For breakfast, give me 2 observational studies and an anti-irritant

Gary Schwitzer is the founder of and has been its publisher for 14 years. He has been a health care journalist for 47 years. He tweets as @garyschwitzer or as @HealthNewsRevu.

This is the way many of my days begin.

I check for messages – across all media – and I see this on Twitter from Adam Cifu, MD, one of our former editorial contributors:

Once again, it’s the New York Times making causal statements about an observational study, confusing a study that shows a statistical association with one that shows cause-and-effect – which this one did not. And once again, it’s in the Times’ Well column. 

In this case, the Times acknowledges the study can’t prove cause and effect – as early as the second sentence.  However, by then the story had already used wording suggesting causation twice:

  • Headline: “Sitting All Day May Increase Your Risk of Dying From Cancer.”  Adding ‘may’ to ‘increase your risk’ doesn’t provide absolution. 
  • First sentence:  “Sitting for hours on end could heighten someone’s risk of later dying from cancer.” Again, adding the qualifying ‘could’ isn’t sufficient.  The story still made a causal link – ‘heighten someone’s risk.’  You can’t prove that kind of effect if you haven’t proven cause-and-effect.  Pretty simple.

So readers heads were spinning like Linda Blair’s in The Exorcist – just in the first two sentences.

But I’ve often felt that the way you end a story is the most important because it leaves the reader with a take-home message. And in this case, the Times turned over the take-home message to one of the researchers:

“The tangible takeaway is that we can tell people they do not have to go out and run a marathon” to potentially reduce their risk of dying from cancer, she says.

I argue that there are no proven grounds to deliver that ‘tangible takeaway.’  It’s an interesting study, blown way beyond the boundaries of evidence with this story and with that quote.

As always, smart readers weigh in with criticism of the Times’ continued Un-Well coverage.  One wrote:

With these correlational studies, you can’t control for every single variable that may be contributing to the picture and I do not agree with the stark message of “Don’t sit too long or you’ll die from cancer.” That’s too simple and terrifying a picture to be useful to anyone.


Even earlier today, I saw this on Twitter:

Well, he didn’t need to wait long to see a headline like this in The Telegraph from the UK:  Alcohol is good for you, study finds. 

No, that’s not what the study found.  It found a statistical association. Not proof of cause-and-effect.

CNN’s story – “ Moderate drinking may improve cognitive health for older adults, study says” –  used a qualifying term  – ‘may improve’ – which still implies that cause-and-effect ‘may’ have been established when it hasn’t.  Overall, though, the CNN story provided context and caveats in the body text.

As I’ve done dozens of times through the years, I refer readers and journalists to the primer that’s been on this website for more than a dozen years: Observational studies: Does the language fit the evidence? Association vs. causation.

Now it’s time to sit down and have a drink. I’m worn out already today and I’m barely beyond breakfast.

You might also like


Please note, comments are no longer published through this website. All previously made comments are still archived and available for viewing through select posts.

Comments are closed.