Almost everyone has been told at some point that coffee is hard on the heart. It is standard advice after a palpitation, a common instruction before a cardiac test, and a widespread personal rule among people who have felt their pulse jump after a strong cup. The unusual thing about this belief is how little evidence ever supported it, and how thoroughly the evidence that finally arrived complicated it.
Why this question stayed unanswered for so long
Coffee and heart rhythm is a hard thing to study observationally, for a reason that runs in the opposite direction from most nutrition research. People who get palpitations from coffee tend to stop drinking coffee. That means that in any large survey, the group of heavy coffee drinkers has been quietly filtered to exclude the people most sensitive to it, while the group of non-drinkers is enriched with people who quit precisely because of symptoms.
This is not speculation. Hyppönen and Zhou tested it directly using Mendelian randomisation in UK Biobank[1] and found that cardiovascular symptoms genuinely do shape habitual coffee consumption, with people experiencing symptoms drinking less. So observational studies showing that coffee drinkers have fewer heart problems are, to some unknown degree, measuring self-selection rather than a protective effect. The only clean way out of that trap is to randomise.
The CRAVE trial
That is what Marcus and colleagues did in the CRAVE trial, published in the New England Journal of Medicine in 2023[2]. They enrolled 100 adults, mean age 39, half of them women, and fitted each with three devices at once: a continuously recording ECG, a wrist-worn accelerometer, and a continuous glucose monitor. Over 14 days, participants received daily text messages randomly instructing them either to drink caffeinated coffee or to avoid caffeine entirely.
The adherence checking was unusually rigorous, and worth describing because it is what makes the result credible. Compliance was verified through real-time indicators recorded by participants, daily surveys, reimbursement for date-stamped coffee receipts, and geofencing that virtually monitored visits to coffee shops. This was not a study relying on people remembering what they drank.
The primary outcome was the mean number of daily premature atrial contractions, the extra beats originating in the upper chambers that are associated with developing atrial fibrillation. If coffee were the arrhythmia trigger it has long been assumed to be, this is where it should show.
The result
It did not show. Coffee days produced 58 daily premature atrial contractions on average, against 53 on caffeine-free days: a rate ratio of 1.09, with a confidence interval from 0.98 to 1.20 and a p-value of 0.10. In plain terms, the small difference observed was well within the range you would expect from chance, and the trial's primary hypothesis was not supported.
The secondary outcomes were more interesting. Premature ventricular contractions, extra beats from the lower chambers, were significantly more frequent on coffee days: 154 per day against 102, a rate ratio of 1.51. Participants also walked more on coffee days, 10,646 steps against 9,665, and slept less, 397 minutes against 432, a difference of about 36 minutes a night. Glucose was effectively unchanged.
So the trial delivered a genuinely mixed verdict, and it is worth resisting the urge to flatten it. Coffee did not provoke the atrial ectopy most linked with atrial fibrillation. It did provoke measurably more ventricular ectopy, whose long-term significance in otherwise healthy people is uncertain but not nothing. And it visibly changed behaviour and sleep, which are themselves cardiovascular risk factors operating on a much longer timescale than fourteen days.
What the extra ventricular beats might mean
The premature ventricular contraction result is the finding most worth sitting with, partly because it is the one that gets dropped from summaries reporting that coffee was cleared. A rate ratio of 1.51 is not marginal, and the increase from 102 to 154 beats a day was well outside the range of chance.
The interpretation is genuinely uncertain. Occasional premature ventricular contractions are extremely common in healthy people and are usually considered benign; most people have some and never know it. At high burdens, however, they are associated with cardiomyopathy over long periods, and the thresholds where that becomes a clinical concern are typically expressed as a percentage of total beats. A jump from roughly 102 to 154 in a day, against a total of around 100,000 heartbeats, remains a very small fraction, which is the reasonable basis for not being alarmed by it.
The honest summary is that the trial found a real, measurable increase in a phenomenon whose long-term significance at this magnitude is unknown, in a group of healthy adults with a mean age of 39. It would be wrong to report that as coffee being harmless to heart rhythm, and equally wrong to report it as coffee being dangerous.
What the trial says about the rest of your day
Two secondary findings had nothing to do with heart rhythm and are arguably more useful. On coffee days participants walked 1,058 more steps, and they slept about 36 minutes less.
The step count is a reminder that caffeine is a stimulant with behavioural consequences, and that some of coffee's favourable long-term associations could plausibly run through activity rather than through any direct biological effect. If coffee reliably makes people move more, and physical activity is one of the best-evidenced determinants of cardiovascular health, that is a mechanism worth taking seriously and one that no biochemical study would ever detect.
The sleep finding cuts the other way. Losing half an hour of sleep a night is not trivial, and short sleep is itself associated with cardiovascular risk. So the same trial found coffee nudging one well-established risk factor favourably and another unfavourably, in the same people, over the same fortnight. We cover the sleep evidence in detail in our article on caffeine and sleep.
What the very large observational studies show
A two-week trial in 100 people cannot say anything about developing heart disease over decades, so the population data still matters. The most detailed analysis by preparation method is Chieng and colleagues' study of 449,563 UK Biobank participants[3], median age 58, followed for around 12.5 years, which separated decaffeinated, ground and instant coffee rather than lumping them together.
The subtype breakdown in that study is worth dwelling on, because it isolates caffeine reasonably well. Ground and instant coffee were associated with reduced arrhythmia; decaffeinated coffee was not. But all three subtypes, decaf included, were associated with reduced cardiovascular disease and reduced all-cause mortality. If that pattern is real, it suggests caffeine may be doing something specific for rhythm while coffee's other compounds carry the broader cardiovascular association. That is a hypothesis generated by the data rather than a demonstrated mechanism, and it is exactly the sort of finding that needs replication before anyone builds on it.
Ground and instant coffee were associated with significantly reduced arrhythmia at 1 to 5 cups a day, with the lowest risk at 4 to 5 cups for ground coffee, a hazard ratio of 0.83, and at 2 to 3 cups for instant, a hazard ratio of 0.88. Decaffeinated coffee showed no significant arrhythmia association. All three subtypes were associated with reduced incident cardiovascular disease, with the lowest risk generally at 2 to 3 cups a day. All-cause mortality was reduced for every subtype, most strongly for ground coffee at 2 to 3 cups a day, a hazard ratio of 0.73.
The umbrella review by Poole and colleagues[4] reached the same broad conclusion from 201 meta-analyses, putting the largest reductions at three to four cups a day, including cardiovascular mortality at a relative risk of 0.81 and cardiovascular disease at 0.85. So the observational picture is consistent and favourable, with the important caveat about self-selection above.
The one preparation detail that does appear to matter
There is a real cardiovascular finding attached to coffee, and it is about how the coffee is made rather than how much of it you drink. Tverdal and colleagues followed 508,747 Norwegian adults for an average of 20 years[5] and separated filtered from unfiltered brewing. Filtered coffee was associated with lower mortality than drinking no coffee at all, with hazard ratios of 0.85 for both men and women. Unfiltered brew sat higher, at 0.96 for men and 0.91 for women, and in men aged 60 and over the hazard ratio for cardiovascular death from unfiltered coffee versus none was 1.19.
What about blood pressure?
Blood pressure is the concern people raise most often, and the evidence is more reassuring than the reputation suggests. Caffeine does produce an acute rise in blood pressure, which is well established and easily measured. What matters clinically is whether that translates into sustained hypertension in habitual drinkers, and the cohort evidence generally does not support it: meta-analyses of prospective studies have not found habitual coffee intake associated with a higher risk of developing hypertension, and some report a modest inverse association.
The usual explanation is tolerance. The pressor effect of caffeine attenuates substantially with regular use, so the spike a naive drinker experiences is not what a daily drinker experiences. This is also why measuring your blood pressure shortly after a cup can be misleading, and why clinics ask patients to avoid caffeine beforehand.
The caveat is the same as everywhere else in this article: people who develop blood pressure problems may cut back, which biases the observational picture in coffee's favour. If you already have hypertension, the guidance from your own doctor takes precedence over a population average, and the acute effect is real enough to matter for measurement even if it does not accumulate.
The mechanism is well established and has nothing to do with caffeine. We cover it in full in the article on cafestol and cholesterol, but the short version is that unfiltered brewing methods leave in diterpenes that raise LDL cholesterol, and a paper filter removes almost all of them. If you drink several cups a day and you care about cardiovascular risk, this is the single most actionable finding in the whole coffee literature.
Who should still be careful
None of the above applies uniformly. The CRAVE participants were healthy volunteers with a mean age of 39, and a trial in that group says little about someone with an established arrhythmia, a cardiomyopathy, or a history of atrial fibrillation. The increase in ventricular ectopy on coffee days was real, and in a person with underlying heart disease that is a more meaningful observation than it is in a healthy 39-year-old.
The trial also ran for fourteen days. Whatever coffee does to a heart over forty years, a fortnight cannot see it.
van Dam, Hu and Willett's review[6] takes the same line: general population findings do not translate into individual clinical guidance, and people with existing cardiovascular conditions should follow their cardiologist rather than population averages. If coffee reliably gives you palpitations, that experience is data about you and outranks a trial average.
The summary
The best randomised evidence available found that caffeinated coffee did not significantly increase the atrial ectopy associated with atrial fibrillation, but did increase ventricular ectopy and did cost participants roughly half an hour of sleep a night. Large observational studies consistently associate moderate coffee drinking with lower cardiovascular disease and lower mortality, though the self-selection problem means those numbers likely flatter coffee somewhat.
It is a more textured answer than either camp in this argument tends to want, and it is the one the evidence actually supports.
It is also, for a question this old, the first time a proper randomised design has been brought to bear on it at all.
The practical takeaways are narrow and worth keeping: use a paper filter if you drink a lot of it, stop drinking it early enough that it is not costing you sleep, and if coffee gives you palpitations, believe yourself rather than the average. For the everyday version of this, see our guide to coffee and health.
None of them require you to give anything up.