Aspirin & breast cancer: another case study in communicating observational studies

There’s undoubtedly going to be a lot of miscommunication about the latest analysis coming out of the Nurses Health Study, which looks at the impact of different lifestyle factors on women’s health.

This time they tracked aspirin use, and then saw how many women were diagnosed with breast cancer. That’s an observational study – not a trial – and it can’t prove cause and effect.

The analysis showed that women who regularly took aspirin had fewer diagnoses of breast cancer.

What you can say is that this was a strong statistical association in a big study. But you can’t say it established a benefit because such a study can’t prove cause-and- effect.

Now here comes the NBC Nightly News – from Vancouver for the Olympics – with the water and the mountains in the background. And, rather than a carefully scripted videotape piece, we get live chatter between anchor Brian Willams and NBC chief medical editor Dr. Nancy Snyderman. Excerpts:

Brian Williams: “What is the benefit shown here?”

Snyderman should have stopped him right there. Benefit? As we’ve said, you can’t establish benefits from a study that can’t prove cause-and-effect. Sorry, but the language is important. See our primer on this topic.

In her answer, Snyderman briefly said, “They didn’t look at cause-and-effect…” but that’s awkward. It takes seconds to say, “They followed women who took aspirin and then saw how many got breast cancer. That doesn’t prove anything, but it does show a strong statistical link that will lead to more study in this area.” This is one of the pitfalls of trying to explain research in a live shot. She had a lot to say and she was live. She didn’t quite get there.

Williams continued: “That sounds almost anecdotal. What do we really know?”

Snyderman: “Well it’s not anecdotal but it is observational.”

Huh? Again, I know she tried, but that’s just jargon to viewers watching her with the water and the mountains in the background. She never explained what observational meant, or what the limitations of an observational study are.

Snyderman: “So here’s what I think you can say. There’s no proof that taking aspirin prevents breast cancer. But if you’ve been diagnosed and you’re on aspirin for another reason, it may be one extra benefit and, frankly, a reason to sleep a little better at night. ..So right now if you’re on aspirin, and you’re tolerating it and you’ve had a diagnosis of breast cancer, this is good news.”

Look at the cumulative language of this NBC story: “no proof that taking aspirin prevents breast cancer” but then again “benefit…reason to sleep a little better at night…this is good news.”

I don’t think you can say that at this point after an observational study.

Again, maybe this is one that should have been carefully scripted and pre-taped – rather than another NBC piece trying to capitalize on the beauty of Vancouver in the background for a live shot.

[2017 Update: This NBC video is no longer available]

For a better example of how to handle such a story, look at how Liz Szabo covered the caveats in her USA Today story:

“A study in August also found that aspirin offered a potential benefit against colon cancer.

Yet neither study proves that aspirin keeps cancers in check, Holmes says. That’s because doctors in each study merely followed patients for several years, noting which patients developed cancer and, of those, which took aspirin. So it’s possible that something other than aspirin controlled their tumors, Holmes says.

For proof, doctors would need to conduct a “gold standard” trial in which doctors randomly assign one group of patients to take a aspirin, then compare their progress with patients randomly assigned to a placebo, says Eric Jacobs, a scientist at the American Cancer Society.”