisions and potentially normalize use of unidirectional DNR decisions and influence clinicians to more regularly invoke this mechanism exists.

My Comments

I highlight this important piece and topic because one of the greatest stresses I have in the hospital is contributing to bad deaths. Interventional cardiologists are often put into this situation because events like myocardial infarction (MI), or heart block, or ventricular tachycardia (VT) are often signals of a dying patient.

We see someone who has no chance of meaningful survival from intervention, and further resuscitation is not only not beneficial, but directly harmful, and we hope that a family discussion goes well and we are allowed to provide comfort in the dying process.

But at times, there is no family, or the family is unsure, and there is a need for some form of unilateral DNR orders. I don’t think I have ever had a direct conflict with a family, but surely, we all have had some degree of assent or covert unilateral DNR. I don’t see the ethical tension in this, do you?

If you do see tension, one example I took from a nice discussion on X from the GeriPal blog was that when we do CPR, and it isn’t working, there comes a time when we decide to stop. That is a pure clinical decision. And I don’t really see it as any different from not beginning CPR in the first place.

There are days when I feel my greatest contribution to patient care is helping a patient and family get DNR orders — so as to avoid harmful end-of-life care.

Yet Another Failure of Targeted Hypothermia

I am sorry, I will keep this brief because I am triggered by the fact that for years my Advanced Cardiac Life Support module kept forcing me to go against the mountains of evidence against targeted hypothermia in survivors of out-of-hospital cardiac arrest.

We are better now, but cooling patients who survive cardiac arrest persisted well after multiple trials found no benefit. I’ve written about; I’ve podcasted about it; I’ve tweeted about it, and no one seemed to care. Cooling continued, and when I asked about it in the ICU it was like asking about not giving the patient clean air or water.

JAMA Neurology has 2-year follow-up from the RCT called TTM2 trial. Recall that the original publication of TTM2 was in NEJM in 2021. A total of 1850 patients randomized to targeted 33 degrees hypothermia or targeted normothermia with early treatment of fever.

There was no difference in mortality at 6 months (hazard ratio, 1.04) and no difference in poor functional outcome at 6 months.

This paper looked at 2 year outcomes in the 84% of patients who had follow-up. Again there were no differences in any 4 measures of neurologic outcome — which strongly suggests no long-term benefit.

I wonder how much money we wasted and how many patients were harmed by aggressive cooling over the decade or so that it was done.

Good News in HFpEF Evidence

HF cardiologist and nice guy Ryan Tedford posted on X news of trial design and rationale paper for a device called the APTURE transcatheter shunt system.

Of course I had no idea what this is. Apparently, the device is a left atrium (LA) to coronary sinus (CS) shunt. The paragraph describing the procedure made me smile because I thought: my gosh, these interventionalists are crazy.

You go out the distal CS, inflate a balloon, then a needle that puts a wire into the LA from the CS. Then you balloon the hole between the CS and LA, and then you place the shunt catheter — which then stays there to shunt LA volume to the CS and right atrium.

It’s shocking in its bioengineering. I have no idea if it will work. I seriously doubt it, but if you were a proponent and you wanted to show that it works, you would do it in TRILUMINATE sort of way: they’d do a trial comparing it to white diuretic tablets and measure quality of life.

But that’s not what the ALT FLOW II trialists will do. Not at all. This trial will include a sham control arm with blinding of patients and doctors.

And this is why I highlight this rationale and design paper. The only way to assess subjective endpoints from procedures like this — including AF ablation — is with proper sham controls.

Congratulations to the trialists, first author Dean Kereiakes from nearby Christ Hospital in Cincinnati, Ohio.

GLP-1 as AF drugs

The journal Europace has an observational study worth thinking about. It’s from the big EP group in Milan, Italy. They did a single-center propensity matched study of patients with obesity (BMI > 30), who had first time catheter ablation for paroxysmal atrial fibrillation (AF) between 2019 and 2024.

One group initiated semaglutide within 3 months before or 1 month after ablation vs matched controls who did not receive any GLP-1 therapy.

All patients had implantable loop recorders (ILR) monitors. The primary outcome was AF beyond the 2-month blanking period.

The final cohort included 181 semaglutide-treated patients vs 181 controls with matched clinical and procedural characteristics.

At 18-month follow-up, freedom from recurrence of AF was 80.2% vs. 65.2%; semaglutide was associated with a significantly lower risk of recurrence (hazard ratio 0.52; 95% CI, 0.34–0.78; P = .002).

Weight and BMI decreased significantly in the semaglutide group (−11.8 ± 3.8 kg; −4.0 ± 1.4 kg/m²) compared with controls (−1.9 ± 1.2 kg; −0.3 ± 0.8 kg/m²; both P < .001).

A substantial proportion of treated patients achieved ≥10% weight loss.

My Comments

Methodologically, this is about as weak as it gets. Single center. Non-randomized. Non-blinded. The ILRs are a positive but otherwise it’s low-level evidence.

I highlight, though, because we know that weight loss is strongly associated if not causal in improving AF outcomes in patients with obesity.

We know that weight loss improves left atrial (LA) structural and electrical properties.

We know that weight loss improves blood pressure and glycemic control, which also have positive effects on LA function and structure.

So the findings are mechanistically plausible.

This week, I met the Novo Nordisk representatives who made rounds in our office because oral semaglutide is now available.

I have yet to prescribe these drugs. We were hesitant with SGLT2 inhibitors because they were felt to be in the realm of diabetes doctors. But then it became clear that they were cardiac and renal drugs.

I think it’s the same story with GLP-1 drugs. It’s time. And I am optimistic that GLP-1 drugs may become one of the best AF drugs ever developed. Surely better than dronedarone or sotalol or amiodarone.

Comments

Commenting is limited to medical professionals. To comment please Log-in.

Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.