Valentine’s Day isn’t the only time observational studies are miscommunicated. It just seems that way.

Health news this week is dripping with warm, gushing claims about the health benefits of chocolate – just in time for Valentine’s Day. Headlines such as:

Chocolate – the love drug.

Dark Chocolate & Red Wine – The food of love and health

Chocolate is good for health and relationships.

But one blogger wrote, I have “Chocolate Concerns.”  Is Chocolate as Healthy as Claimed?Excerpt: 

Most cocoa studies are funded by chocolate manufacturers because that industry wants to sell you chocolate by convincing you of all the health properties of cocoa.

Good for her.

But we’d also remind readers to remember that almost all such claims come from observational studies that cannot prove a cause-and-effect relationship.

The pervasiveness of the wrong language used to describe observational studies drove us to post a primer on the topic several years ago.

But it’s hard for journalists – or everyday consumer/readers – to read between the lines when even medical journals fail to adequately emphasize the limitations of observational studies.

We’ve written about BMJ news releases on observational studies before – in fact, chocolate hype was one of the topics. They did it with a news release about a white rice and diabetes study, too. And again with a news release about alcohol and arthritis.  And, most recently (at least of the ones we’ve caught), a BMJ journal’s news release has delivered mixed messages on another arthritis study – this one headlined, “Sunlight may help ward off rheumatoid arthritis in women.” Mind you, this was creative research from a huge sample population, but it based its conclusion on where people lived (including latitude, altitude, and cloud cover) and then estimated UV exposure.  That may result in a statistical association, but it can’t prove that sunlight “warded off” anything in the women studied.  No cause and effect could be proven by this kind of study.  Yet “effect” was mentioned 3 times. “Association” in its various forms was also mentioned 3 times.  But there wasn’t a word about this kind of study being unable to go beyond association into causation.

So how should you interpret the claims that doing housework hurt men’s sex lives?   Read about this association, which did not prove causation.

And how should you interpret the news that “Another benefit of exercise – at least for Caucasian men – is that it may cut the risk both of developing prostate cancer and having high-grade disease”? You’d have to be really good to pick up on the hints in the MedPage Today story:

  • about a “relationship” or “association” between activity and prostate cancer.
  • about the activity levels being self-reported.

Because the story may have confused readers by using phrases like:

  • cut the risk
  • there was no significant benefit

It danced around the edges, mentioning “that the cohort was small, which raises the possibility that the findings are the result of chance” and that “factors that were not measured, such as diet, might have played a role” and that the questionnaire used in the self-reporting was “subject to the participant’s interpretation.”

Bottom line:  There are built-in limitations in observational studies.  They don’t make the research invalid or unimportant by any means.  But they don’t warrant causal verbs.

OK, you’ve slogged through another one of these lectures.  You deserve a candy bar.

You might also like

Comments (2)

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

Alia Bucciarelli

February 14, 2013 at 9:21 am

Thanks, Gary. We are reviewing an observational chocolate study in the class I teach at Tufts called “Writing about Health and Medicine.” On tap is to discuss absolute vs. relative risk and caveats about observational studies. I’m bringing this post to share!