Feb 17, 2023 This Week in Cardiology Podcast

John M. Mandrola, MD


February 17, 2023

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 17, 2023, John Mandrola, MD comments on the following news and features stories.

Screening with ILR

The first paper and topic highlights one of the most important challenges of the modern age of medicine. That is, the (loosening) connection between diagnosis of a condition and our ability to improve outcomes. I don’t want to call it over-diagnosis. It’s not exactly that. It’s more an issue with our technology redefining disease and perhaps normality.

Before I tell you about a super-important paper in the atrial fibrillation (AF) space, I want to point you to a Tweet from Dr. Ade Adamson, an academic dermatologist at UT. This was a picture he published in the NEJM in 2021. It is a graph showing two lines. One was a steep rise in the diagnosis of cutaneous melanoma over the past 40 years. The other line was a flat line showing absolutely no change in mortality from melanoma.

A remarkable post-hoc analysis of the LOOP trial details a similar phenomenon in cardiology. First, I want to say that many times, post-hoc analyses are data-dredging exercises aimed at marketing not science. Not this one.

First author Soren Diederichsen, a young MD, PhD, from Copenhagen, along with other co-authors of LOOP, were interested in the prevalence and prognosis of bradyarrhythmia in patients implanted with an implantable loop recorder (ILR) in the LOOP trial.

Slow the treadmill or bike speed, this paper will teach us important lessons. First a brief review of LOOP, one of the seminal trials in cardiology this decade.

  • Published in Lancet in 2021, the trialists randomly assigned more than 6200 patients without AF but with high stroke risk features to screening with an ILR or screening with usual Danish medical care.

  • After more than 5 years, they found 3 times more AF in the ILR group. That led to nearly 3 times more use of oral anticoagulants (OAC). Sounds great. This had to be good.

  • Yet the main result was a 20% reduction in stroke that did not meet statistical significance P=0.11, and a 26% increase in major bleeding that also did not reach statistical significance.

  • In other words, LOOP was a clearly non-significant trial. Even if you believed the 20% reduction in stroke was not noise, it was countered by an even greater increase in major bleeding.

  • Screening for AF with the best modality for screening — the always on ILR — failed to change outcomes. Despite the 3 times increase in diagnosis.

See the pattern; it’s just like the melanoma picture of Dr. Adamson.

Now to the post-hoc study. The authors, and anyone who uses ILR devices, know that you may implant them looking for AF, but they also find other things. Scary things, like sinus pauses and heart block. Surely, knowing these things will help people, right? Well, let’s delve into the findings.

The primary endpoint of this study was brady episodes, pacer implants, syncope, sudden cardiac death.

“Brady episodes were classified as one of (1) SND defined as inadequate heart rate compared with physiologic need due to sinus bradycardia, pause or arrest, exit block, or chronotropic incompetence; (2) low-grade AVB defined as first- or second-degree heart block with a P:QRS ratio of 2:1 or lower; or (3) high-grade AVB defined as complete heart block or second-degree heart block with a P:QRS ratio of at least 3:1.”

They excluded asymptomatic sinus bradycardia (SB), sinus pauses or atrio-ventricular (AV) block during sleep of anything less than 3.5 seconds or longer. They also excluded heart rate (HR) < 30 during sleep.

  • Bradyarrhythmia was detected in 4% vs 21% in the ILR group – hazard ratio of 6.21.

  • The detected bradyarrhythmia was not symptomatic in 80% of the ILR group vs only 24% of the control group. (Which makes sense because you’d expect that the control group makes diagnoses far more often based on symptoms. I will come back to that.)

  • The most common bradycardia was sinus node dysfunction (SND) followed by high-grade AV block.

  • Pacers were placed in 4.5% of the ILR arm vs 2.9% of the control arm; hazard ratio of 1.53, which was highly significant.

  • But there was no difference in syncope (2.7% vs 2.2%) nor sudden cardiovascular (CV) death (1.1% vs 1.2%).

Comments. I cannot emphasize enough how important these observations are. Important in cardiology but important also in the philosophy of modern medicine. First let’s do the cardiology part.

  • This analysis shows that if you do more rhythm monitoring you pick up things we once thought were abnormal. I used the past tense there on purpose.

  • The problem was that these norms were established in an era where ILRs and smart watches could not even have been thought of.

  • It’s clear now that patients and doctors alike have not re-calibrated what we call normal.

  • One of the revelations of this paper is that when you monitor people with higher risk features you find that bradycardia—as historically defined-- is common. Six times more common than those in a routine care arm.

The study was done in Denmark, so the rate of pacemakers was only 53% higher in the ILR arm. Had this study been done in the United States, you can be sure it would have been higher. Still, to this day, I am referred patients for consideration of pacers due to bradycardia without symptoms. The authors speculate that there may have been over-treatment with pacemakers. Maybe their Danish-ness makes them subtler than normal.

I’d say it differently. Of course, there was over-treatment. If you put in oodles more pacers and find no difference in syncope or sudden death, then by definition there was over-treatment. But there is so much more to say.

  • The study’s primary aim was to assess the value of finding AF in patients-most-likely-to-benefit. And it did not show benefit.

  • But this post-hoc analysis elegantly shows the off-target effects of screening. You pick up other stuff too. Like bradycardia.

  • Another example: CT scans for coronary calcium populate lung nodule clinics. Lung nodules lead not only to anxiety about cancer, but also to downstream shenanigans, such as more CT scans, biopsies, even surgeries. The shenanigans are much more likely to benefit the health system than the patients.

Ok, now let’s move to the philosophical matters. Here I see clash of forces... of modernity and medicine.

  • Modernity – especially in the US population — is creating diets of mostly processed food, increasing rates of obesity, and lower rates of exercise. Modernity is creating a population of metabolically unhealthy people.

  • But modernity has also created amazing technology, such as loop recorders, smart watches, CT scanners, and even devices that can measure troponins through the skin. These technologies are being used to diagnose more and more conditions, which then leads to more and more intervention.

  • It’s like a printing press for dollars for anyone in the healthcare industry, but a tsunami of unhealth for patients. I know I am sounding emotional about this, but gosh, I hate to see what is happening.

I love what technology has allowed us to do in caring for the sick. The advancements have been astounding. But this move to use technology transforms healthy people into patients. The 60-year-old who was cycling every day to work now worries about having coronary calcium. The patient with heart failure doing okay on meds, now is told they need an internal monitor in their pulmonary artery. And perhaps the worst of all: the elderly patient with unsteadiness of gait who is convinced to have a foreign body placed in the left atrial appendage (LAA) instead of taking an OAC.

If you are young, and listen to this podcast, it is not too late. You can rise up and tell people that the emperor has no clothes. This is all nuts. The solution to better health is not more technology; it is to eat better food and go out and play.

If an ILR in high-risk patients doesn’t help, how in the world do we think digitally monitoring younger healthier people will work? Health is not complicated: there is food, exercise, and luck. Most everything else is incremental or worse.

Finally, Dr Diederichsen and colleagues state that they will share the raw data upon request.

Thrombolysis in Mild Stroke

Sue Hughes covered an interesting randomized controlled trial (RCT) called ARAMIS that was presented at this year’s International Stroke Conference (ISC). This was thrombolysis vs dual antiplatelet therapy (DAPT) with clopidogrel loading in patients with mild nondisabling stroke (National Institutes of Health Stroke Scale [NIHSS] score ≤ 5).

The bottom line is that the more patients in the DAPT arm achieved a high-functioning level modified Rankin score (mRS) of 0-2 than in the alteplase arm. ARAMIS was a non-inferiority trial and it easily met non-inferiority.

I am not going to say much more about this because it’s not published, and we need the details. The study was done in China and the definition of non-disabling stroke was debated.

But I want to at least mention that there is a very provocative study coming to a journal. It’s provocative because right now these patients get lytic therapy.

It is also similar to the PRISMS trial, published in JAMA in 2018.

PRISMS randomly assigned patients with NIHSS scores of 0-5, mean enrolled 2.3.

  • There were about 300 patients.

  • There were no differences in the primary outcome of mRS 0-2 or good functional capacity.

  • But there was a 3-fold higher rate of intracranial hemorrhage.

  • PRISMS was terminated early by the sponsor for “financial” reasons, so there are wide confidence intervals.

This new study is another opportunity to remind everyone that the practice of lytic-therapy-for-stroke sits on very shaky evidence. Many trials have been done, and only two yielded positive results — the NINDS study, which had substantial differences in baseline characteristics that favored tPA. Last year neurologist Ravi Garg connivingly showed that these baseline differences were not likely due to the play of chance but rather randomization errors.

Garg also showed that in the most representative meta-analysis of lytic trials by Hacke and colleagues in 2018, in the International Journal of Stroke, there was enough missing outcome data to warrant uncertainty regarding the benefit of this therapy.

  • Finally, neither of the two positive lytic trials (NINDS or ECASS) have withstood reanalysis, NINDS due to baseline differences in stroke severity; ECASS due to statistical error wherein the authors did not account for baseline imbalances.

  • Alper and colleagues used the raw data to make these adjustments and their re-analysis found no significant differences.

  • I know the stroke neurologists disagree with me. They’ve published reanalyses that upheld the positive findings.

I am not saying lytic therapy does not work in some patients. I may be wrong about my doubts. There may be a good explanation why the majority of lytic trials aren’t positive or why independent authors who re-analyzed ECASS and NINDS did not come to the same positive results as the proponents. The missing data in the meta-analysis may not matter. Nor the randomization errors in NINDS.

Mechanical thrombectony trials are much stronger. Yet the whole “stroke center” industrial complex turns on the use of lytic therapy. And the data for lytic therapy is shaky.

Imagine if there were enough investigators with the courage to redo the NINDS study in the modern era, we might be surprised. And... if lytic therapy for stroke was overturned, what a massive reversal that would be! A therapy that doctors get sued for not using fast enough could be proven ineffective.

I will let you know when the new paper comes out. And I warn you: if you look deep into this data, you cannot unsee it.

Advanced Practice Nurse vs MD Care

A new study with super-interesting methods of simulating randomization suggests that I may have been wrong about advanced practice RN (APRN) vs MD-led care.

In 2020, and again in 2022, I argued that nurse-practitioner (NP)-led care is not inferior to MD-led care. The columns generated love from the APRN community and angst from my doctor friends.

This new paper is a working paper in the NBER journal – National Bureau of Economic Research. The authors are not medical doctors.

  • They studied use of NPs in emergency departments (ED) in the Veterans Administration (VA) system.

  • They found that NPs increase resource use and achieve worse outcomes compared with doctors. The differences were substantial.

  • Length of stay increased by 11%; costs of care in ED went up by 7%; and the 30-day preventable hospitalization rose by 20%.

The controversy of this paper comes in its design. The authors used an instrumental variable analysis to achieve quasi randomization. Here it gets complicated.

A well as a community electrophysiologist can understand it, the authors leveraged two factors: NPs in the VA system can treat independently, and there is an instrument measuring the exogenous variation in the availability of NPs on duty to decide whether a case is treated by an NP or a physician.

  • In a sample of 1.1 million ED visits, their approach compares patients arriving at the same ED and during similar times that differ in the number of NPs on duty.

  • They show that the number of on-duty physicians declines with the number of on-duty NPs, and the number of NPs on duty strongly predicts whether an arriving patient will be assigned to an NP versus a physician.

  • Under the plausible assumption that patients arrive quasi-randomly within cells of ED stations and time categories, this instrumental variable design allowed them to study the effect of NPs on patient resource utilization and health outcomes.

The big assumptions were that arrival of patients are random and no other exogenous factors affect the MD:NP staffing on any particular day. The authors did careful analyses and made strong arguments that their instrumental variable was sound.

I’ve communicated with many experts. The study only can succeed if the instrumental variable is uncorrelated with everything else. I am not sure they convinced me. Because it seems NP and MD staffing would not be truly random.

I really need to better understand this paper. Econ papers read differently than medical papers, but my guess is that if you really get deep into this, you will learn a lot.

  • The issue of NP care has immense impact on medical education. This study suggests that NPs in an ED setting do not compare favorably with MDs on an economic basis, even given the lower salaries.

  • But if, in other settings, the performances are similar, as many studies have suggested, then the question becomes, why do doctors need more than 10 times the hours in training? And why does there need to be such restriction into the guild of who can practice medicine.

The number of NP is now about one-third the number of MDs. The need for healthcare is outstripping the supply of doctors. So there is clearly need to have more empirical data in this situation.

Can LAA Really Reduce Mortality?

The prominent journal Circulation has published a comparative effectiveness study of older patients with AF who were treated with OAC vs LAA occlusion (LAAO) by sex. Comparative effectiveness is a buzzword for non-random comparisons. Here they used Medicare claims data from 2015 to 2019 of LAAO-eligible patients.

One group got OAC. The other LAAO. Of course, they needed to use statistical techniques to match these patients. Propensity matching in this case. The outcomes of interest were mortality, stroke, and bleeding. They matched about 4000 women and 5000 men, 1:1.

  • In men and women, LAAO was associated with a 50% reduction in death!

  • In men and women, LAAO was associated with 35% reduction in stroke.

  • Bleeding risk was higher after the device but lower if you exclude the 6 weeks after the procedure. But total bleeding ) was no different in females and higher in males in the LAAO group.

  • The Kaplan Meier (KM) curve that the journal tweeted out was of mortality. The separation started in 6 months and was robust at 12 months.

  • The analysis was funded by an investigator-initiated grant from Boston Scientific.

The conclusion which resides in Circulation is thus: “In a real-world population of older Medicare beneficiaries with AF, compared with anticoagulation, LAAO was associated with a reduction in the risk of death, stroke, and long-term bleeding among women and men. These findings should be incorporated into shared decision-making with patients considering strategies for reduction in AF-related stroke.”

Comments. I don’t know how to feel about this other than to be terribly sad for the scientific enterprise. You would think we make progress, not regress.

The problem here is that KM curve that the journal credulously tweets out undermines the entire paper. And I know that the authors know this. I know that the journal editors know this. I don’t know what the peer reviewers know, but that too is scary.

  • My friends, even if you believed LAAO provided benefit, it is in stroke and bleeding reduction over years. Plugging an appendage would be unlikely to ever reduce mortality but surely not in the first year.

  • What’s more, the stroke reduction in this paper was only 35%. How can that lead to a 50% reduction in death? It can’t. If you posit a procedure reduces death by 50%, you’d want 100% reduction of stroke. In this study, there were no differences in bleeding in women and it was higher in men.

  • In LAAO3, a real trial, with real randomization: Surgical appendage closure substantially reduced stroke rates over no closure but had no effect on mortality.

  • Why? because stroke is such a small contributor to death. It’s great to reduce stroke but reducing stroke will not reduce death.

The teaching point is that when you see an observational study with implausible separation of KM mortality curves, you know that healthier patients were in one group — here, the LAAO group.

What follows is not personal. I am sorry and I know it won’t increase my invites to meetings. But. The conclusion that this data should be incorporated into shared decision-making with patients is perhaps the worst thing I have read this year.

This data is obviously biased and implausible. The study should not have been done or published. The only way you should publish this data is as an example of what biased observational data looks like.

My gosh. Maybe I am too harsh, or emotional, if you think so, let me know. If you can find anything positive from this paper, let me know.


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.