Of all the daily habits people feel guilty about, coffee is the one that deserves it least. Not the sleep debt, not the bourbon, not the two hours of scrolling before bed. The coffee.
It is one of the few consumption habits where the public health message and the actual evidence point in opposite directions, and the evidence has been pointing the same way for a long time now.
The umbrella review
Poole and colleagues published an umbrella review in the BMJ in 2017 that pulled together 201 meta-analyses of observational research covering 67 distinct health outcomes, plus 17 meta-analyses of interventional research (1). That is about as wide a look at a single dietary exposure as exists.
The finding: coffee was more often associated with benefit than with harm, across nearly every way of slicing the exposure. The association was non-linear, with the largest reduction at three to four cups a day compared with none. All-cause mortality relative risk 0.83. Cardiovascular mortality 0.81. Cardiovascular disease 0.85 (1).
High versus low consumption came with an 18 percent lower risk of incident cancer (1). There were also lower risks reported for several specific cancers and for a set of neurological, metabolic and liver outcomes.
And the harmful associations that did turn up were largely explained away once smoking was adequately adjusted for, which is the oldest confounder in this literature. Coffee drinkers used to be smokers. That is much less true now than it was in 1980, and the older studies never sorted it out properly.
The genetics question, which is the one I find most persuasive
Here is the objection that always came up. Some people metabolize caffeine slowly because of common variants in CYP1A2 and elsewhere. If coffee were genuinely protective you would expect it to matter how fast someone clears it, and if the benefit were confounding you would expect genotype not to matter at all.
Loftfield and colleagues tested this in the UK Biobank, published in JAMA Internal Medicine in 2018, using a genetic caffeine metabolism score and ten years of follow-up (2). The inverse association between coffee drinking and mortality held regardless of genotype, and it held in people drinking more than five cups a day.
You can read that two ways. Either the benefit is not caffeine-dependent, which points at the several hundred other compounds in coffee, or the whole thing is confounded in a way genotype does not touch. I lean toward the first, and I hold that loosely.
Five reasons, honestly stated
Mortality. Three to four cups a day sits at the bottom of the curve for all-cause and cardiovascular death in the pooled data (1).
The liver. This is the most consistent signal in the whole coffee literature and the one most often missed, with lower rates of fibrosis, cirrhosis and hepatocellular carcinoma in coffee drinkers and a clear dose relationship, which is why I bring it up with anyone who has fatty liver disease and has already been told, correctly but unhelpfully, that the answer is to lose weight and exercise more. Coffee costs nothing. It is one of very few things on that list that does not require them to change their life.
Type 2 diabetes. Inverse association, dose dependent, and present for decaffeinated coffee too, which again argues that caffeine is not the active ingredient.
Neurological outcomes, where the associations are real and the mechanisms are less settled.
And the one nobody counts: coffee is a small structured pleasure that people build a morning around. I am not going to pretend that is a clinical endpoint. It is still worth something.
Where the caveats actually are
Pregnancy is the genuine exception. The harmful associations that survived adjustment in the umbrella review were in pregnancy (1), and the standard advice to stay under 200 mg of caffeine a day stands.
Then there is what people put in it. A drink with 400 calories of syrup and cream is not the exposure any of these studies measured. “Four coffees a day” and “four of those” are different sentences, and the second one belongs in a conversation about calories rather than a conversation about coffee.
Sleep is the other one. Caffeine has a half-life of roughly five hours, longer in some people, and the afternoon cup is a real contributor to the insomnia that patients then treat with something else.
Reflux, palpitations, anxiety. Real for individuals, not population-level reasons to stop.
For patients
You do not need to quit coffee for your heart. Three to four cups a day is where the association looks best, and there is no evidence that you should push toward that number if you do not want it.
Watch what goes in the cup, keep the last one well before bed, and stay under 200 mg a day if you are pregnant.
For colleagues
The liver association is the most actionable piece of this and the least used. Patients with MASLD get told to lose weight and exercise, both of which are hard, and rarely get told that coffee has consistent evidence behind it.
The UK Biobank genotype analysis is worth having on hand for the patient who has read that slow metabolizers should avoid coffee.
The Bottom Line
Coffee comes out ahead in almost every outcome anyone has pooled, with the largest benefit around three to four cups a day and an all-cause mortality relative risk of 0.83. The effect does not depend on how fast you metabolize caffeine, and it shows up for decaf in the diabetes and liver data, which suggests caffeine is not what is doing the work. Pregnancy is the real exception. The rest of the caveats are about sleep and about what you add to the cup, not about the coffee.
Related Reading
How Do You Actually Lose Weight? A Doctor Explains
Does Fasting Slow Aging? What the Science Says
Longevity Medicine: What Actually Works, and What Doesn’t
Scott Rennie, D.O.
Board Certified in Obesity Medicine and Family Medicine
This blog is for educational purposes only and does not constitute individual medical advice. Always consult your own physician before making changes to your health, medications, or treatment plan.
Sources
1. Poole R, Kennedy OJ, Roderick P, et al. Coffee consumption and health: umbrella review of meta-analyses of multiple health outcomes. BMJ. 2017;359:j5024. PMID 29167102. https://pubmed.ncbi.nlm.nih.gov/29167102/
2. Loftfield E, Cornelis MC, Caporaso N, et al. Association of Coffee Drinking With Mortality by Genetic Variation in Caffeine Metabolism: Findings From the UK Biobank. JAMA Intern Med. 2018;178(8):1086-1097. PMID 29971434. https://pubmed.ncbi.nlm.nih.gov/29971434/