Haircare Churnalism on NPR
A story that upset my Sunday Morning routine
I wrote this article at the behest of my family. On Sunday, 10/23/22, we were enjoying our usual Sunday morning - breakfast, Sunday newspaper and NPR Weekend Edition Sunday with Ayesha Rascoe. Ms. Rascoe is, by far, my favorite host since the Susan Stamberg days ended in 1989.[i]
In the midst of listening I started ranting at a story entitled “Researchers Have Found a Link Between Chemical Straighteners and Uterine Cancer.” My family told me to please shut up and write an article about it so they could go back to enjoying their morning.
Sensible Medicine is a reader-supported publication. If you appreciate our work, consider becoming a free or paid subscriber.
The NPR story covered an article just published in Journal of the National Cancer Institute. The journal article reported data from the Sister Study, a cohort study. The researchers enrolled 33, 947 women, aged 35-74, during the years 2003-2009. In baseline questionnaires, participants were queried on their use of hair products (dyes; straighteners, relaxers or pressing products; permanents or body waves) in the prior 12 months. The results reported in this study were the associations between hair product use and uterine cancer over an average of 10.9 years of follow-up.
The results showed that there was only a relationship between uterine cancer and hair straighteners. Ever use vs never use was associated with uterine cancer with an adjusted HR = 1.80, 95% CI = 1.12 to 2.88. There also was a dose response curve. The absolute risk increase was about 1.2% for a number needed to harm of about 85.
A few interesting points to the study:
All patients enrolled in the study had at least 1 sister diagnosed with breast cancer (hence, “Sister Study”).
The demographics of the study: 7.4% Black/African American, 4.4% Hispanic/Latina non-Black, 85.6% non-Hispanic White, and 2.5% all other race and ethnicity.
The results were adjusted for race, education BMI, physical activity, menopausal status at enrollment, parity, age at menarche, and use tobacco, alcohol, oral contraceptives and hormone replacement therapy.
The researchers looked not just at whether you had used hair products but whether you had applied (non-professionally) them on other people. While the main results showed that use of hair straighteners was associated with uterine cancer (HR, 1.80; 1.12 to 2.88), application to others showed a similar, though non-significant, trend (HR, 1.42 0.72 to 2.79). Either arguing for a highly toxic product or the presence of residual confounding.
There was negative confounding in this study. The HR went from 1.63 to 1.80 when confounders were statistically removed.
Brief take-away: This is an observational study of a very particular population. It shows a small association (causation cannot be demonstrated in an observational study) with wide confidence intervals between use of hair straighteners and uterine cancer. There is certainly residual confounding (no adjustments for socioeconomic status among other things) but the confounding may be negative. The trend toward an association in those who non-professionally applied straighteners is concerning for confounding.
The NPR Report
If you did not hear the report, please stop reading and listen to it or read the transcript. What got me so aggravated was that this was an interview with a researcher, not a journalistic evaluation of a study. The NPR story raised concerns about a common practice, almost certainly causing fear, without pointing out the weak points in the study . The only mention of the inability to infer causation was helpfully volunteered by the author.[ii] The study author was encouraged to make behavior change recommendations based on these results. She obliged, as authors of observational studies often do.
This is certainly an example of churnalism. We define churnalism as the careless, incurious reporting of poorly done biomedical research. Churnalism trades the real story -- why this study is unimportant or proves something other than it contends -- for the easy headline. There are seven deadly sins of churnalism and this story certainly managed to commit a few.
The seven deadly sins:
Observational studies almost never prove causation
Extrapolation and generalization
Ignoring confounding, selection bias and other epidemiological errors
The ‘Disclaim and Pivot’ maneuver.
Keep testing; report just once
In this report, we witness sin #1, noted in passing by the author, not sought by the journalist. We see sin #2 as data from a very specific group of woman, those with a sister who developed breast cancer, is extrapolated to all woman interested in getting their hair straightened. There was no discussion of confounding, sin #3. Sin #6 comes in to play when a relationship is found be examining data in multiple ways until something this found. Let’s give the authors credit for doing the research well but we do not know if this relationship has not been found in other studies. The 7th sin, being incurious, was definitely committed and could have been remedied with the 5 questions I would love to have heard.
I’d suggest 5 questions that would have made this report health journalism rather than churnalism:
1. This was an interesting, and I have to say frightening, result. Besides the ill effect of straighteners, are there other explanations for the results you found?
2. The Sister Study looks at an important, but unique population. I’d expect these women are at higher risk for cancer. Should we be comfortable extrapolating these results to other populations?
3. This is an observational study and we know that observational studies demonstrate association rather than causation. What in this study makes you think the relation is causative and what makes you think it might just be an association?
4. I found the trend toward an increased risk of uterine cancer in people who apply straighteners interesting. Could this because people who used straighteners, and their friends and family, are at higher risk for cancer irrespective of their use of straighteners?
5. The absolute risk here, even in this population, is small. What are your next steps in trying to prove there is a causative relationship at play here?
Now that would have led to radio that would have made my bagel go down well!
[ii] It is true that when you look at harm, we generally rely on observational data, we cannot randomized people to an exposure we expect is harmful.