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 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 September 9, 2022, John Mandrola, MD comments on the following news and features stories.
Factor XI Inhibitors
The benefit of anticoagulation is that it reduces clot formation, and this decreases clinical events, like stroke; the harm of anticoagulation is that it reduces clot formation and increases the risk of bleeding.
Perhaps because it is unseen, it’s easy to forget that our clotting system is like an elaborate dance or balancing act. Perturbing this balance with anticoagulants can be beneficial but it is surely precarious.
The Factor XIa inhibitors are supposed to the be Goldilocks anticoagulants. Here is how the hope works: factor XI is involved in clot amplification once an injury occurs but activated factor XI has only a minor effect on clot consolidation during hemostasis. The observation, dating back more than a decade, that individuals with deficiency of Factor XI, so-called hemophilia C, have apparent protection from ischemic events, such as stroke, but low rates of spontaneous bleeding. This has led to a new class of drugs in development called Factor XIa inhibitors, which prevent the activation of Factor XI. There are multiple forms of drugs that can do that.
The obvious goal would be to reduce thrombotic events in the coronaries and brain without the Achilles heel of typical oral anticoagulants (OAC) — bleeding.
At the American College of Cardiology (ACC) meeting in the Spring, we heard results of the phase 2 study called PACIFIC AF, in which the small molecule Factor XIa inhibitor asundexian resulted in numerically lower rates of bleeding compared with apixaban.
Three hotline trials at the European Society of Cardiology (ESC) meeting presented early data on Factor XIa inhibitors. This podcast rarely discusses phase 2 trials but given the massive potential of this class of drugs, and their placement in the hotlines at ESC I will give them brief comments. One trial was called PACIFIC MI, led by the Duke team, first author Sunil Rao. They randomized 1600 patients with recent MI in a phase 2 dose-ranging study. The four arms all included dual antiplatelet therapy (DAPT) with aspirin plus a P2y12 inhibitor. The small molecule inhibitor tested is called asundexian. Arm 1 used 10 mg + DAPT; arm 2 used 20mg + DAPT; arm 3 used 50mg + DAPT; and arm 4 used placebo + DAPT.
The primary safety endpoint was bleeding vs placebo. The primary efficacy endpoint was major adverse cardiac events (MACE).
Three main findings:
There was a dose-dependent inhibition of FXIa activity and 50 mg resulted in > 90% reduction.
Over about a year follow-up, there was no increase in major bleeding relative to placebo.
There was also no signal of reduction of thrombotic events.
This was a phase 2 study, which are mostly used for dose-ranging and safety. They don’t recruit enough patients to power for hard clinical endpoints. In the pooled asundexian arms for example, there were 1200 patients, but in COMPASS there were more than 9000 patients per arm.
The authors concluded, rightly, that further study of the factor XIa inhibitor, asundexian, in adequately sized phase 3 studies for patients following an acute myocardial infarction is warranted.
Asundexian was also studied as secondary prevention after stroke in the PACIFIC stroke study. It used a similar methodology, with about 450 patients in placebo, asundexian 10 mg, 20mg, and 50 mg groups.
The primary efficacy outcome was ischemic stroke or “covert infarct on MRI” at 6 months.
There were no sig differences in the event rates of the primary outcome. More worrisome, there was no dose-response noted either.
In a post-hoc exploratory outcome of recurrent ischemic stroke or transient ischemic attack (TIA), there looked to be a dose response, with events lowest in the 50 mg dose.
And in an exploratory post hoc subgroup, patients with atherosclerosis had fewer recurrent ischemic strokes or TIAs in the 50 mg vs placebo arm.
Bleeding rates trended higher with asundexian but did not reach statistical significance.
The authors called these results promising and required validation in adequately powered phase 3 trials, which are trials powered to detect differences in clinical outcomes.
Also at ESC, we heard results of another phase 2 dose-finding study of the small-molecule oral factor XIa inhibitor called milvexian.
The methodologies were similar. The results were similar.
No dose response seen for efficacy though each milvexian dose had lower rates of the primary outcome except for the highest dose vs placebo.
There were numerical increases in major bleeding (BARC Type 3) at milvexian doses of 50 mg BID and higher; the majority were gastrointestinal (GI) bleeds
Surveillance Stress Tests
Last week I told you about a unique screening program for cardiac disease that held promise: DANCAVAS. Well, doing stress tests in asymptomatic people after revascularization is similar: the idea is to pick up disease and intervene before it causes problems.
And for the first three-quarters of my 26 years in electrophysiology, stress testing after bypass or percutaneous coronary intervention (PCI) was routine — a therapeutic fashion if you will. It didn’t hurt that that stress tests — especially those with imaging — are basically money printing machines for cardiology offices.
With the advent of drug eluting stents (less restenosis) and the COURAGE and BARI 2d trials, which found no added benefit to revascularization over optimal medical therapy in stable patients, routine cardiac stress testing declined, but did not stop.
Even the most recent guidelines allowed a 2b recommendation for image-based stress testing “may be considered” in high-risk patients. Of course, high-risk is in the eye of the beholder.
At ESC, we learned the results of the POST-PCI trial, which also made it to the New England Journal of Medicine (NEJM).
POST-PCI trialists, first author, Duk-Woo Park, randomly assigned about 1700 patients with high-risk PCI (either by anatomy or clinical characteristics) to a follow-up strategy of routine functional stress testing at 1 year post PCI or to standard care alone.
The primary outcome was death, myocardial infarction (MI), or unstable angina (UA). Enrolled patients were definitely high risk. One in five had left main PCI, slightly less than half had bifurcation disease, 70% had multivessel disease and/or diffuse long lesions.
At 2 years, the primary outcome occurred in 5.5% of the stress testing arm, and 6.0% in the standard care arm.
This 0.5% absolute risk reduction and 10% relative risk decrease did not meet statistical significance.
We can declare this a negative trial and the end of the foolishness of post-revascularization stress testing, right?
Sadly, it’s more complicated and what follows is a super-important concept in trial interpretation. The clue to the problem comes when you look at the confidence interval (CI) surrounding the hazard ratio (HR).
The point estimate of the HR is 0.90, a 10% reduction. That’s not big but recall that last week the DANCAVAS trial nearly met significance with a 5% reduction in hazard.
In post PCI, the HR of 0.90 had a CI ranging from 0.61 to 1.35. What does that mean? Roughly, it means that if you were to repeat this experiment an infinite number of times, there’s a 95% chance that post-PCI testing results in a 39% reduction in MACE or a 35% increase in MACE.
My friends, this is a terrible outcome of a trial. The outcome is inconclusive. All that money and experimentation yielded a result that tells us little.
Why did this happen? Well, it gets to the tension between having enough patients and paying for a trial.
The authors “powered” their trial based on assumptions. One was that they wanted to detect a 30% reduction in the primary outcome. That’s a huge difference. They also estimated a 15% rate of events in the control arm. But in reality, there was only a 6% rate of MACE.
These two assumptions led to enrolling not enough patients to detect differences, and the wide CI, ranging from a huge benefit from post-PCI testing to a substantial harm from post-PCI testing.
The authors wrote one sentence in the NEJM:
“A 30% relative lower risk of a primary-outcome event with active surveillance with stress testing than with standard care may be too ambitious with contemporary medical therapy.”
They justify this assumption by saying data were scant at the time. They cite two trials — PROMISE (functional stress testing vs coronary computed tomographic angiography [CCTA] in patients with chest pain) and the FACTOR-64 trial (screening CTA for MACE reduction). Both of these trials assumed 20% to 40% reduction in events but both reported no significant differences.
A reader of the NEJM might just focus on the conclusion:
“Among high-risk patients who had undergone PCI, a follow-up strategy of routine functional testing, as compared with standard care alone, did not improve clinical outcomes at 2 years.”
That is misleading if not totally wrong. A better conclusion would be:
Among high-risk patients who had undergone PCI, we could not determine the efficacy of a follow-up strategy of routine functional testing, as compared with standard care alone.
Then in the discussion section, the authors could explain their decision on how many patients to enroll, and what the spread of the CI actually means.
What is also weird is that the chosen editorialist missed this issue. She did not mention the wide spread of the CI and the inconclusiveness of the results. And wrote:
“The POST-PCI trial provides compelling new evidence for a future class III recommendation for routine surveillance testing after PCI.”
While my gut tells me that post-PCI stress testing probably provides benefit only to cardiologists and hospitals, not the patients, this trial does nothing to strengthen or weaken that gut feeling. My prior belief or gut feeling is informed by the trials of patients with stable ischemic disease, such as COURAGE, BARI 2d, and now ISCHEMIA. All the available data tell us that intervening on asymptomatic disease when patients are on optimal medical therapy is not beneficial. In post-PCI, there were higher rates of intervention in the testing arm, and my priors would lead me to believe this would not benefit these patients, even though they are high risk. But, again, the lower bound of the CI went to 0.61 or 39% benefit, but it also went to a 35% risk increase. So, we don’t know.
Maybe it is my age, but I love trials that show that old therapies work for common problems. That I am retired cyclocross racer and the ADVOR study group hails from Belgium — the crucible of cyclocross — only intensifies my interest in the ADVOR trial, first author Wilfried Mullens.
Patients who present with acute decompensated heart failure need fast aggressive action, mainly diuretics. We’ve known for a decade or more that high-dose loop diuretics are the best option. But these don’t always work to decongest the congestion, especially in patients already on diuretics or with kidney issues.
ADVOR randomly assigned about 500 patients with acute heart failure to loop diuretics plus IV acetazolamide vs loop diuretics alone.
Acetazolamide is a carbonic anhydrase inhibitor that reduces sodium reabsorption in the proximal tubule. The drug is old as dirt. It’s been used for altitude sickness and glaucoma.
The investigators measured successful decongestion defined as the absence of signs of volume overload, within 3 days after randomization and without an indication for escalation of decongestive therapy.
Successful decongestion occurred in 108 (42.2%) in the acetazolamide group and in 30.5% in the placebo group (risk ratio, 1.46; 95% CI, 1.17 to 1.82; P<0.001).
Duration of hospitalization was better by a day in the acetazolamide group.
ADVOR was not powered for clinical outcomes but there were no statistically significant differences.
There were no significant differences in safety outcomes.
Comments. These are actionable results. Acetazolamide costs almost nothing. It helps. There were no safety signals. Abrupt and rapid diuresis is important in these patients.
Table 1 is always important: Patient characteristics. These were 78 patients already on loop diuretics who had big BNPs and advanced heart failure with mild chronic kidney disease (CKD), not low-risk diuretic-naïve patients. So, let’s not go overboard.
There was discussion about applying this trial in the era of SGLT2 inhibitors. I don’t quite understand the discussion. While I see benefit of SGLT2 inhibitors, especially in patients with heart failure with reduced injection fraction (HFrEF), I am not sure an oral drug is going to have an immediate impact over the first 3 days. Oral drugs are often ineffective when there is congestion because it can affect GI absorption.
These are patients who are struggling with acute congestion. Emphasis on acute. IV acetazolamide helps get them acutely decongested. Use it for a few days, get the patient out of the hole, and then start SGLT2 inhibitors in patients who have HFrEF. What’s more, I understand SGLT2 inhibitors as much “milder” diuretic agents than acetazolamide. Nephrologists can correct me if I am wrong. What I would like to ask the Belgian trialists is how they got this idea? I mean acetazolamide has been around forever. How come no one thought of this 20 years ago?
I say proficiat to the trialists. I look forward to trying this strategy.
AI in Cardiology
At ESC we heard the results of the first blinded randomized controlled trial (RCT) of artificial intelligence (AI) in cardiology. David Ouyang from Cedars Sinai in Los Angeles, California, presented results of Echo-Net.
This was a blinded RCT of sonographer vs AI assessment of left ventricular ejection fraction (LVEF). Some brief comments before I tell you about this very interesting and as yet unpublished trial.
One of the most common questions in all of cardiology is cardiac function. Knowing the strength of the heart contraction is the key risk stratifier in many decisions in cardiology.
But accurate assessment of EF is one of cardiology’s biggest secrets. Here’s the truth: EF assessment varies a lot. Some of the variability is because of pre-and after-load. Say the patient is excited or volume depleted. EF can vary.
But EF can also vary based on human characteristics. A sonographer typically traces the outline of the heart in diastole and systole and a computer calculates the difference. Lots of squeeze means a high EF. Back in the day, when they let me read echos, we would visually estimate — that looks like a little sluggish: EF, 40%. That looks a lot off, say 25%.
Sit down for this: that still happens today. The sonographer does the planimetry, and the preliminary report outputs an EF, then a cardiologist over-reads it and simply says I agree or don’t agree. Often when they don’t agree it’s because it “looks” more or less vigorous.
I often feel it’s a bit like judges who give tougher sentences before lunch. There’s a lot of play in this. Back at Indiana University in the 90s, Feigenbaum would not allow us to put a number on an EF; we had four categories: normal, mild, moderate, and severe LV dysfunction. So, I hope to have convinced you that EF reading is ripe for non-human automation.
EchoNet was elegant. Echocardiograms that had been performed in 2019 were used to compare sonographer assessment of EF vs an AI assessment of EF. The randomization was blinded. The AI program had been previously validated and published in Nature.
The final step was the blinded cardiologist over-read. The primary outcome was the proportion of studies in which the cardiologist changed the EF more than 5%.
The trial had more than 3700 echocardiograms.
The results favored AI.
The cardiologist changed the AI LVEF reading more than 5% in 17% of images vs 27% for the sonographer.
This was not only noninferior, but also superior.
Some side notes:
Cardiologists could not distinguish between AI and sonographer assessments.
AI saved the sonographer 2 minutes. They do about 20 echos per day, so that’s nearly an hour.
AI saved the cardiologist some seconds.
Comments. This is a single center study. There was no comparison to MRI, which may be more accurate.
EchoNet is the beginning, I think. The tip of the iceberg. Recall that we could not imagine the internet before the internet. Although some may posit that AI can replace human reading, for the immediate term, I see AI aiding human reading.
Pattern recognition like echocardiogram and electrocardiogram (ECG) reading are places where AI will help us. Dave Albert and his Kardia device are doing great work with AI in reading rhythm strips. The Mayo Clinic group is publishing great stuff with AI and ECG reads.
Congrats to the Cedars Sinai team. Please tag the podcast when the paper comes out.
© 2022 WebMD, LLC
Any views expressed above are the author's own and do not necessarily reflect the views of WebMD or Medscape.
Cite this: Sep 9, 2022 This Week in Cardiology Podcast - Medscape - Sep 09, 2022.