Gary Schwitzer is founder and Publisher of HealthNewsReview.org. He tweets as @garyschwitzer, or using our project handle, @HealthNewsRevu.
The limits of observational studies–once again missing from health care PR news releases. And from health care news stories.
Two examples of each caught our eye in this pre-Thanksgiving news slowdown period. As real news slows down, news about observational studies often fills the void. Just what we didn’t need.
Here are some things that journalists–and their readers–should know.
The University of Missouri sent out a PR email to journalists with the subject line, “Can bacteria in yogurt calm your anxiety?” It began:
Zebrafish could hold some insight into the way we handle stress – and probiotics might help us decrease our anxiety levels; this according to a recent study released in Scientific Reports, a journal published by Nature. … researchers at the University of Missouri, using a zebrafish model, determined that a common probiotic sold in supplements and yogurt can decrease stress-related behavior and anxiety.
Point of emphasis #1: Zebrafish ≠ humans. So “your anxiety”–teased in the email–is on an entirely different plane. The relevance of this animal model to “your anxiety” requires a huge leap, one readers should be cautious to make.
Point of emphasis #2: Correlation ≠ causation. So suggesting that probiotics “can calm your anxiety” is an overstatement, drawn from observations of fish in a fish tank. Please! The more complete news release to which the Mizzou email links is headlined, “Common Probiotics Can Reduce Stress Levels, Lessen Anxiety.” It, too, is over the top.
Point of emphasis #3: This is not a knock on the importance of basic research. It is a knock on overstating the results of basic research.
The University of North Carolina Health Care System and the Cambridge Health Alliance sent out a news release headlined, “Alarming disparities in health outcomes could be prevented by breastfeeding.” It began:
A new study published in the Journal of Pediatrics showed that black infants had more than twice the deaths of whites attributable to lack of optimal breastfeeding. Black infants also had more than three times the rate of necrotizing enterocolitis, a devastating disease of preterm infants, attributable to suboptimal rates of feeding with their mother’s own milk.
The end of the news release revealed:
(The study) used computer modeling to simulate hypothetical cohorts of black, white and Hispanic women and the children they bore, based on current literature linking disease risk and breastfeeding.
It struck me that not only was this another case of an observational study which cannot establish cause and effect, but it was one in which other factors–confounding factors–may explain what the researchers thought they found. And the computer modeling simulation introduces another question mark.
We recently added two pediatricians to our team of editorial reviewers/contributors and I turned to one of them, Alan Schroeder of Stanford, for a reaction to this news release. He wrote:
Studies that have taken clever approaches to circumvent the residual confounding issue that has plagued most investigations of breastfeeding outcomes have demonstrated less of a positive health impact of breastfeeding. Given the uncertainty around the exact effect sizes for breastfeeding benefits, and given the possibility that there are several factors that might confound the association between race/ethnicity, breastfeeding rates, and health outcomes, I would be cautious about interpreting the results presented in this study. It is misleading to state: “Suboptimal breastfeeding is associated with a greater burden of disease among non-Hispanic blacks and Hispanic populations” from a hypothetical cohort of patients.
Schroeder once wrote a blog post for the American Academy of Pediatrics on “residual confounding” in which he focused on breastfeeding. He wrote:
That breastfeeding is good is a central tenet of pediatrics. But we are basing almost all of our evidence on observational studies. …it is certainly possible that there is something unmeasurable about women who elect to breastfeed their babies (and/or about babies who are able to successfully breastfeed), and this unmeasurable variable or variables may in fact be at least partly responsible for the health benefits that are observed in children who were breastfed.
… Like any intervention, it is imperative that we are honest and transparent about possible complications. When this intervention goes well, complications are minimal; but without proper support and resources, it doesn’t always go well.
You can read his blog post to learn more about what doesn’t always go well.
The point here is that PR news releases can paint a very simple, clear picture of a situation that, in reality, may not be simple nor clear.
Finally, journalists make the causation leap often enough on their own, even without PR prodding. One example this week: HealthDay headlined a story, “Bonus From Your Blood Pressure Med: Fewer Fractures?” The story began:
High blood pressure and weakened bones are two big health issues for seniors. Now, new data suggests that one class of drugs (thiazides) might help protect against both.
Another example: Fox News & Men’s Health reported, “Your Netflix binges are lowering your sperm count.”
Headlines and incomplete stories like this lower readers’ health literacy IQ.
Both of the studies reported on in these cases were observational, providing no proof of cause and effect. But there was no mention of that limitation in either story. Further proof of how there are many parties in the polluted stream of research news to the public who need to do a better job of communicating research results.