WedMD has an article about a possible link between the HPV Virus and subsequent diagnoses of squamous cell and basal cell carcinoma. The article reports findings from research published in the British Medical Journal.
This is a link to the full BMJ research paper which, not surprisingly, reports nuanced and careful findings that cannot be easily translated and simplified. Among other things, the findings are qualified by the specific type of HPV virus and the specific type of carcinoma.
And just to show you how perilous it is to reach any medical or scientific conclusion simply from a quick reading of a headline or article title, take a look at the two headlines from WedMD and from BMJ (British Medical Journal).
They seem clear enough. One can easily imagine someone glancing at the headline, registering the finding, and moving on. The only problem is that farther down in each article, the authors report that people on long-term steroid medications for chronic conditions such as rheumatoid arthritis and asthma are at special higher risk of HPV-associated skin cancer.
My health science colleagues will almost certainly have quickly realized that these drugs have a mild, immunity-suppressing effect. In fact, HPVs have already been linked to skin cancer in transplant patients on immune-suppressing therapy and in people with a genetic disease (epidermodysplasia verruciformis) that suppresses immune responses.
But do you see how quickly a headlined “finding” got more complicated?
Which leads to an ongoing serious problem: Medical findings, when reported in the quick, brief manner typical of the mass media, are almost never treated with the nuance they deserve. The public reads a headline about an alleged connection between “virus A” and “heart ailment B” and almost automatically leaps to a belief in a direct, causal elationship:
“Oh no, I had virus A so I am going to get heart ailment B!”
Those of you in the health or social sciences, though, already know where I am headed. The connection might be complex and involve multiple variables. All sorts of other intervening variables might also be at work complicating the possible causal relationship.
And, while all of these findings might be clear in the actual research article, few of them find their way into media coverage.
In this hypothetical example, it almost certainly would be a very specific genus of virus A and a specific type of heart ailment B. And the connection might only be seen in people of certain ages and even certain ethnic backgrounds. And the correlation might or might not rise to the level of clear, unambiguous causation.
This would be just another standard media critique, but something quite serious is at stake. Many news consumers, for good or bad, will be using these findings to reach conclusions and take action. And in the case of health and science news, even highly educated, statistically literate consumers very well might not read the actual research article with all of the qualifications.
I can’t suggest an easy solution, although I do strongly feel that we need to be concerned about statistical literacy and the place of statistics in everything from elementary to higher education. It is astounding how much news — not only health and science — rests on embedded assumptions about correlation, causation, frequency distributions, variance, and all the rest.
Armed with such knowledge, people might begin to see that in many cases, the alarming headline they have seen isn’t actually as alarming as they might think.
Without better statistical literacy, we will continue to be plagued by social panics and anxiety about alleged perils that simply don;t hold up to statistical scrutiny.
It happens that every guy I know who has gotten a slightly elevated PSA test is also a frequent eater of chicken. But we can’t assume that …….
Steve, stop! They get the point.