Booze Out, Coffee Okay to Outsmart Atrial Fibrillation?

The I-STOP-AFib and CRAVE trials

Interviewer: Jagmeet P. Singh, MD, MSc; Interviewee: Gregory M. Marcus, MD, MAS


December 09, 2021

This transcript has been edited for clarity.

Jagmeet P. Singh, MD, MSc: I'm Jag Singh from Medscape Cardiology. It's a pleasure and a privilege to be here. It's a delight to have my colleague and friend, Dr Greg Marcus, from the University of California San Francisco School of Medicine. Greg is a professor of medicine and the associate chief of cardiovascular research. Welcome, Greg.

Gregory M. Marcus, MD, MAS: Thanks. Thanks for having me.

Singh: Greg, you presented a phenomenal late-breaker clinical trial at the AHA 2021 a day ago, and this trial was called the I-STOP-AFib trial. This looked at individualized studies of triggers of atrial fibrillation (AF) and its impact on AF episodes.

For those who were not able to be there for the presentation, I was wondering if you can give us an overview of the study design and the results. We can really dig into the details after that.

Marcus: Sure. The idea actually arose in very close collaboration with patients with AF who had told us that they were especially interested in understanding discrete triggers of acute AF episodes. We sought to really test the notion that we could look at n-of-1 trials within individuals. We therefore enrolled individuals from all over the US using a mobile app. The consent was all done remotely.

They were then randomly assigned to just have their AF tracked vs to undergo their own individualized n-of-1 trials. If one was randomized to those n-of-1 trials, they first were presented with a menu of potential triggers. We had developed that menu by serving up to 1000 AF patients where we just asked them, "What do you think triggers your atrial fibrillation?" They also had the opportunity to write in their own trigger if their trigger of interest wasn't represented on that menu.

They were randomly assigned 1 week at a time via daily texts to, "This is your week to go ahead and expose yourself to your trigger. You told us that, for example, two glasses of red wine triggered your AF." On other randomized weeks, they were instructed, "Make sure to avoid your trigger this week. Don't drink any alcohol this week."

After the end of 6 weeks, they were presented with their own individualized results where we showed them a calendar that illustrated "Here's where you were randomly assigned to test your trigger vs avoid it; here's where you told us whether you had AF or not." We asked every day, "Did you have any AF yesterday?” We also provided an AliveCor KardiaMobile device to help confirm whether there was AF or not.

They then had 4 weeks to digest that information and change their lifestyle if they wished. At the same time, we had the control arm doing the same thing, but just letting us know whether they had AF or not, just from monitoring. They also had the AliveCor KardiaMobile device.

At the end of that 10-week period, they were all given an opportunity to test another trigger if they wanted. Even those who were initially randomized just to tracking had an opportunity to test a trigger to receive their results.

The primary outcome was the AFEQT score. This is a validated survey instrument that reflects overall AF-related quality of life in the previous 4 weeks. We found no significant difference between those randomly assigned to do their n-of-1 testing vs tracking. However, when we looked at those last 4 weeks and the real-time number of AF episodes that were recorded, those randomized to the n-of-1 trigger of testing actually did exhibit statistically significantly fewer AF episodes.

As another part of the study, we wanted to dive into the actual n-of-1 test to see if we could identify particular triggers that really did influence the risk of a near-term AF episode. In those analyses, alcohol arose as the one consistent trigger, albeit only in the per-protocol analysis, not in the intention-to-treat. We asked them every day, "Did you actually expose yourself to your trigger?" When we looked at when they said they did, they were more likely to have AF essentially that day or the next day.

Of potential interest, caffeine was the most commonly selected trigger for testing, and yet that showed no relationship with atrial fibrillation.

Singh: You mentioned that this is open-label, patients select their own triggers, and they expose themself with their triggers. What is the potential for bias that could influence the outcomes of these studies?

Marcus: There are a few. One is that there could be some preconceived notion. Of course, the whole point of having random assignment is to try to circumvent that. There's still that possibility that there's some predetermined thinking that "I know that this is my trigger." Are there other things going on there that make that more likely the case?

One of the interesting ideas is that we've started to think about it in terms of the utility of the n-of-1 trial. The intuition is that it's useful for patients to identify what their triggers are so they can avoid them. The flip side is that they may be able to test something that they're not sure of that they might enjoy, such as drinking coffee. It may be reassuring, then, for them to learn, "Oh, it turns out this isn't associated with my AF, and so I can enjoy the coffee."

The other thing in terms of bias that's important to mention is that we didn't have perfect follow-up. Retention and engagement of our research participants is always a challenge. It's always an important consideration. Especially in these remote-based studies, it's really challenging.

Singh: Getting back to the topic of triggers, whenever I think of arrhythmias, I look at it as an abnormal substrate with a trigger and then an abnormal autonomic tone, which helps perpetuate the rhythm disturbance usually.

Here, you have an n-of-1 trial, so you have that same patient with that substrate, you have the same trigger. We always say that if you have an arrhythmia, the stars really need to line themselves up. You need to have that same sort of balance or the level of angst or the level of emotional turmoil at that point in time for everything to fall into place. That often doesn't happen when you're testing a trigger, and the absence of a trigger initiating AF doesn't necessarily rule it out as a potential trigger in different circumstances.

How would you relay the results of a population-based study like this at the individual level? What would you advise your patients?

Marcus: Yeah, that's absolutely right. We don't think that these substances are both necessary and sufficient to cause AF, but rather all of us have a propensity to AF and there's a broad spectrum. Some people are never going to get AF no matter what they're exposed to. Some people are going to get AF no matter what they avoid. What we're trying to find is a bit of a sweet spot and what's idiosyncratic to a given patient that might help them with their AF.

It's probably worth mentioning a little bit about alcohol specifically, given the wealth of data that we now have related to alcohol in AF. We recently published a study demonstrating, among people wearing continuous ECG monitors and an alcohol sensor, that acute alcohol consumption really did predict an AF episode a few hours later.

There's now a randomized trial from Alex Voskoboinik and Peter Kistler demonstrating that cessation of alcohol can reduce AF burden, so there really is very robust evidence of a causal relationship there.

We did a randomized trial where we delivered intravenous alcohol vs placebo. We showed that, acutely, pulmonary vein electrophysiology changes in ways we would expect to be important. That, in conjunction with these data, suggests that while there almost certainly is much heterogeneity among individuals, if there's any substance that can be attributed to a higher risk of near-term AF, it's alcohol.

I tend to tell my patients, "Look, if you want to do everything you can to avoid AF, reducing or eliminating alcohol is probably the right thing to do." There are always going to be some patients for whom alcohol is just important to their quality of life. Of course, I always strongly recommend avoiding excessive alcohol. Whether consuming one drink a day or a glass of wine a day is better or worse, we still need to figure that out.

When it comes to the other potential triggers and how to counsel our individual patients, or think about this on a general population level, I think it's reasonable to suggest that it's okay to experiment, such as with coffee or caffeine, where there isn't clear, consistent evidence of harm, where quality of life is an important issue, where trying to understand one's individual level triggers for their AF may be important.

I always reassure my AF patients that an episode of AF is not imminently life threatening. We want to think about stroke prevention. That's why we prescribe anticoagulants as needed, but that's really more of a chronic thing. The reason to avoid discrete AF episodes is primarily to help with quality of life. In the long run, it's not unreasonable to experiment. I think that heterogeneity is somewhat the name of the game, which makes it hard to extrapolate from the general population to our individual patients with AF.

Singh: Exactly. Just picking up on the word "heterogeneity," I think there is much heterogeneity in what kind of alcohol also causes AF. I have patients who say that every time they drink red wine, they've got into AF. If they switch to beer, they still consume alcohol but they don't have AF.

How do you resolve those differences? Is it the alcoholic content? Is it the tannins or something else that is contributing to it?

Marcus: Considering our randomized controlled trial data from our intravenous alcohol study, we titrated everyone to a breath alcohol concentration of 0.08% and found this relationship. My sense is that it is the alcohol that still could be mediated by autonomic changes, but it's still true. Whether there are other constituents that may be important, I don't know. There might be.

There are some patients who also describe that simply having a cold drink, regardless of what's in that drink, contributes to AF. Perhaps it is activation of the esophagus, which we know lies right next to the left atrium. That's why transesophageal echocardiograms are so useful to rule out left atrial thrombi. Again, I suspect it's also multifactorial.

As I often tell my patients that AF is almost certainly a common final pathway of multiple different mechanisms. We know that different families with Mendelian inheritance of AF can exhibit very different mutations. You can have a mutation in ANP, a sodium channel, or a potassium channel. It could be a depolarization problem, or it could be repolarization. Not all AF is the same. That also fits with this idea of heterogeneity and idiosyncratic relationships that our patients are telling us about.

Singh: That's terrific. Greg, I know we're running short of time. You had a stellar AHA conference. You didn't have one late-breaker; you had two late-breakers! The other late-breaker was the CRAVE study, correct?

Marcus: That's correct.

Singh: That's looking directly at coffee. Can you provide us a very quick overview for our listeners so they can get a two-for-one out of you?

Marcus: I'm very happy to. CRAVE was a lot of fun, frankly, and I think our participants actually enjoyed it. These were 100 healthy volunteers that we fit with a Zio patch to record their ECG continuously, Fitbits to record their step counts and sleep, and a continuous glucose monitor. They were randomly assigned also in an n-of-1 sort of paradigm, day-to-day, to avoid vs consume coffee.

We assessed adherence to the randomization assignment in several ways. We asked them to push the button on the Zio patch whenever they took a drink. We asked them the next day, "Did you actually have any coffee yesterday, whether you were supposed to or not?" They didn't really have any incentive to be dishonest. We reimburse them for all coffee consumed as long as they gave us a date-stamped receipt. Again, we said, "Look, even if you weren't supposed to, we'll still pay for it," so that we could record that. Number four, we tracked their geolocation and actual visits to coffee shops.

It turned out that while compliance wasn't perfect, it was generally very good. It was much more common that they followed their randomization assignment than not.

Skipping to the results, there are several. We found no relationships with premature atrial contractions (PACs). Actually, in per-protocol analysis, less supraventricular tachycardia (SVT) with more coffee consumed. This fits with some recent data. We have previously shown in a community-based cohort that PACs are the single most important predictor of AF. In the UK Biobank, we found that those who reported consuming more coffee exhibited a lower risk for AF and SVT, which fits with these data.

In contrast, based on intention-to-treat, days randomly assigned to coffee were associated with, on average, 50% more PVCs, which was highly statistically significant. Per protocol, the more coffee consumed, the more PVCs the people had.

Looking at the step count, based on intention-to-treat, on days randomly assigned to coffee, people took, on average, about 1000 more steps on those days. We know from other studies that a difference of 1000 steps is highly clinically relevant. In fact, there's some evidence that, if you compare those with a major disease vs not, the differences can be about 1000 steps.

At the same time, days randomly assigned to coffee were associated with significantly less sleep in the subsequent evening — about a half-hour less sleep — also highly statistically significant and clinically relevant.

We did also genotype these individuals for their expected caffeine metabolism. There, one of the most telling findings was that it was the slow metabolizers who experienced the worst sleep deprivation when exposed to coffee, and the fast metabolizers exhibited no reduction in sleep when exposed to coffee.

Finally, we didn't find any evidence of any relationships with glucose.

Singh: Spectacular results. If I participated, you wouldn't have to geofence me. I have a coffee machine in my office. I want to thank you once again, Greg. Two remarkable studies and certainly, I think, very enlightening and practice-changing. Thank you, again, for being with us. Until next time.

Marcus: Thank you so much.

Follow Medscape on Facebook, Twitter, Instagram, and YouTube


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.