December 18, 2012
Please, Grey Lady, don’t spill more coffee observational studies on us
The New York Times – in its print edition today and on its Well blog – reported, “Risks: Coffee Linked to Fewer Oral Cancer Deaths.”
That is technically accurate. An observational study like this – actually a questionnaire-based survey of a large number of people – can point to a statistical association – a “link” as it were. It cannot, however, prove a cause-and-effect relationship between coffee consumption and fewer oral cancer deaths.
But stories like this can be technically accurate while falling far short of helpfully complete.
First, it’s based on a study in the American Journal of Epidemiology, “Coffee, Tea, and Fatal Oral/Pharyngeal Cancer in a Large Prospective US Cohort.” If you really care, and if you can find your way around journal articles without getting hurt, you can read the entire study online.
Here’s what the story could have told readers, to be most helpful:
- It could have explained the inherent limitations in observational studies. We have a published a primer on the topic of the importance of the proper language to describe such studies. Even a journal group joined in this refrain recently, warning researcher-authors about the importance of language. While this story did not use the wrong language, it certainly didn’t use enough of the correct language in our review.
- The closest the story came was this: “The authors acknowledge that they could not distinguish whether coffee drinkers were less likely to get oral or throat cancer or more likely to survive it. The lead author, Janet S. Hildebrand of the American Cancer Society, said that the mechanism was unclear, but that coffee contains compounds that may have anticancer effects.” But why not just add a line such as “This kind of study cannot prove cause-and-effect.” ???
- Its use of numbers from the study was also not the most helpful. It reported: “the researchers found that the risk of death from oral or throat cancer was 26 percent lower among those who drank one cup a day, 33 percent lower among those who drank two to three cups daily, and 50 percent lower among those who drank four to six cups daily, compared with those who drank no caffeinated coffee.” 26 % lower than what? What is the baseline risk? The numbers used frame this a possibly huge effect size (even though it didn’t cross the line and claim a cause-and-effect!) – but they don’t tell readers how big a threat is oral cancer to begin with.
On its blog, at least, if not in print, the New York Times has all the room in the world to explain things like this. Use links if you must. But please, Grey Lady, don’t let your writers contribute to the back-and-forth ping-pong games of “coffee lowers risk…coffee heightens risk” stories that seem to endlessly pour forth from the coffee pot of observational studies.
Follow us on Facebook,
and on Twitter: