Jun 07 2024 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 June 7, 2024 John Mandrola, MD, comments on the following news and features stories:  Cannabis and cardiovascular (CV) outcomes, post-cardiac surgery atrial fibrillation (AF), embolic protection devices, emulation of randomization, and an preview of heterogenous treatment effects.

First, I want to offer huge thank you to Professor Andrew Clark and the British Cardiovascular Society (BCS) for inviting me to give the Paul Wood lecture at the BCS annual meeting on Monday.

Paul Wood practiced cardiology in the 1940s to 1960s. In 1950, he wrote what was then the definitive textbook in cardiology. It was 1000 pages by one author!

He became famous in Britian for his bedside skills, strong thinking, and lack of fear of speaking candidly.

He died at age 55 in 1962 of an acute coronary occlusion. When he saw his ECG, he called it “irreversible” and told his caregivers that he should not be resuscitated. He had ventricular fibrillation later that day.

I started my lecture by saying Dr Wood had a different problem than we have now. Dr Wood’s problem was that he did not have enough to do for his patients. One of his papers recommended treating angina by telling the patient to cut down to 10 cigarettes per day from 20 to 25.

Our problem now is that we have tons to do for (or to) our patients. The question is whether we should do these things. And that’s where evidence appraisal comes in.

So, I told stories about evidence appraisal — including many of the major mistakes we made as a field. One theme was how we are misled by overconfidence in observational studies.

I’ve been honing this lecture for months. There are good stories to tell about using evidence. It’s why I like doing this podcast.

Cannabis for Chronic Pain and CV Safety

There was a tale of two observational studies recently. One strong. One weak.

Dr. Anders Holt and his team led by Morten Lamberts at the University of Copenhagen have published a nice study of CV safety for medical cannabis use as a pain reliever.

This is an important topic to study, for multiple reasons:

• Chronic pain is a serious problem, not least because the relief of pain comes with the potential for adverse effects. Non-steroidal anti-inflammatory drugs have known CV and renal complications. The dependence issue with opioids surely is an even more serious complication. The utter tragedy of our US opioid crisis is under-appreciated.

• Second, cannabis is pharmacologically active and has CV effects. And there is little empirical data on its use.

• Third, as always, you’d want to have proper randomized controlled trials (RCTs) to really know how cannabis stacks up against other pain relievers, but there are no RCTs, and there are not likely to be any in the near future.

Observational data is going to be all we have with cannabis for a long while. Many countries have legalized medical uses of cannabis, Denmark included.

A recent JAMA Network Open paper, first author, Lillian Gelberg reports that in the UCLA system, nearly one in five people report cannabis use. Of those, about one-third meet criteria for cannabis use disorder.

Cannabis use is a big deal, and worthy of study.

The Danish paper:

The DANISH registry allows for high quality observational data. Part of the universal Danish healthcare system involves a national registry, which captures prescriptions filled as well as outcomes in all Danes that use the universal healthcare system.

When I was in Copenhagen, Dr. Lambert showed me the registry, and a bit about how it works. It’s an amazing resource for asking questions about associations, finding adverse effects of therapy, or describing temporal trends. 

• Dr. Holt and his team identified patients (median age 59) with chronic pain who initiated a first-time treatment with medical cannabis between 2018 and 2021. They had about 5400 patients.

• They then matched these in a 1:5 ratio with 26,000 control patients who also had chronic pain. Matching also included age and sex.

• The two major outcomes of interest were arrhythmia of any sort and acute coronary syndrome (ACS).

• Diagnoses of pain included mostly musculoskeletal, but also cancer and neurologic pain. About one-third were unspecified.

• The 5400 patients who filled a cannabis prescription could have taken three different formulations — CBD alone, CBD/THC, or THC alone.

• The key finding was a doubling of the risk of any arrythmia in first time user’s vs controls. Absolute incidence was 0.8% vs 0.4%; risk ratio 2.07 (confidence interval [CI] 1.34 to 2.8).

• Assessing just new onset AF/flutter as an outcome yielded a similar results risk ratio of 2.04 (CI 1.19 to 2.89) comparing medical cannabis use with no use.

• The Kaplan Meier curves separated not immediately but over time, as you’d expect if cannabis was causal. The higher risk ratios were seen in all subgroups but did not reach statistical significance in all of them.

• But. The authors found no significant association with medical cannabis and ACS; risk ratio 1.20 (CI 0.35 to 2.04). The absolute risk of an ACS was quite low at 0.2% in both groups. They also found no association with stroke or heart failure.

Concomitant treatment with other pain meds did not change the doubling of risk. But when they restricted the exposure to only count patients as exposed if they filled a second prescription of cannabis, the risk ratio declined to 1.4 and was no longer significant.

Comments: This paper was originally published 3 months ago. I saw it because Holt and colleagues ran another analysis that the European Heart Journal (EHJ) published this month.

It’s a nice story – a writer wrote to EHJ to ask about what the analysis would look like if Holt and colleagues considered lifestyle factors such as smoking and alcohol use.

And guess what? Holt and colleagues obliged and ran more analyses. It turns out that patients exposed to medical cannabis were more likely to have a history of smoking than control patients whereas alcohol abuse rates were similar.

Adding the smoking and alcohol proxies as variables to the statistical models yielded comparable results compared with the main analyses. Likewise, if the association between medical cannabis use and cardiovascular risk was assessed in a subgroup free of smoking and alcohol history, results were comparable as well

What should we make of this?

• First, I believe it’s plausible that cannabis could increase the risk of AF. The active components have effects on cardiac electric properties.

•  It’s also possible that medical cannabis has no association with atherosclerotic cardiovascular disease events, at least acutely.

•  Positive effects on pain relief could balance any adverse effects on likelihood of vascular events. The editorialist points out that this observation differs from other observational studies that find higher rates of myocardial infarction (MI) with cannabis.

The analysis, of course, is observational. Unmeasured confounding is always possible, as the authors write. There’s also the issue of how cannabis is delivered; smoking would seem more hazardous than other ways.

The second paper:

There’s a second paper on cannabis to tell you about. It’s in JAMA Network Open, remarkably by only one person, Alexandre Vallee, an epidemiologist from France.

He used the UK Biobank to ask about cannabis use association with all-cause, CV death, and cancer mortality in males and females.

• It’s a simple association study. He divides cannabis use into high, moderate, and low categories in the t sexes and then looks at rates of death, CV death, and cancer death.

• He finds a positive association between CV death and heavy cannabis use among females.

• He spends many words describing the limitations of this analysis. It’s really weak.

•  There are adjustments but it seems obvious that self-reporting cannabis users would be different in many ways from non-users.

• He makes a case for reverse causation wherein the CV disease causes cannabis use. He writes that there was no information on the type of cannabis, or its delivery. He also included no falsification analyses.

Comments. I am surprised that such flawed analyses are published. You should read the Danish paper and this one and contrast the quality. I report on these observational analyses because there is so little data with cannabis. We can’t make definitive conclusions from any cannabis paper yet.

• The Holt paper from Denmark is stronger but it is still non-random and only looks at 180 days. Chronic pain therapies must be proven safe and effective over a lot more than half a year.

• But perhaps the lack of short-term ACS signal is mildly reassuring. If cannabis helps with chronic pain, and the only costs are a small absolute risk increase in AF and no elevated risk of MI, then that is a good thing, because patients with chronic pain have so few good options.

• The US population has just recently woken up to the dangers of opioids and boy did we get bamboozled by marketing. In recent years, I’ve become frightened about the dangers of opioids. If cannabis helps reduce opioid exposure and it proves reasonably safe, it will be a win.

But this is the first mile of a marathon in understanding cannabis as a therapeutic.

Post Cardiac Surgery AF

AF that occurs after cardiac surgery has to be one of the most vexing problems in cardiology. I’ve come to believe the only way to prevent AF after heart surgery is not doing heart surgery. Nothing seems to work for prevention.

And one major question that comes up often is what to do about stroke prevention. Should patients get oral anticoagulation (OAC)?

On the one hand, cardiac surgery patients often have many stroke risk factors. Their CHADSVASC scores are almost always high. Plus, they just had surgery, so thrombotic risk would seem to be high.

On the other hand, they just had surgery, so bleeding risk is elevated.

It would be an awesome question to study with randomization since there is equipoise. But I know of no such study. We have only observational data.

A group of authors, from the Netherlands, first author, Mileen van de Kar, have put together a huge meta-analysis of observational data on the effects of OAC for post-coronary artery bypass graft (CABG) AF.

• The study included 1.7 million patients in 28 studies.

• The overall mean incidence of AF was 24%; range 8% to 38%, which seems low, but perhaps is low because it’s just CABG and not valve surgery.

• They first summarized the incidence rates of stroke.

• They then compared stroke, death, and bleeding rates (per 100 patient years) for post operative AF (POAF) patients vs non POAF patients over 5 years.

  • POAF patients had higher rates of thromboembolic events (1.7 vs. 1.1 events), mortality (3.4 vs. 2.2 events), and bleeding (2.0 vs. 1.6 events) per 100 person-years.

  • Look at how low the rates of stroke are at 1 month and 1 year; 1% or less. And notice how little difference there is in stroke and bleeding over 5 years. There are clues there.

• Four studies allowed for pooling of data to compare outcomes using OAC vs not using.

• There was no association with lower stroke or mortality rates with OAC vs non OAC, but there was a 32% higher rate of bleeding in the OAC users.

Comments. This data tempts you to conclude that POAF is (a little) similar to subclinical AF in that it is associated with low risk of stroke and no benefit from OAC. I use the verb tempts because it’s still observational data. The way to answer this is with an RCT.

But this is all the data we have. So, we must use our brains. What can we say about this?

Here is how I see it. This is opinion level evidence. The weakest.

• First, I don’t think the CHADSVASC should apply. That score was not derived in post-bypass patients with AF.

• Second, we can be reasonably confident that OAC indeed increases the rate of bleeding.

• Third, we can also be confident that the signal of low stroke incidence after POAF is probably pretty close, because stroke within 30 days is likely not missed too often.

The problem comes in sorting out the lack of stroke benefit. Clinicians, not randomization, decide to use OAC so the two groups will likely differ in important ways and thus, it is hard to make much of the lack of signal of benefit. It could be a false negative (though I doubt it).

In the end, what I do when asked about this, is to understand that we have no evidence of benefit, but we know the risk of stroke is low (in general) and we know bleeding risk is high, so I lean slightly against use of OAC.

We need a proper RCT. Perhaps listeners know if one is ongoing or planned.

Embolic Protection After TAVI

It shocks me that stroke rate is so low after transcatheter aortic valve replacement (TAVR). These are older patients with thickened, calcified valves. At the risk of dumbing this down, the new device squishes and crushes the diseased valve. How in the world does this not spray debris everywhere?

If I did not know better, an embolic protection device (EPD) placed between the aorta and brain would be a necessity. But it’s not quite that obvious. Things in medicine are not often as they seem.

EPDs do indeed catch debris after TAVR. These make dramatic pictures, like coronary angiograms, but the PROTECTED TAVR trial failed to show a statistically significant reduction in all-cause stroke. Though there were wide confidence intervals, allowing for benefit, and there was a signal (a 62% reduction) that the device reduced disabling stroke — you know, the kind that would be prevented if the device caught large debris.

Circulation Outcomes has published a new observational study looking at use and non-use of EPD. First author Neel Butala.

Even though it’s observational, I like the paper. There are worthy things to say about the methods and the results.

Butala and colleagues used the Society of Thoracic Surgeons/American College of Cardiology/ Transcatheter Valve Therapy registry to compare outcomes between use of EPD and nonuse. The authors were interested in disabling stroke, but that’s not labeled in datasets, so they used in-hospital stroke resulting in death or discharge to a non-home location as a surrogate. That’s good thinking.

• They had data on more than 400,000 TAVRs. About 13% or 54,000 had an EPD and the rest did not.

• These are obviously not randomized comparisons, so the authors had to try and simulate the normal balancing of factors that randomization provides.

•  They did this in two ways. One was the usual propensity matching, wherein patients are matched using variables from the datasheet.

The other way is to use what’s called an instrumental variable analysis. I am going out on a limb here trying to explain instrumental variable analysis. Give me some leeway, please.

An instrumental variable analysis can be used to overcome confounding by indication. An instrument is something that is relevant to the outcome and determines the treatment but is not affected by the treatment.

Physicians’ preferences might be a good instrument, because when there is equipoise of a treatment, the choice of using or not using the treatment could be doctor’s choice, sort of like a fashion or practice pattern, if you will, and that can be quasi-random. And then any differences in outcome would be due to the treatment, not other factors, like the overall health of the patient.

The instrument they used in this case was site-level preference for EPD. Site-level EPD use would seem a good instrument because theoretically, it leads to quasi-random groups.

In the supplement, they have two columns of patients — those with hospital EPD use within the same quarter and those without hospital use of EPD within the same quarter. This was 168,000 in the EPD use group and 250,000 in the non-use within the same quarter group.

The main findings:

• The unadjusted rate of in-hospital disabling stroke was 0.7% in the EPD group vs 0.9% in the no EPD group.

• Now for the adjustments. In the instrumental variable analysis, the relative risk was 0.87, a 13% reduction in disabling stroke, CI 0.73-1.00, so not quite significant.

• In the propensity matched analysis, the hazard ratio (HR) was better At 0.79, a 21% reduction, with CI going from0 .70 to 0.90.

• As with PROTECTED TAVR, EPD use was not associated with non-disabling stroke.

• In subgroups, the benefit of EPD was much greater in those with prior stroke vs those without. For instance, in the instrumental variable analysis the EPD reduced disabling stroke by 35% vs only 6% in patients without prior stroke.

The authors concluded:

In the largest study to date, among patients undergoing TAVR, EPD use was associated with a small, borderline significant reduction in stroke associated with death or a discharge to a non-home location (a proxy for disabling stroke) that is likely to be causal in nature. Taken together with previous mechanistic and clinical studies, these findings provide credible evidence that EPDs benefit patients undergoing TAVR.

Comments. I like this analysis. While observational, it is a thorough analysis, with good effort to simulate randomization. Senior author David Cohen wrote to me on Twitter that “the instrumental variable analysis does not assume that hospital use of EPD is random. It assumes that patient distribution across hospitals is random.” That looks to be born out in the appendix in that the two groups are well matched.

The other thing I like is that it’s clear that the authors are trying to compare EPD use and non-use. They are candid about their attempt to make causal conclusions. They don’t disguise this randomization emulation attempt with mere associations. I think they get quite close to randomization. 

And they emulated the results of the PROTECTED TAVR RCT. But they also did something else. They showed that this device is not cost effective. Not at all.

• The tip off for me came in the middle of their discussion, when the authors noted that the nice-sounding 13% reduction in the instrumental variable-adjusted analysis translates to a number needed to treat (NNT) of 833. Even if the device cost very little, which it doesn’t, we’d have to treat 833 patients to prevent one disabling stroke.

• That was such a huge NNT, I went back to look at the absolute risk reduction, and indeed, it was 0.79% vs 0.91% in the EPD vs no EPD group. That is a delta of 0.12%.

• The authors spend the next paragraph telling us how big the next study would have to be based on these small differences. It would have to be very large. Like more than 100,000 patients large.

• They tell us that there is enough equipoise to continue the UK-based British Heart Foundation PROTECT TAVI trial, which has a target of just 7,700 patients, so, it is likely to be underpowered.

In conclusion, this is a tough dilemma, isn’t it?

The device probably has a tiny effect on disabling stroke. Real world data with careful adjustment replicates the findings in the RCT.  But how can we justify the expense when it’s that small of an effect? There has to be a line somewhere.

As I said at the outset, Paul Wood did not have enough to offer. We have plenty. But when the NNT is 800, well, that too is a problem.

The authors tell us to consider the smaller NNT in patients with prior stroke. But a) it’s still going to be pretty small, and b) this is a subgroup analysis of an observational study so it’s hard to put much weight on that.

A Preview

Finally, a preview of next week. Speaking of subgroups within studies, I saw on Twitter this week an incredibly interesting paper on treatment effect heterogeneity, first author Herbert Weisberg, published in the American Heart Journal.

The reason it’s important pertains to the EPD story I just told. Overall, the devices have little average effect. But proponents, and really me, too, would like to try and hone into patients who might benefit most. In other words, we seek evidence for and direction of a treatment effect heterogeneity.

The paper delves into ways to find such things. It’s the essence of medicine, because average effects from trials help us somewhat, but we often see patients who are not at all average or similar to those in trials.