The annals of confusing news stories about observational studies showing an association between coffee and…fill in the blank…have a new entry.
Do a Google search for “coffee and melanoma” and you’ll get thousands of returns.
Many of these stories inappropriately used causal language – suggesting that a cause-and-effect had been proven, when it hadn’t.
TIME, for example, headlined it, “This Drink Could Protect You From Skin Cancer.” Yes, it could. But perhaps it would not. And you need to evaluate the evidence, which was observational and could only point to a statistical association, not causation. So “could protect” is misleading in our view. So is “may help thwart…may also be doing your skin a favor…coffee can protect against melanoma” – other phrases used in the story. “Association” was mentioned once, but without explanation or discussion of limitations about drawing causal conclusions from this kind of research. This ending section is more confusing than helpful:
“The group says that their results need to be repeated and confirmed, and that it’s too early yet to change your coffee habits to protect yourself from skin cancer. But the findings support the idea that there might be more you can do to protect yourself from the sun’s harmful rays than only slathering your body in sunscreen.”
CBSNews.com stated that “protective benefits of coffee increased the more a person drank.” It is simply inaccurate to discuss protective benefits when you have not proven cause-and-effect. It’s interesting to note some of the wisdom of the crowds – this user comment left online in particular:
“I don’t think they can draw such conclusions (cause and effect). Coffee drinkers may simply be working more (indoors, less UV) at a job whereas those who work less may be outdoors more and have no need for coffee. Not saying this is true, but their conclusions aren’t necessarily either.”
FoxNews.com repeatedly used the word “protection.” We need to be protected from such misleading language.
On ABC’s Good Morning America, Dr. Jennifer Ashton finally got around to explaining association versus causation, but you have to sit through about 90 seconds of anchor chit-chat and giggling to get to it. Nonetheless, she eventually got there.
HealthDay gets high marks for succinctly stating: “The study only uncovered an association between coffee consumption and melanoma risk; it didn’t prove a cause-and-effect relationship.” One sentence…20 words…made all the difference.
If you really want to learn from an in-depth (and I don’t think you’d find it too academic) explanation of the limitations of the research, see Dr. Perry Wilson’s video on MedPage Today. Excerpt:
“But if we dig a bit deeper, we might not be ready to dig for the extra cup just yet.”
He discusses confounding factors and multiple looks at the same dataset, something he explains very well. Listen and learn.
Coffee is the poster child for abused translation of observational research. If you search for “coffee” on this blog, you come up with dozens of examples of the wrong language used to describe “coffee and xyz” research through the years.
To try to help, we offer a primer, “Does The Language Fit The Evidence? – Association Versus Causation.”
Follow us on Twitter: