Paul Alper points to this news article reporting on an observational study of 170,000 British people, finding that “Those who drank 1.5 to 3.5 cups of coffee per day, even with a teaspoon of sugar, were up to 30 percent less likely to die during the study period [approximately seven years] than those who didn’t drink coffee.”

A reduction of mortality by 30% seems implausibly high; on the other hand, it’s not impossible. Indeed, the news article quotes the journal editor who says, “It’s huge. There are very few things that reduce your mortality by 30 percent.” Alternatively, the news article states, “there may be other lifestyle factors contributing to that lower mortality risk among people who drink coffee, like a healthy diet or a consistent exercise routine.” I guess that smoking and drinking would be the biggest factors. I took a quick look at the research paper, and they do adjust for smoking and drinking, along with many other factors, but I don’t know exactly how that adjustment was done.

I’m not saying they did anything wrong; indeed, the authors of the research paper are in a damned-if-they-do, damned-if-they-don’t position, where if they find a small difference, people can say it’s no big deal, and if they find a big difference, people can say they don’t believe it. Still, they found a big difference, and I don’t know what to believe.

Reporting this sort of observational pattern seems like a good start, in any case. I’d like to reduce my own risk of dying in the next seven years by 30%, but I can’t quite bring myself to start drinking coffee every day.

So, yeah, I have nothing useful to add here. What do you expect?—It’s just a blog post! Seriously, I think it can be valuable for me to post sometimes just to express my ignorance and uncertainty. It’s a big world out there, and I have no insight, intuition, or expertise on this one.

35 thoughts on “coffee-study-lower-dying-risk.html

  1. “up to 30%” doesn’t mean 30%. It means one subset of the data (the folks who eat only yelllow M&Ms) have that value. (There’s an xkcd on this…)

    Also, presumably, how much coffee people drink is self-reported, not measured. And people who eat yellow M&Ms are known to be liars.

    As a tea drinker, though, I definately believe the self-reported tea drinker studies…

    • This was discussed in the comments here before.

      In the UK coffee drinking is relatively rare, and it typically means you are not a tea drinker.

      So actually they discovered tea drinkers are 30% more likely to die.

        • I always get in trouble with Andrew when replying to these comments from the peanut gallery, but sometimes they are funny so I will see what happens.

          Your link requires a login so I have no idea what numbers it shows, besides this quote which indicates relatively low coffee consumption (vs in the US and compared to tea in the UK):

          Some 27 percent of respondents drink two cups of coffee per day at home during the week.

          Anyway, the data in the paper showed that coffee drinkers drank less tea. This was further discussed in the accompanying editorial. Also, I said “relatively”. So numbers just for coffee from the UK cannot possibly address what I wrote anyway.

          Plus, it was sarcasm…

        • Anonyd00d, you’re the top hat wearing, monocle-doffing, black cane-carrying representative of this blog’s peanut gallery.

        • What does it mean when people misspell words on purpose? Not typos but phonetically: “Shurely”, etc then “anonydood”. Is it a version of leetspeak? I can’t imagine its that effective at avoiding bots/filters in this day and age.

          I’ve seen this a lot on zerohedge, which seems to be mostly trolls and old people with dementia in the comments. But since covid a few people started posting here with that “style”.

        • I’ve never read Private Eye. A google search indicates “The publication is widely recognized for its prominent criticism and lampooning of public figures.”

          I guess the usage is pretty consistent between this blog and Private Eye!

      • Coffee drinking in is much more common than in the old days. It’s very common to see people wandering around clutching a paper cup. But it’s a lot less common in the age group with highest death rates – ie 75+. I’m 71 and probably have one coffee per day on average but 5+ mugs of tea. Tea is far more refreshing when made properly.

  2. One thing strikes me as a little odd.

    “During a median follow-up of 7.0 years …”

    How were they able to do the follow-ups so that the median came out to be 7.0 on the nose? Or was that exact number a result of chance?

    • They followed up with exactly half of the population at 7.0 years and the other half randomly selected for longer times? They followed up with people at random times but some of those times were selected to be exactly 7 years and that bumped them above the 50% mark?

      Medians are pretty easy to control relatively speaking.

  3. The statistics of the health field is replete with the “C’s”—Coffee, Cancer, Circumcision, Chocolate, Celery. Endless studies of usually self-reported numbers. And, note the linguistic slipperiness of “up to 30 percent less likely to die” which is often in the advertising game read/interpreted as “at least 30 percent less likely to die.” If that last sentence is confusing, reconsider the legal nature of an advertisement which guarantees “up to 30 percent off” of the usual price but says nothing about an actual INcrease in price.
    This particular study is strange in a way I just now noticed. Namely, the data comes entirely from the UK, everyone of the authors is from China and additional explanations come from Harvard

    I am no more suspicious/paranoid than the next man, provided the next man is Richard Nixon.

  4. “Seriously, I think it can be valuable for me to post sometimes just to express my ignorance and uncertainty. It’s a big world out there, and I have no insight, intuition, or expertise on this one.”

    Yes agreed, quite valuable–if only some of the readers and responders of this blog would learn this from you!

  5. > Reporting this sort of observational pattern seems like a good start, in any case.

    Well, I think that’s the important point; it’s a start.

    It’s a starting point from which to further investigate, hopefully by following up with studies designed to test theories of causality. It’s a basis from which to look for mediating, moderating, and interaction effects – ideally in longitudinal interventions through which we can get closer to understanding causality.

    Is it the authors’ fault if they explicitly state that causality can’t be inferred from their observational study, and people infer causality regardless? Is it the fault of science by press realease? Is it the fault of the media? Is the public to blame? Let’s blame perverse incentives and journals and peer review? About about credentials and elitist “experts?”. Shouldn’t they shoulder the blame?

    I say none of the above (because all of the above). Better, imo, is to understand that there are all kinds of predictions and cognitive biases in play and its unrealistic to just hope that there weren’t.

    Amd to work on accepting ignorance and uncertainty without appealing to ignorance, which is a bit of a balancing act.

  6. This is an observational study, and in some ideal world it would be a prelude to a prospective randomized trial. Not all observational trials need that next step; observation of people who eat amanita verna wouldn’t require a follow up. Death is a good endpoint in clinical trials because it’s clearly defined. Over a large population, there are many causes of death. In this trial there were 1725 cancer deaths and 628 CVD deaths. This is a very different ratio from USA population data that consistently shows more CVD deaths than cancer deaths. UK mortality also shows this. What was special about their study population?

  7. The tale most told: In the past, the most cited benefit of coffee drinking was because it replaced the morning consumption of alcoholic beverages in England. The alcohol consumption was due to the lack of trust in the purity of the water.

  8. Interesting. In Table 1 it shows that the non-consumer group reported drinking around half a cup (mean = 0.4, sd=0.9) of coffee per day over the last year. Does this imply that many of the non-consumers were former coffee drinkers that had quit over the past year before the study?

    From methods:
    “Participants were invited to complete the questionnaire on 5 occasions over 1 year to account for seasonal variations in dietary intake between April 2009 and June 2012.”
    “Because participants could complete the 24-hour dietary recall up to 5 times, we classified a participant drinking coffee at any 1 dietary recall as a coffee consumer; all others were classified as non consumers.”

    So, a non-consumer was someone who didn’t happen to drink coffee on the day of the questionnaire (which could happen up to 5 times in a year). It doesn’t mean that they didn’t consume coffee. Just less frequently, as seen in the Table 1 results. Also, as noted in the study, this group appeared to consume more tea.

    Also, did anyone notice that when they clicked the link to the article, that the little ‘page loading’ animation was a coffee cup filling? haha

  9. The mean age was 55.6 with an sd of 7.9. Most deaths would have been amongst the elderly. Older people drink less coffee than younger people in the UK. Cox models with penalized splines seems unnecessary.

    • Yes, that’s a very good point. Coffeee drinking amongst the young is very prevalent amongst the young(ish) in the UK. It’s very common to see young(ish)sters carrying a drink in a cardboard cup with plastic lid and you can be 100% sure that ain’t a cardboad cup of tea.

      20 years of watching Friends most likely. This habit hasn’t come from that great coffee drinking nation Italy.

  10. “on the other hand, it’s not impossible. ”

    Seems pretty close to impossible. If I were asked by a supervisor to follow up on this, I’d put my effort into finding out what went wrong with the analysis rather than searching for causal connections between coffee and mortality. That seems like the best approach with any claim derived from purely from statistical analysis.

    • > That seems like the best approach with any claim derived from purely from statistical analysis.

      What’s the “claim” of which you speak? The Times article, and I assume the study itself (behind a paywall), explicitly spoke of the limiting nature of an observational study.

    • Exactly. This is really old “news”. Drinking more tea (especially green tea) and to a lesser extent coffee has been shown (in self-reporting studies, of course) to be “good for you” over and over and over again.

      What was really hilarious, though, was that coffee usually contains something arguably noxious (acrylamide) and there was a court case in California trying to make coffee chains stop selling coffee. Even though it’s fairly reliably shown that increased coffee drinking has a positive dose effect on overall health, including cancer.

      (This article, while covering the basics of the story reasonably well, isn’t as good on acrylamide as the Wiki article thereon. For those not interested in checking wiki, the bottom line is that “as of 2019 evidence from epidemiological studies suggests that dietary acrylamide is unlikely to raise the risk of people developing cancer”, and that smoking results in blood levels of acrylamide three times higher than any known dietary source.)

  11. SES is not observable and correlates with coffee drinking. Indicators like education and income are noisy. My bet is this is mostly SES to the extent that it’s “real”.

      • Now that’s a causal mechanism. Or…people who drink coffee in the U.K are cultural heretics, having declined tea, and thus they must be more resilient, gritty if you will, and so they have learned to steel themselves and resist death.

        The authors used the Total Deprivation Index to estimate SES and this does not seem like a tool sensitive to detecting anything other than poverty; it includes home ownership, car ownership, unemployment, and home overcrowding.

        There seems to be a strong connection between SES measures (race, wealth, education) and coffee consumption, so I’m guessing the model struggled to “control” for everything. The authors tossed in like 30 variables, many of which are likely collinear. Andrew, you’re the pro, so help me out here: What does it mean to ask about the impact of coffee consumption after accounting for age, education, material deprivation, waist size, BMI, fruit consumption, vitamin use, physical activity level, depression, smoking status, alcohol consumption, red meat consumption, diabetes, etc.? How do we interpret that output in real-world terms?

        The authors don’t report the effects of any of the other variables, so maybe the headline finding could’ve look a bit suspect if, say, they had found that vitamin usage had a 25% decline in mortality or whatever. We don’t know; it’s not reported, and the data are unavailable.

Leave a Reply to Andrew Cancel reply

Your email address will not be published. Required fields are marked *