Listen on: Apple Podcasts | Spotify | Amazon Music | Podcast Index | Overcast | YouTube

A few weeks ago, headlines everywhere declared that drinking two or three cups of coffee a day reduces your risk of dementia by 20%. Social media ran with it. Morning lattes were suddenly rebranded as brain medicine. Then, almost as quickly, other voices chimed in to warn that caffeine triggers dangerous heart rhythms. So which is it — should you be drinking more coffee or less?

I’ve spent a good part of my career learning to read medical evidence carefully, and this story is a perfect illustration of why that skill matters. Let’s dig into the actual science — the dementia study, the heart rhythm trials, and what I personally do with this information.

The Coffee-Dementia Study: Impressive, But Flawed

The study generating all the buzz was published in JAMA, one of medicine’s most prestigious journals. Researchers drew on two long-running Harvard databases — one following nurses, another following physicians — for a combined cohort of 131,000 people tracked for up to 43 years. People who reported drinking two to three cups of coffee per day had a 20% lower likelihood of developing dementia than those who drank little or none. That’s a headline worth paying attention to.

But before you order a double shot in the name of neuroprotection, let’s talk about what this study can and cannot tell us.

First, it’s an observational study. Participants were not randomized to coffee or no coffee — researchers simply recorded what people reported drinking and watched what happened over time. That design can reveal associations, but it cannot prove causation. The history of nutrition science is littered with observational findings that didn’t survive the test of a controlled trial.

Second, what exactly counts as ‘a cup of coffee’? The study didn’t distinguish between a tiny espresso and a giant 20-ounce drip brew. The caffeine content alone can vary tenfold. This kind of exposure misclassification weakens any conclusions we try to draw.

Third — and this is where it gets genuinely interesting — tea drinkers showed a similar reduction in dementia risk. Tea has meaningfully less caffeine than coffee. Yet decaf coffee drinkers showed no benefit at all. So if caffeine were the protective ingredient, tea drinkers shouldn’t do as well as coffee drinkers. And if it’s some other bioactive compound like polyphenols, then decaf drinkers should have benefited too. Neither pattern is clean. That inconsistency is a real problem for the study’s main narrative.

There’s also the question of what statisticians call confounding. Coffee drinkers may be systematically different from non-drinkers in ways that affect dementia risk: education levels, exercise habits, sleep quality, overall health consciousness. Good studies try to adjust for these differences, but they never fully succeed.

And then there’s a lesser-known issue I find equally important: the ‘garden of forking paths’ problem in data analysis. When researchers have a large, general-purpose database, they can test hundreds of hypotheses. Often only the exciting findings get published. One classic study deliberately analyzed 1,200 variations of the question ‘does red meat increase mortality?’ and found that one-third of those analyses showed it increased mortality, while two-thirds showed it reduced mortality. All were legitimate analyses of the same data. The point isn’t that researchers are dishonest — it’s that statistical noise in a large database can be made to tell almost any story you want.

My bottom line on the dementia data: the study is real, it’s from a credible journal, and it’s worth knowing about. But I won’t be adding a second cup of coffee to my daily routine based on it. I drink one cup in the morning, and anything more disrupts my sleep. Sleep is one of the strongest levers I know for brain health. Trading quality sleep for a marginally better dementia odds ratio is not a trade I’m willing to make.

Coffee and Your Heart: What the Randomized Trials Show

For decades, patients who had heart palpitations or atrial fibrillation (AFib) were routinely told to cut out coffee. The reasoning seemed intuitive: caffeine stimulates the nervous system, speeds the heart rate, and might trigger dangerous arrhythmias. AFib is a serious condition — it raises stroke risk fivefold and accounts for enormous numbers of hospitalizations each year. So caution made sense.

But here’s the thing about intuitive reasoning in medicine: it’s often wrong. Two well-designed randomized trials have significantly updated my thinking on this.

The first is the CRAVE trial (Coffee and Real-time Atrial and Ventricular Ectopy). Researchers took 100 healthy adults and randomized them day-by-day — via smartphone notifications — to either consume or avoid coffee. Participants wore continuous cardiac monitors throughout. The findings were nuanced: coffee did not increase atrial arrhythmias, which is the territory where AFib lives. It did increase premature ventricular contractions (PVCs) by about 50%. PVCs are those ‘skipped beat’ sensations that many people notice — real, sometimes bothersome, but in otherwise healthy hearts, generally benign. Importantly, coffee was not associated with any increase in dangerous sustained arrhythmias.

One secondary finding from CRAVE deserves mention: coffee drinkers took about 1,000 extra steps on coffee days, and slept 30 to 40 minutes less. The extra activity is a nice side effect. The sleep loss is not — and it reinforces my personal calculus about sticking to one cup early in the day.

The second trial is the DECAF trial, which focused specifically on patients who already had AFib. After electrical cardioversion successfully restored normal heart rhythm, participants were randomized to either continue or stop their usual coffee habit. After six months, the coffee group actually showed lower rates of AFib recurrence than the abstinence group.

Now, I want to be precise here: I don’t think we should conclude that coffee prevents AFib from coming back. The DECAF trial was not large enough or blinded enough to support that conclusion. What it does strongly suggest is that the old advice — ‘you have AFib, stop drinking coffee’ — is not supported by controlled evidence. The case against coffee in cardiac patients is far weaker than we long assumed.

What I Actually Do — and What You Might Consider

I’m a one-cup-a-day person, and that won’t change based on the new dementia study. My sleep is too important to sacrifice for speculative brain benefits from an observational analysis. I also pay attention to timing — my coffee goes down early, well before it can interfere with sleep onset.

What this evidence does tell me is that my morning coffee is almost certainly not hurting my brain or my heart. That’s genuinely reassuring. The dementia findings, even with all their limitations, don’t support the hypothesis that coffee accelerates cognitive decline. And the randomized cardiac data says that for most people — including many with known arrhythmias — moderate coffee intake is not the danger it was once thought to be.

If you’re a coffee drinker who enjoys it, current evidence gives you no good reason to stop. If you’re debating adding a cup or two specifically to protect your brain, I’d hold off — the evidence doesn’t yet clear that bar. And if you have a cardiac condition and your doctor has told you to avoid coffee, it’s worth having a conversation about whether that advice is still supported by the best available evidence.

The deeper lesson here, as always, is about reading the evidence clearly. Large observational studies with exciting headlines are often the beginning of a scientific question, not the answer to it. The questions to ask: Was this randomized? What’s the absolute risk difference? Could confounders explain the finding? Is the result biologically coherent across all subgroups?

Ask those questions consistently, and you’ll navigate the noise of health headlines far better than most.

—

Listen to the full episode at drbobbylivelongandwell.com. To sign up for the newsletter or send Dr. Bobby a note, visit the site directly.