Skip to main content

Coffee Is Bad For You. Coffee is Good For You. Why Am I So Confused?

Coffee is one of the most widely consumed substances in the world.  I suppose many drink coffee because they believe it tastes good.  But for many, the appeal of coffee is for its medicinal properties-specifically the stimulant effects of caffeine.

The popularity of coffee and its dual use as a beverage and drug make its health effects an important public health issue.  So, it is not surprising that a recent study in the New England Journal of Medicine received a lot of attention.  The study had two important findings:

(1) People who drink coffee are more likely to die
(2) People who drink coffee are more likely to live


Well, it is actually true.  The study really did say both of these things.  And in the end, the study does not really answer whether coffee is good or bad for you.  But it is worth taking a few moments to try to understand these contradictions.  Because if you do understand, you will be a better consumer of the research you read about in the medical literature and the popular press.

The study design is simple enough.  The researchers administered a dietary questionaire to nearly 400,000  people between the age of 50 and 71 in 1995.  They asked them if they drank coffee, and if so, how much.  Then they determined who died by 2008.  They compared death rates in those who did not drink coffee, light coffee drinkers, and heavy coffee drinkers.

And after 13 years, coffee drinkers did worse--actually a lot worse.  In men, 13.1% of those who did not drink coffee died.  18.8% of those who drank 6 or more cups of coffee per day died.  In women, 10.4% of those who did not drink coffee died.  15% of those who drank 6 or more cups of coffee per day died.  This sure seems bad.  No more Ventis for me.  Oh Well.  Perhaps I can put that $2.25 to better use.

But wait a second.  Have you ever heard of the concept of confounding?  In medical research studies, a confounder is something that is strongly associated with the risk factor (coffee drinker), that explains its association with the outcome (death).  One way to think about this:  Up to now, we were assuming that the higher rate of death in coffee drinkers was due to the coffee.  But what if coffee drinkers are very different from non coffee drinkers?  And what if those differences, rather than coffee consumption is what really explains the higher death rate in coffee drinkers.

Well, as it turns out, coffee drinkers are MUCH different from non coffee drinkers.  They actually are a fairly unhealthy group.  Most notably, coffee drinkers are much more likely to smoke.  For example, in women, 8% of non coffee drinkers smoke, while 48% of those drinking more than 6 cups a day smoke.  Coffee drinkers are also less physically active and have worse dietary habits.  Could it be all this bad stuff (confounders), rather than the coffee that is leading to more deaths in coffee drinkers?

It turns out that if one can measure a confounder, it is pretty easy for the researchers to account for it.  The  researchers are able to use statistical tools to adjust for these confounders.  This makes it possible to see what the difference in mortality would be if rates of smoking and other measured health markers were the same in coffee drinkers and nondrinkers.  The point of these statistical procedures is to get closer to the truth by making apples to apples comparisons.  For example, how would mortality rates compare in nonsmokers who don't drink coffee and nonsmokers who do drink coffee.

The result of this statistical adjustment is a rather stunning reversal of fortune for coffee drinkers.  After accounting for all of the bad health habits of coffee drinkers, those who drink coffee actually have a lower risk of death than those who do not drink coffee.  Men who drink 6 or more cups of coffee a day have a 10% lower risk of death than those who do not drink coffee.  6 or more cups in women is associated with a 15% lower risk of death.   If you smoke and drink coffee, you live longer than if you smoke and don't drink coffee.  If you don't smoke and drink coffee, you live longer than if you don't smoke and don't drink coffee.

So, coffee is good for you?  Well not so fast.  While it is true that researchers can statistically correct for confounders like smoking, they can only adjust for confounders that they actually measure.  What if there are other unmeasured factors that make the coffee drinkers live longer?  For example, what if wealthy people are more likely to drink coffee?  What is people who have more medical problems drink less coffee?  The study was able to do little to account for either of these highly plausible scenarios.  It is quite possible the better outcomes of coffee drinkers have nothing to do with coffee.

So, the study first showed that coffee drinkers are more likely to die.  When the researchers accounted for smoking, coffee drinkers were less likely to die.  It is possible if they were able to account for additional differences between those who do and do not drink coffee, the results would swing back in the other direction.

So, in the end, we still do not know whether coffee is good or bad for you.  (My best guess--it is probably neutral--drink coffee is you like coffee.  If you don't like coffee, drink something else).

In the end, the most important lesson:  Have some healthy skepticism when you read the latest news in your medical journals or newspaper about which foods are good or bad for you.

by: Ken Covinsky (@geri_doc)


Great explanation of research principles. This study is a classic Cum Hoc, Ergo Propter Hoc fallacy, or With This, Therefore Because of This. It wrongly assumes that things that occur together cause each other; without taking into account other, or confounding, explanations for the outcomes. The real question is why studies like this are even published in the first place.
Mike Steinman said…
Coffee once again appears to be the ultimate geriatric health intervention - extending both lifespan and quality of life (the second effect wasn't formally studied, but I'll take it as a given based on personal experience).
diet clinics said…
I feel this is among the most significant information for me. And i am glad reading your article. But wanna commentary on some common things, The website taste is perfect, the articles is in reality nice. Good process....
Eric said…
I wonder if a key piece of the puzzle is that someone who drinks 6 cups of coffee a day is less likely to drink a lot of soda than a non coffee-drinker?
Brad Stuart said…
Ultimate solution: coffee in the morning, red wine at night. That's the N. CA way. Seriously, nice piece, Ken.
Anonymous said…
I have not seen any were, the consideration of what most people put in coffee. Milk, sugar, aspartame. All know irritants! Particularly the latter. Straight black organic coffee is like every thing else we eat or drink, moderation. Only my opinion of course, I have no medical back ground. In my opinion, if we took complete notice of all "medical" research, we would die early any way, from dehydration and starvation. Again, only my opinion.
Camaro said…
Very good article :)
Victor said…
As a non-selective A1 and A2A adenosine receptors antagonist in the brain and the most widely and legally used psycho-stimulant, caffeine increases loco-motor activity, alertness, vigilance and arousal via complex pathways of adenosine-dopamine, adenosine-serotonine,adenosine-nicotinic/cannabinoid, and adenosine-glutamate receptors binding and depending on genes(genome) coding for the availability of the receptors.Furthermore, caffeine metabolites such as paraxanthine, theobromine and theophylline are ten times more powerful then caffeine itself. Like ethanol consumption is regulated by alcohol dehydrogenase racial and ethnical genomic controlled, caffeine use is regulated by cytochrome P450 genotype and it metabolized by CYP1A2 enzyme(fast or slow metabolizers). As you all know from the epidemiology surveys and questionnaires are very biased(collection, response, cultural bias etc.)and geno-typing is only needed to infer from high quality evidence the valid results.
From Russia with love
Dr. Victor Ananyev
Gail Cooney said…
Where do you find a $2.25 venti, Ken??
Mark Bright said…
I guess that goes to show that even in the well-calculated sections of the medical profession, there are still parts that tend to screw us up. The best thing to do, I guess, would be to do and consume things in moderation, and keep a skeptical eye on anything new popping up.

Popular posts from this blog

Lost in Translation: Google’s Translation of Palliative Care to ‘Do-Nothing Care’

by: Cynthia X. Pan, MD, FACP, AGSF (@Cxpan5X)

My colleagues often ask me: “Why are Chinese patients so resistant to hospice and palliative care?” “Why are they so unrealistic?” “Don’t they understand that death is part of life?” “Is it true that with Chinese patients you cannot discuss advance directives?”

As a Chinese speaking geriatrician and palliative care physician practicing in Flushing, NY, I have cared for countless Chinese patients with serious illnesses or at end of life.  Invariably, when Chinese patients or families see me, they ask me if I speak Chinese. When I reply “I do” in Mandarin, the relief and instant trust I see on their faces make my day meaningful and worthwhile.

At my hospital, the patient population is about 30% Asian, with the majority of these being Chinese. Most of these patients require language interpretation.  It becomes an interesting challenge and opportunity, as we often need to discuss advance directives, goals of care, and end of life care options…

Delirium: A podcast with Sharon Inouye

In this week's GeriPal podcast we discuss delirium, with a focus on prevention. We are joined by internationally acclaimed delirium researcher Sharon Inouye, MD, MPH. Dr Inouye is Professor of Medicine at Harvard Medical School and Director of the Aging Brain Center in the Institute for Aging Research at Hebrew SeniorLife.

Dr. Inouye's research focuses on delirium and functional decline in hospitalized older patients, resulting in more than 200 peer-reviewed original articles to date. She has developed and validated a widely used tool to identify delirium called the Confusion Assessment Method (CAM), and she founded the Hospital Elder Life Program (HELP) to prevent delirium in hospitalized patients.

We are also joined by guest host Lindsey Haddock, MD, a geriatrics fellow at UCSF who asks a great question about how to implement a HELP program, or aspects of the program, in a hospital with limited resources.  

You can also find us on Youtube!

Listen to GeriPal Podcasts on:

Are Palliative Care Providers Better Prognosticators? A Podcast with Bob Gramling

Estimating prognosis is hard and clinicians get very little training on how to do it.  Maybe that is one of the reasons that clinicians are more likely to be optimistic and tend to overestimate patient survival by a factor of between 3 and 5.  The question is, aren't we better as palliative care clinicians than others in estimating prognosis?  This is part of our training and we do it daily.   We got to be better, right? 

Well, on todays podcast we have Bob Gramling from the Holly and Bob Miller Chair of Palliative Medicine at the University of Vermont to talk about his paper in Journal of Pain and Symptom Management (JPSM) titled “Palliative Care Clinician Overestimation of Survival in Advanced Cancer: Disparities and Association with End of Life Care”.

Big findings from this JPSM paper include that we, like all other clinicians, are an optimistic bunch and that it actually does impact outcomes.   In particular, the people whose survival was overestimated by a palliative care c…