Time Stamps

  • 00:00 Challenges of interpreting nutritional research
  • 02:22 Best practices for evaluating studies in nutrition
  • 12:35 Delve into the CRAVE trial as an example of critically appraising nutritional investigations
  • 26:41 Applying this to clinical practice for your patients

Sponsor: Use code CORE25 for 25% off on ACP MKSAP Audio Companion!

https://www.coreimpodcast.com/MKSAP

Show Notes

Why Should I Care? 

  • Interpreting nutritional research well = Crucial for promoting health for your patients!
    • Patients (and colleagues) will ask you about the latest studies they saw on the news or other media (such as social media). 
      • PRO TIP:  Have tools to evaluate the original studies ,while also noting the challenges of interpreting this research. 
        • Especially because nutritional guidelines often change!

Nutritional Research: Types of Study Designs

  • Randomized control trials (RCTs)
    • The gold standard but not always feasible for nutritional research. 
      • Limitations:
        • TIME: Requires long follow-up for certain outcomes of interest, such as cardiovascular disease or cancer.
        • ADHERENCE: It can be difficult to maintain adherence with assigned diets over time in a trial 
          • Tracking adherence is difficult!
        • It is not possible to test ONE variable of interest. 
          • If you tell someone to eat more of one type of food, then they are eating less of something else or more calories overall; you can’t do the typical placebo versus drug comparison. 
  • Epidemiologic studies (Case-control, Cohort Studies)
    • Very commonly used in nutritional research! 
    • BENEFIT: Can capture dietary patterns through 24-hour dietary recall and food frequency questionnaires.
      • Example of trans fatty acid intake and heart disease: 
        • Controlled feeding studies of about 50 people 
          • Looked at short-term changes
        • Result: Diet high in trans fat → 
          • Elevated LDL cholesterol 
          • Reduced HDL cholesterol
          • Elevated blood triglycerides 
        • Clinical Application: From other epidemiologic studies, this pattern of lipid changes is associated with increased risk of heart disease
          • Suggests that trans fats = Risk factor for heart disease 
          • These findings have been replicated in additional studies
    • Consider whether your study is prospective, large, long-term.
        • Also look at the effect size.
      • If NOT long-term: combination of long-term epidemiologic studies with shorter term studies with intermediate endpoints can be helpful. 

What are the challenges of interpreting nutritional research?

  • Recall Bias: Limitations with accuracy 
      • Food frequency questionnaires 
      • 24-hour dietary recall
    • Once people have a disease, they can “misremember” their previous diets. 
    • Can try to minimize this type of bias by collecting information before people are diagnosed with disease.
  • Forms of bias
    • Healthy user effect
      • When a specific food or dietary pattern has a connotation of being “good for you,” health conscious participants may be more likely to consume it.
      • People who sign up for nutritional research may be more health-conscious. 
        • They may be more likely to participate in other healthy behaviors. 
    • Confounders: 
      • Health care use 
      • Socioeconomic status
    • Crossover effects
      • Multiple Risk Factor Intervention Trial, or MRFIT (MRFIT JAMA 1982):
        • Multicentered RCT that looked at a multi-factor intervention which included: 
          • Question: Effect of replacing saturated fat with polyunsaturated fat on coronary heart disease mortality. 
          • Intervention: No effect! 
            • But the people in the control group had already changed their diet on their own after reading the news.
  • Translation of results:
    • We cannot determine which aspect of food is being evaluated!
      • There are many factors involved in a food study
        • Example: Coffee has caffeine, antioxidants, polyphenols, and other bioactive compounds. Studying coffee cannot be extrapolated to conclude specifics about any individual ingredient. 

What additional features strengthen studies?

  • Reproducibility
    • First study might be the most unreliable study 
      • Have the findings been reproduced by other investigators and using additional approaches? 
  • Consistency and intermediate end-points
    • Do we have some studies with intermediate endpoints that show the same findings? 
  • Biologic plausibility
    • Sometimes animal studies can help, but they can be removed from human biology so may not be as applicable to studies with humans. 
  • Effect size
  • Validation
    • Have the methods for assessment been validated? 

CRAVE and DECAF: Example of Randomized Trials in Nutritional Research 

  • Background studies
    • Observational studies on coffee and Afib
      • Many have shown NO increased risk
        • Large cohort studies and meta-analyses (Caldeira et al. Heart 2013; Larsson et al. BMC Med 2015). 
        • Some have even suggested lower AFib risk among certain coffee drinkers (Bodar et al. J Am Heart Assoc 2019).
      • Observational study on coffee consumption and mortality 
        • Inverse association between coffee consumption and mortality (Freedman et al. NEJM 2012).
  • CRAVE (NEJM 2023): Coffee and Real-Time Atrial and Ventricular Ectopy
    • Study design: 
      • Prospective, randomized, single-center (UCSF), 14-day trial with a case-crossover design
    • Population:
      • May 23, 2019 – March 25, 2020
      • 18 years or older 
      • Inclusion criteria
        • Have a smartphone
        • Able to use the Eureka mobile application
        • Willing to provide  saliva sample for genetic processing
        • Drink coffee or coffee-based products at least once a year
        • Willing to abstain from coffee, caffeinated products, or minimally caffeinated products (decaffeinated coffee) for at least 2 days when instructed
      • Exclusion criteria
        • History of AFib, heart failure
        • ICD or pacemaker
        • Treatment with beta blockers, non-dihydropyradine calcium channel blockers, or class 1 or 3 antiarrhythmic medications
        • Medical reason to avoid coffee
    • Randomization: 
      • Over a 2-day period, randomly assigned (via daily text messages over 14-day period) to either 
        • 1) Consume caffeinated coffee, or
        • 2) Avoid caffeine (including caffeinated coffee, decaf coffee, chocolate, tea) 
      • Participants received their assignments at 8 p.m. for the next day and received a follow-up reminder on the assignment day at 8 a.m.
      • To avoid cumulative effects and enhance enrollment and retention, the randomization was done in pairs of off–on or on–off days to ensure that a participant had no more than 2 consecutive days of coffee consumption or abstinence.
      • How was adherence monitored? Multiple ways!
        • Participants press a button on the ECG patch when consuming a standard 8-ounce cup of coffee or two standard 1-ounce shots of espresso
        • Asked each morning about coffee consumption on day prior
        • Offered reimbursement for coffee drinks purchased for immediate consumption if provided date-stamped receipts, 
          • Regardless of how those dates corresponded to their assigned consumption
        • If at baseline visited particular coffee shops, used the Eureka app (developed and maintained by investigators at the UCSF) on smartphones to continuously monitor geolocation to track visits to coffee shops
    • Primary outcome: 
      • Mean number of daily PACs
        • Continuously recording ECG patch (Zio XT Patch, iRhythm), a clinical-grade monitor
        • Arrhythmias confirmed by manual review 
          • By a blinded board-certified cardiac electrophysiologist
    • Secondary outcomes: 
      • Daily number of PVCs
      • Daily number of episodes of non-sustained SVT or VT
      • Daily step counts
        • Measured by Bluetooth-enabled, wrist-worn accelerometer to quantify step counts
        •  
      • Daily minutes of sleep
        • Inferred from actigraphy measured with the use of Fitbit INSPIRE devices
        •  
      • Mean daily serum glucose levels
        • Continuously recording glucose monitoring device
    • Results: 
      • Population
        • 100 participants 
        • Mean (±SD) age of the participants was 39±13 years
        • 51% were women
        • 51% were non-Hispanic White
      • Adherence monitoring
        • Majority of participants followed the instructions to consume caffeinated coffee or avoid caffeine on the majority of days 
        •  
      • Primary outcome
        • 58 daily PACs vs. 53 daily PACs; no difference (caffeinated coffee vs. no caffeine) 
          • Rate ratio, 1.09; 95% confidence interval [CI], 0.98 to 1.20; P = 0.10
        •  
      • Secondary outcomes
        • Caffeinated coffee VS. No caffeine 
          • Increased daily number of PVCs
            • 154 vs 102 daily PVCs, respectively (rate ratio, 1.51; 95% CI, 1.18 to 1.94)
          • NO difference in daily number of episodes of non-sustained SVT or VT
            • Non-sustained SVT: 
              • Mean, 0.17 episodes vs. 0.20 episodes, respectively (rate ratio, 0.83; 95% CI, 0.68 to 1.02) 
            • Non-sustained VT: 
              • Mean, 0.01 episodes vs. 0.01 episodes, respectively (rate ratio, 1.14; 95% CI, 0.43 to 2.99)
          • Increased daily step counts
            • 10,646 vs. 9,665 daily steps, respectively (mean difference, 1058; 95% CI, 441 to 1675)
            • Each additional coffee drink was associated with 587 more steps per day (95% CI 355-820)
          • Decreased daily minutes of sleep
            • 397 vs. 432 minutes of nightly sleep, respectively (mean difference, 36; 95% CI, 25 to 47)
              • Every additional coffee drink was associated with 14 minutes less sleep per night (95% CI 10-18)
          • NO difference in mean daily serum glucose levels
            • 95 mg/dL vs. 96 mg/dL (mean difference, −0.41; 95% CI, −5.42 to 4.60).
    • Clinical impact: 
      • Pros
        • Clear intervention 
        • Adherence well tracked
        • Objective outcomes
      • Limitations
        • Done in a healthy population
          • Cannot apply it to patients with heart disease 
        • Short-term study 
          • Does not study long-term effects of coffee and cardiovascular risk
          •  
        • There may be limited clinical relevance of measuring PACs in relation to AFib. 
  • DECAF (JAMA 2025): Caffeinated Coffee Consumption or Abstinence to Reduce Atrial Fibrillation
    • Study design: prospective, open-label, randomized clinical trial
    • Population: 
      • 200 current or previous (within past 5 years) coffee-drinking adults with 
        • Persistent AF
        • Atrial flutter with a history of AF 
        • Planned for electrical cardioversion 
      • 5 hospitals in the US, Canada, and Australia 
      • November 2021 through December 2024
    • Randomization: 
      • Randomized in a 1:1 ratio to: 
        • 1) Regular caffeinated coffee consumption
          • Encouraged to drink at least 1 cup of caffeinated coffee daily
        • 2) Coffee and caffeine abstinence for 6 months 
          • Encouraged to completely abstain from both caffeinated and decaffeinated coffee and other caffeine-containing products
  • Primary outcome: 
      • Clinically detected recurrence of AF or atrial flutter over 6 months
  • Results: 
      • Population: 
        • 200 patients
          • Mean 69 years (SD 11 years) 
          • 71% male
          • Baseline coffee intake was 7 cups (IQR, 7-18) per week in both groups
      • Primary outcome: 
        • AFib or atrial flutter recurrence was LESS in the coffee consumption 
          • 47% vs. 64% recurrence, coffee vs. abstinence
            • 39% lower hazard of recurrence (hazard ratio, 0.61 [95% CI, 0.42-0.89]; P = .01)
        • No significant difference in adverse events
    • Clinical impact: 
      • Pros: 
        • Done in population with disease (in contrast to CRAVE) 
          • May have more relevance to that population
      • Limitations: 
        • Clinically detected primary endpoint
          • Does not detect subclinical AFib
        • Only 40% of the patients reported that coffee ever triggered their AFib in the first place

Applying This for Your Clinical Practice

  • Remember to ask your patients about nutrition and diet! 
    • What do you say to these questions? 
      • Your patient with AFib asks if they can drink coffee!  
        • If they like coffee, it’s probably not harmful and they can probably continue to enjoy it. 
        • BUT, they don’t have to start drinking coffee if they aren’t currently AND they don’t have to drink it if it creates symptoms.  
      • Your young, healthy patients ask if they can drink coffee! 
        • Coffee is probably fine with respect to irregular heartbeats.
        • But it might cause a few more PVCs and might impact their sleep. 
    • What else should you consider when counseling patients? 
      • Socioeconomic barriers = Impact access to healthy food 
        • Cost 
        • Availability 
  • Put YOUR thoughts in the comments!:
    • What have been your experiences with patients (or friends) asking you about nutrition? 
      • How do YOU counsel them? 
    • What are some nutritional studies you’ve heard about recently in journals (or the media) you are wondering about? 

Transcript

Challenges of interpreting nutritional research

Dr. Walter Willett: I think almost everyone picking up the newspaper, looking online, especially online, is likely to be very confused about how diet affects health or anything else because you’ll find for almost every food that if you look on the internet, it’s either going to kill you or make you almost immortal, and everything in between. 

Dr. Shreya Trivedi: Yes, the headlines, the social media reels, feel like they’re all over the place, and the topic of nutrition research feels quite pressing. And so we’ve got a great episode lined up featuring the CRAVE trial on coffee and arrhythmias published in the New England Journal of Medicine (NEJM) in March 2023. 

Dr. Greg Katz: And with that, we’d like to welcome you to another episode of Beyond Journal Club, a collaboration between CORE IM and NEJM Group. 

Dr. Clem Lee: The goal of Beyond Journal Club is to take landmark clinical trials and to put them into context, telling the story of how we got to where we are and what it means for how we take care of our patients. 

Dr. Shreya Trivedi: I’m Dr. Shreya Trivedi, an internist at BIDMC.

Dr. Clem Lee: I’m Dr. Clem Lee, a Med-Peds hospitalist within the Mass General Brigham system and a Deputy Editor for NEJM Clinician. 

Dr. Greg Katz: I’m Dr. Greg Katz, cardiologist at NYU. 

Dr. Katerina Lin: And I’m Dr. Katerina Lin, NEJM Editorial Fellow. 

Dr. Greg Katz: So, I’m so excited that we get to delve into nutritional research today because this is what patients ask us about all the time. I mean, I cannot tell you how many people come to see me who are confused about what to eat or drink. Is red wine good or bad? What about meat? What about eggs? What about coffee? 

Dr. Katerina Lin: And we’re so fortunate to have Dr. Walter Willett here with us to delve more deeply into nutritional research.

Dr. Walter Willett: I was told when I was a doctoral student that probably diet’s important, but it’s just too complicated. But I sort of like complicated things.

Dr. Katerina Lin: Dr. Walter Willett is Professor of Epidemiology and Nutrition at the Harvard TH Chan School of Public Health. 

Dr. Greg Katz: And he’s been doing nutrition research for decades, and he’s one of the most cited people in literally just all of science, not just nutrition.

Dr. Shreya Trivedi: Yeah, a pretty big deal. So today we’re going to start by getting into why it’s so hard to ask the big questions in nutrition. 

Dr. Clem Lee: And then we’ll discuss what we can glean and what we have to look out for in both large RCTs versus observational trials.

Dr. Greg Katz: And finally, we’ll get into the CRAVE trial to help answer the question that our patients care really about: Can I have a cup of coffee in the morning? Or is it going to give me irregular heartbeats?

Best practices for evaluating studies in nutrition.

Dr. Shreya Trivedi: Alright, yeah, I don’t know about you guys, but like I was saying earlier, with all these health influencers, right, they say things with such certainty sometimes, that even with my own education and degrees that I have, I can sometimes even get persuaded. And this is all saying that I also know, at the same time, nutrition research cannot be simplified into these 60-second clips. And so let’s get into nutritional research and why it’s so hard to study, and oftentimes, way too complicated to draw quick, zinger conclusions from. 

Dr. Walter Willett: Well, in the perfect, ideal world, we would just do randomized trials, but for practicality, they’re very often (in fact, usually) going to be not possible to do. For example, for red meat consumption, probably you really need to be on a bad diet for three or four decades before you actually have a myocardial infarction. And we know that because you don’t have those diseases before age 30 or 40, usually. And so again, if you’re doing a randomized trial, that’s not going to be feasible to randomize many thousands of people at birth and follow them.

Dr. Shreya Trivedi: So the effects of diet on outcomes that we care about–whether it be cardiovascular disease, cancer, mortality–take decades to show up. And most people, we can barely make a change in our diet consistently for a week, let alone 30 years.

Dr. Greg Katz: And even a randomized controlled trial that has perfect adherence in nutrition has its own set of problems. Because it’s really hard to test just one thing. If you tell somebody to eat more of one thing, that means they’re eating less of something else, or they’re eating more calories overall. And so ideally, we would love to change one variable at a time and see the effect, but that’s impossible in nutrition. And so you can’t do the typical placebo versus drug that you would do in other types of studies. 

Dr. Katerina Lin: And then when these trials are run, we need to consider whether people are actually following their assigned interventions. Take, for example, the MRFIT (Multiple Risk Factor Intervention Trial) trial from the 1980s. It was a RCT that included replacing saturated fat with polyunsaturated fat (MRFIT JAMA 1982). 

Dr. Walter Willett: And in the end, there was no effect. The people who joined mostly had already changed their diet because they had read the news and had made the changes already. And then as the trial went on there, the intervention group did reduce their saturated fat intake, but the control group did it right in parallel. So in the end, there was really never any meaningful difference between the intervention and the control group. This is hard business. So the best alternative will usually be long-term epidemiologic studies like we’re doing, plus combining that information with the results from shorter term studies with intermediate endpoints.

Dr. Clem Lee: Yeah, it’s good to hear from Dr. Willett why we often must resort to observational studies. And we reviewed the Bradford-Hill criteria on a previous episode on microplastics. These are criteria to help identify strong associations in observational studies that can point towards causation, though we can never really say that the Bradford-Hill criteria prove causation.

Dr. Shreya Trivedi: Yeah, and with all that in mind, let’s get into when we look at a nutritional study, how we can look at it with a critical eye and really build up that chain of causality with some level of confidence?

Dr. Walter Willett: For the average person or physician looking at that, I think there’s several things to consider. First of all, usually the first study is the most unreliable study. It’s the replicated finding that is going to be what you, I think, trust more in terms of making any decisions, that it’s been reproduced by two or three other groups, or sometimes the same investigators, using additional approaches. So reproducibility is really important—not just the most recent finding. And then also how does it fit with other evidence? Do we have some studies that show, for example, in controlled feeding studies that an intermediate endpoint points in the same direction? Sometimes animal models, consistent animal models, can help, but usually that’s so far from human biology that those are not going to be so reliable. So mostly, I think, consider first the study itself. Is it prospective, is it large, long-term? And if you can, have the methods for assessment been validated (although that may not be in the story). But then how does it fit with other information, including has it been reproduced by other investigators?

Dr. Shreya Trivedi: Nice, so I appreciate hearing that thought process. Like when you see a new headline in the world of nutrition or a new research study. And, you know, he’s mentioned a bunch of times, these “short-term intermediate endpoint” studies, and I just want to clarify what exactly he meant by these. 

Dr. Walter Willett: I can use an example: trans fatty acid intake and heart disease. People were doing some short-term studies looking at changes in blood lipids as the outcome. These were studies, controlled feeding studies, of about 50 people and putting them on diets for about three weeks and then looking at short-term changes. And those short-term studies did show that trans fat elevated LDL cholesterol, reduced HDL cholesterol, and elevated blood triglycerides. And we know that that pattern of lipid changes is associated with increased risk of heart disease. So putting those two kinds of studies together (and each of those were replicated over time), that gave us a really strong foundation to conclude that trans fats were an important risk factor for heart disease.

Dr. Katerina Lin: So we’re using short-term intermediate endpoints in addition to the larger epidemiological studies. And for many of these studies, we’re recording diet with methods like the 24-hour dietary recall or food frequency questionnaires. 

Dr. Clem Lee: Yeah, and a lot of these food frequency questionnaires, there’s a lot of bias that comes with them. So let’s listen to how Dr. Willett tries to minimize the bias for these recall methods. And then, perhaps, we can apply a more rigorous lens to nutritional studies. 

Dr. Walter Willett: In our studies, what’s really important is that we are minimizing bias, because they’re prospective studies. We’re collecting the information before people are diagnosed with disease. Once they’ve got a disease, then we’re very likely to get biased information: they are thinking about their diet differently. So the important point is that we’re getting the information in a way that’s unbiased with respect to disease. 

Dr. Shreya Trivedi: Yeah, so prospectively, before someone has the disease in question can help, but there are also other steps that Dr. Willett takes, specifically when it comes to these self-reported questionnaires and recalls. 

Dr. Walter Willett: And then we’ve done from the very beginning a series of what we call validation or calibration studies where we take (actually now our most recent one) about 1300 people who were already in our study. And then we’ve gone back to them and collected very detailed data using weighed diet records, recording everything that they eat over two one-week periods over a year. We have lots of biomarkers from blood measurements, urine measurements, doubly labeled water measurements, which gets at energy intake. And so we can compare our simple questionnaire provided by several hundred thousand people with this very detailed measurement, and then we can actually do statistical corrections to adjust for measurement error. But what’s really very important is that we have repeated these measurements every four years. Very few, in fact, I don’t think any other studies have actually done that before, and that does turn out to be really important because people’s preferences change over time. Manufacturing changes over time.

Dr. Shreya Trivedi: So the next time I look at a nutritional study, I know after hearing the steps that he takes, I’ll also try to see: did these investigators, you know, take other measures to actually verify the self-reported data collection?

Dr. Clem Lee: Okay, it’s time for me to be the naysayer in the room again. I hate doing this, but I just have to bring up: even though we now have some extra tools from Dr. Willett to help us make the data more trustworthy, I think there are still a lot of caveats with nutritional data we need to discuss. 

Dr. Greg Katz: So for most of these observational studies, the data is collected from a food frequency questionnaire. And we need to keep in mind that even really well validated data collected over a short period of time, and then it’s extrapolated to all of these years. And we just need to be honest that even if you’re measuring people for 2 weeks, and you’re tracking every morsel that goes into their mouths, you don’t know what’s happening for all of the other years in their lives. I mean, most people’s diets change from day to day or week to week or month to month. And we’re not following people around for decades, figuring out what they’re putting in their mouths.

Dr. Katerina Lin: Right, Greg, and another thing to keep in mind is that people signing up for these nutritional studies might be more likely to do other healthy behaviors, meaning the healthy user effect.

Dr. Clem Lee: Yeah, and we also need to think about some very common confounders and to make sure that they were controlled for. These confounders might include health care use and socioeconomic status.

Dr. Greg Katz: And all of that rigor is really helpful. But we’re still left building a chain of evidence from imperfect data sources. And even if we’re talking about the same food, we might be talking about different things. And so if you tell me you eat ground beef once a week, 80/20 ground beef is a very different product than 93/7 ground beef. And so the level of complexity with all of this stuff is just profound, and sometimes it makes it hard to know if we’re even assessing the same thing. 

Dr. Clem Lee: Yeah, and something else to bring up, is that we have evidence linking entire diets to lipids, blood pressure, and cardiovascular events. But that still leaves a lot of uncertainty when patients ask about specific foods, not diets. So we can study the Mediterranean diet, but then people have specific questions about eggs.

Dr. Greg Katz: But even if you’re talking about eggs, sometimes that raises more questions than it answers. And so with eggs, what’s the bioactive compound? Is it the entirety of the whole food matrix? Is it a specific preparation? Is it the dose? Is it the timing? And coffee is another great example. It’s so commonly consumed, but coffee has a million things in it, and it’s really hard to study. 

Dr. Shreya P. Trivedi: Yeah, exactly. Coffee isn’t just caffeine, right? It’s the antioxidants, the polyphenols, the other bioactive compounds. So, yeah, it is a complex question when we ask, “Is coffee good or bad?” That is a total oversimplification. 

Delve into the CRAVE trial as an example of critically appraising nutritional investigations.

Dr. Shreya Trivedi: Alright, so after all the discussion about nutrition research and the complexities of it, let’s make it a bit more concrete and do some application. Let’s talk about coffee. And that oversimplified question: is it good or bad?

Dr. Greg Katz: If I had a dollar for every time a patient asked me if coffee was good or bad, I think that swim lessons for my kids would feel so much more affordable. Like I literally have patients asking me every single day of my life whether coffee is bad or good for them, if they’re allowed to have a cup. Every single day. And the concern really makes intuitive sense. I mean, caffeine is a stimulant. It affects autonomic tone, calcium handling, catecholamines. And so for decades, so many of my patients, their whole lives, they’ve been told that coffee might provoke irregular heartbeats.

Dr. Clem Lee: Okay, but I also have to admit, whenever I drink coffee, Greg, I also get palpitations. So I really do sympathize with your patients.

Dr. Katerina Lin: And a lot of people will still drink coffee, even if it gives all the symptoms. 

Dr. Clem Lee: Yeah, like me. 

Dr. Katerina Lin: Well, let’s look at the evidence. So it’s been really popular, a long-studied topic. Many observational studies haven’t shown increased risk between coffee and AFib, including large cohorts and meta-analyses (Caldeira et al. Heart 2013; Larsson et al. BMC Med 2015). And surprisingly, some even suggest lower AFib risk among certain coffee drinkers (Bodar et al. J Am Heart Assoc 2019).

Dr. Shreya Trivedi: Yeah, and speaking of the overall benefit of coffee, I just want to put it out there, this one big observational study that was actually published in the NEJM back in 2012. And it showed that those who drank more coffee actually tended to have a lower risk of death (Freedman et al. NEJM 2012). 

Dr. Greg Katz: And you can read a lot about coffee and all of coffee’s effects on health. And, you know, sometimes I feel like you can take any of the individual components and find a mechanism. And then you pick catecholamines and you say it increases sympathetic stress. You pick antioxidants and you say it helps with inflammation. 

Dr. Clem Lee: Right, as Shreya mentioned before, there are so many ingredients: polyphenols, antioxidants, phytochemicals. And so some of these might have anti-inflammatory or cardioprotective effects. So even mechanistically, I don’t think it’s obvious which ingredient explains any purported benefit of coffee.

Dr. Greg Katz: And so if you want to be an influencer, basically pick a random molecule in coffee, find a plausible biologic mechanism and then the epidemiology you like, and then post about it on social media, and then boom, it goes viral, you have 5000 comments on Instagram, including people fighting with each other in the comment thread. 

Dr. Shreya Trivedi: Oh, people, people. All of it. Yeah, so I think we can all agree that observational studies, especially the diverse group of chemicals that actually make up coffee, those observational studies aren’t perfect. But the CRAVE trial was, you know, really asking a more targeted question. And this question was “Does short-term coffee consumption lead to arrhythmias?” This is a much easier question to answer because we can have more control over short-term coffee consumption, right, even if we can’t control all of the specific ingredients in said coffee.

Dr. Katerina Lin: Right, and so investigators of this RCT had a more focused goal in mind: What actually happens to heart rhythms on days when people drink coffee versus when they don’t?

Dr. Clem Lee: Some of you like me might be wondering what the acronym CRAVE stands for. And so I’m here to tell you that it stands for Coffee and Real-Time Atrial and Ventricular Ectopy, which I think described the trial succinctly. So I give this 10 out of 10! Premature atrial contractions were the primary outcome, but ventricular ectopy was also measured as a secondary outcome.

Dr. Greg Katz: Clem, the investigators thank you for your compliments on their acronym. And so after the trial was done, the media headline basically simplified this to “Caffeinated coffee is safe—there’s no increase in premature atrial contractions.” But that headline (it’s true), but it hides a lot of nuance.

Dr. Clem Lee: Yeah, so, let’s get into it! Who were the participants in this trial, Katerina?

Dr. Katerina Lin: So there were 100 healthy people. And they were on average 39 years old. Most were non-Hispanic White, and the median BMI was 24. And very few had chronic health conditions like diabetes or hypertension. And at baseline, about half drank 1-3 cups of coffee daily.

Dr. Clem Lee: Okay, so far nothing crazy there. I also found this to be really interesting: that the participants didn’t know their assignments until the night before. So the night before, they would get a text message telling them that they were to drink coffee or not drink coffee for the next day. There was also a second text message at 8 AM the next morning. And so for the 14 day trial, each day was randomized, and researchers made sure that no person got the same assignment of coffee or no coffee 3 days in a row.

Dr. Shreya Trivedi: Hmm, can you just imagine getting a text message being like, “You can’t drink coffee tomorrow?” That brings up a good point. How did the investigators make sure the participants who got the text message to drink coffee were actually drinking coffee and the ones that were randomized to abstaining from coffee did not drink their coffee? 

Dr. Greg Katz: If I had been in this trial, I would have certainly been excluded because I would have withdrawn consent. So it’s super clever how they tracked these folks. And so participants were given Fitbit watches, ZioPatches, continuous glucose monitors, and a smartphone app called Eureka that tracked their geolocations.

Dr. Clem Lee: Yeah, so the researchers basically tracked the participants using this app to see if they went into a coffee shop or stayed away from a coffee shop. And this was specifically for people who reported in the beginning of the study that they went to coffee shops to get coffee. This didn’t really apply to people who made their own coffees at home. They also gave participants surveys to fill out and then had them press the button on their ZioPatch when they drank coffee. And both of these we know is a weaker form of validation since people can forget to press a button or lie on a survey. 

Dr. Katerina Lin: Interestingly, they were offered reimbursement for coffee even if they didn’t follow their assigned coffee days as long as they provided receipts. It’s a smart way to try to make coffee reporting more accurate since people were reimbursed regardless of whether they followed their randomization.

Dr. Shreya Trivedi: Yeah, that is smart. I would definitely fess up that I accidentally drank coffee, even if I wasn’t supposed to, knowing that, hey, I got some reimbursement out of it. Now let’s get into the endpoints, though. Start with our primary outcome: PACs, premature atrial contractions. I got to say, why? And maybe this is embarrassing to say, but, I don’t know, I kind of like ignore PACs when I see them on EKGs. 

Dr. Greg Katz: I basically ignore PACs also, unless they’re really symptomatic. So I don’t think that you’re wrong there. But the thing that really everybody is worried about with caffeine is the risk of AFib. And, you know, if you look at the burden of PACs across the population or for an individual, the burden of PACs is a pretty decent predictor of future AFib. So from a clinical perspective, PACs are a pretty reasonable surrogate if you want to think about the risk of AFib down the road. It’s not perfect, and ideally, you’d power the trial and follow people for long enough to actually see AFib. But looking at PACs isn’t so bad. 

Dr. Shreya Trivedi: Okay, thank you for explaining that a bit more. And, I guess, Katerina, I turn it over to you. Would you do the honors of telling us: what did they find with regard to coffee drinking and PACs on those Ziopatches?

Dr. Katerina Lin: So for the daily number of PACs, they found no difference: so for coffee versus no-coffee days, there were 58 versus 53 PACs over 24 hours. 

Dr. Greg Katz: That’s like 2 PACs an hour. So basically a minimal burden.

Dr. Clem Lee: Hm, interesting. And there were multiple secondary outcomes. So they looked at PVCs (premature ventricular contractions), step counts, sleep, and glucose levels. And these were all using the gadgets they had put on the participants.

Dr. Greg Katz: Yeah, so there was no difference in the glucose levels. But the other secondary outcomes are really interesting. There were more PVCs on days drinking coffee (154 versus 102 on average). The people in the study also took more steps during their coffee days (about 1000 more steps each day on average), but they had 36 fewer minutes of sleep per night, both of those measured via their Fitbit. 

Dr. Clem Lee: Hm. And when they analyzed the data per coffee drink, for each cup of coffee someone drank, they had 587 more steps but 14 minutes less sleep daily per cup of coffee.

Dr. Shreya Trivedi: Ah, interesting. I can’t help but wonder, like, were people just walking more to coffee shops and that’s why they had more steps?

Dr. Clem Lee: Yeah, I think that is certainly one theory that has been brought up. The other is that maybe coffee gives you more energy and helps you exercise more. And to put that into perspective, the average American walks about 5000 steps a day. So if coffee makes you walk 1000 steps more, that’s like a 20% increase in exercise.

Dr. Shreya Trivedi: That is awesome, right? Like, I feel like Dunkin, Starbucks, they got a new motto! Like “Ditch your trainers! Increase your step count by 20%!”

Dr. Clem Lee: Yeah!

Dr. Katerina Lin: Before we get carried away, it’s worth mentioning again that this was a small, time-limited study. There were only 100 people, and the trial was only 2 weeks. 

Dr. Clem Lee: Yeah, but I think we gotta give the researchers credit where it’s due. We talked a lot about how nutritional studies are really hard to perform.

Dr. Greg Katz: And to that end, there’re multiple reasons from a research perspective why this study is rigorous. The first, it’s a very clear intervention. Second, the adherence was really well tracked. Third, these outcomes (PACs, PVCs, step counts), they’re objective and so it’s not based on symptoms. There’s no placebo effect. It’s not palpitations. And four, everyone served as their own control: coffee on some days, no coffee on others. And so we don’t really need to worry about balancing the groups. This is nice; it’s the type of nutrition question that is truly testable. This is not “Does coffee cause cancer” but “Does coffee, over a short-term period, cause this concrete thing that we can truly measure?” So I’m personally really happy that we have this trial.

Dr. Shreya Trivedi: Hm, and that is a big deal coming from Greg Katz, if anyone knows him. Alright, so where does this leave us with coffee and then the messaging to our patients? 

Dr. Katerina Lin: It’s a complicated question. CRAVE was done on healthy people. So we can’t extrapolate it to people with cardiovascular disease. And so it’s really more complicated than blanket statements like: “Coffee is positive, no increase in PACs, people were more active,” or on the flip side, “Coffee is negative, more ventricular ectopy, less sleep.”

Dr. Clem Lee: Yeah, Katerina, and both of those interpretations could actually technically be correct. If I’m counseling like a young, healthy person who values energy and activity, I think this data feels reassuring. But if I’m talking to someone with symptomatic PVCs or a lot of sleep issues, I think this data leads to a very different conversation.

Dr. Shreya Trivedi: Yeah, like Clem, you have symptomatic PVCs (we’re presuming) every time you drink coffee to rounds. But what does that mean for your long-term cardiovascular health? Again, like, I’m still stuck on those endpoints, right? I don’t know the relevance of the PACs and PVCs per se. 

Dr. Greg Katz: My honest assessment is that we can’t really tell anything about long-term cardiovascular risk. And more than that, there’s a difference between saying that PACs are associated with the AFib versus “something is associated with more PACs so it’s also associated with AFib.” Those aren’t really the same thing. And, you know, going back to the study. 50 PACs over a 24-hour period. That is a normal number of PACs for a healthy, like, totally well patient to have. 

Dr. Shreya Trivedi: Yeah, what about those PVCs, right? There was a difference. So do we think of PVCs differently?

Dr. Greg Katz: Yeah, it’s really interesting. Because PVCs are often thought of as a harbinger of bad cardiovascular outcomes. But even 100 PVCs in 24 hours is kind of a nothingburger for a sick patient who has heart failure, let alone a young, healthy patient who doesn’t. There’re 100,000 heartbeats in a 24 hour period, so I’m not going to even spend 5 seconds thinking about whether 100 PVCs, or 150 PVCs, is really anything much at all in the grand scheme of things. 

Dr. Shreya Trivedi: So I guess what we’re saying is Clem is not going to go into heart failure, you know, with the 150 PVCs he’s having every day drinking coffee. 

Dr. Clem Lee: Yay!

Dr. Greg Katz: But, also if you told me that I could get my patients to walk 1000 more steps a day by just giving them coffee, then there’s a decent chance I would sneak into their houses and just force them to drink coffee every morning.

Dr. Clem Lee: Okay, that’s creepy, and I’m going to be locking my doors from now on.

Dr. Greg Katz: You’re not my patient, Clem. 

Dr. Clem Lee: Yet. 

Dr. Shreya Trivedi: That’s true. Arrhythmia clinic. 

Dr. Clem Lee: Exactly. I do wanna bring up one more study. It’s coffee in patients who already had AFib. This was cleverly called DECAF by the authors, which stands for “Does Eliminating Coffee Avoid Fibrillation?” It was published in JAMA in November 2025 (Wong et al. JAMA 2025).

Dr. Greg Katz: So the DECAF study took 200 coffee-drinking patients who had AFib and were going for cardioversion. They had half drink caffeinated coffee for 6 months and another group not drink coffee for 6 months. Fortunately, they saw that the coffee group actually had a lower rate of recurrence of AFib or atrial flutter.

Dr. Katerina Lin: One thing was that the AFib and atrial flutter were only clinically defined. There wasn’t a more structured or continuous way of measuring heart rhythms. 

Dr. Shreya Trivedi: Ah, so we wouldn’t have known if, like, a patient had, you know, a short bout of AFib while they were sleeping or if it happened subclinically?

Dr. Katerina Lin: Exactly, and since most patients reported few AFib symptoms, even if AFib reoccurred, they may not have noticed it. 

Dr. Clem Lee: Yeah, and also, maybe it’s just me, but I feel like if you want to study if coffee triggers AFib, then you should probably include more people who said that coffee triggered their AFib. And sadly, in the DECAF trial, only 40% of the patients reported that coffee triggered their AFib in the first place.

Dr. Greg Katz: But even with all those caveats, these two studies are probably some of the best data that we’re going to have on the question of coffee and arrhythmias. And so, my take is that even if you take the most pessimistic read of the data, for this group of patients who are pretty young and pretty healthy, that coffee is probably fine for you with respect to irregular heartbeats. But it might cause you to have a few more PVCs and mess up your sleep a little bit. 

Dr. Katerina Lin: So what are you going to tell your patients the next time they ask (which for you, Greg, may be tomorrow) if drinking coffee is good or bad for arrhythmias?

Dr. Greg Katz: I bet you, it will be tomorrow. And so when my patients, even those who have AFib, ask me if they can drink coffee, I tell them that if you like coffee, you can drink coffee. But you shouldn’t start coffee because “it’s a therapy for AFib,” because it’s not. And if it causes palpitations for you like it does for Clem, then you don’t have to drink it. But if you like coffee, I don’t think that there’s any compelling evidence (even including these trials) to say that it’s harmful for you, even if you have irregular heartbeats. 

Applying this to clinical practice for your patients

Dr. Clem Lee: Alright, so let’s zoom out a little bit. I think it’s step one to be able to look at the data and see what they actually say and to interpret that correctly. And hopefully, we’ve given you enough tools up until now to interpret data with all of its caveats. But then the next step, which arguably is harder, is actually talking to patients about nutrition. 

Dr. Walter Willett: So there’s a lot of what might be called implementation research that needs to be done where we do know a good direction, but how to make it easy for people to move in that direction. Certainly good advice is important as a starting point, but almost in all cases, just advising people alone is not sufficient. It has to be, there are barriers in terms of cost availability for many people, and we want to try to remove those barriers as much as possible. The role of physicians here, of course, is a key issue. And one of the things I’ve realized for quite a while is we actually haven’t given physicians the tools that they need to incorporate nutrition counseling into their practice. Or even though I’ve had some really good physicians during my life, none of them has asked me about what I ate. That’s a starting point, only a starting point, for moving onwards. How can you go start the discussion without knowing what a person’s eating? 

Dr. Shreya Trivedi: Yeah, I think we’ve all been guilty of this and not asking our patients what they are eating or drinking for that matter. And I think on top of it, the other, like, common scenario is when our patients are bringing up a very specific nutrition question. And AI won’t be able to help us with this. Like how do we communicate in a way that really resonates with the patient about what the nutrition research is lacking and what it might be telling us, and how do we say it in some level of confidence, knowing that there’s all these gaps? 

Dr. Greg Katz: So I think a good rule of thumb is that when the data are mixed (or even when there are confusing messages or contradictory messages about the topic), that probably means that the effect size is weak and so the stakes of any decision that you’re making about that individual item are probably pretty low. And so that’s why (I know this is not going to sound satisfying), but when I am talking with patients about nutrition, and we’re focused on details that I don’t think are all that important, I try to redirect them to things that are probably a better use of our time.

Dr. Shreya Trivedi: Yeah, and how do you respond when a patient is like, “Oh, my wife is putting garlic all over my food? Dr. Katz, is that good for me?” 

Dr. Greg Katz: I’ve seen that, too. And what I tell them, is if you like garlic, then you should eat garlic. But garlic isn’t a superfood because there’s no such thing. And so, when people are asking me questions about very specific foods or phytochemicals or nutrients, you know, I try to redirect them to the bigger picture nutritional advice that’s consistent across a whole bunch of data sources. But if you don’t like something, you shouldn’t be choking it down because again, there’s no such thing as magical food. And so sometimes, I’ll have patients do a food log for a week or a 24-hour diet recall, and I usually find that a fair number of my patients have very, very straightforward things that could be improved in their diet, whether they are drinking too many calories, they’re eating baked goods, cookies, pastries, candies. And so, if you look at the sort of consensus across all of the different diet trials, there’s a whole bunch of things that seem to be pretty consistent. You know, not too many calories. Drinking your calories is bad for you. Soda is junk food. Legumes, vegetables, fruits, whole grains–nobody disagrees that those are good. And so, think about the things that none of the diet gurus recommend, and try to avoid those things. And so when I’m talking with my patients, I really try to focus on the low-hanging fruit of their diet improvements. I think that one message is like, it’s really easy to confuse yourself, and a lot of questions about healthy eating come from people who know what junk food is and who are trying to find loopholes in that concept. 

Dr. Shreya Trivedi: Yeah, this is so hard, and I am so glad we’re talking about this. Because I think this is tricky on many fronts. The science is hard, and on top of that, the conversations can be hard. I’m always amazed, by the way, like what we tell our patients, and then like what our patients actually like hold onto. So I’m really curious from our listeners: what other approaches do you have in terms of handling these conversations, knowing the limits of the research and how hard some of these outcomes can be to really study?

Dr. Clem Lee: Yeah, and this is my personal opinion, but I believe that we as clinicians don’t think critically enough about nutrition. So I really hope that we’ve given you all some food for thought (you know, pun intended). And the next time you pick up your morning cup of coffee, or when you hear about the latest diet in the news or on social media, we hope that you now have more tools to help you think more critically about your food and diet. 

Dr. Greg Katz: And that is a wrap for today! Please feel free to share today’s episode with your friends and colleagues. 

Dr. Katerina Lin: And special thanks to Dr. Walter Willett for kindly dedicating his time and sharing his wisdom with us.

Dr. Shreya Trivedi: Thank you to our peer reviewer, Dr. Jane Leopold, and Dr. Jimin Hwang for the accompanying graphic. 

Dr. Clem Lee: If you have any feedback or suggestions, please email us at Hello@CoreIMpodcast.com. 

Dr. Greg Katz: Opinions expressed are our own and do not represent affiliated institutions.

References