Episode 4: A Tale of Uterine Cancer, Hair Products, and Bad Stats

Ever heard of noise? I’m certain you have. In science, when we talk about noise, we’re normally talking about the variance or the variability when trying to measure something. In toxicology, that noise is usually the variability associated with a control or unexposed group. If you were to look at heights across a large number of people, you’ll no doubt see a distribution that looks kinda like a bell-shaped or “normal” curve. Most people are huddled together near the middle, and then we have some shorter and some taller people — the tails. Now, consider that distribution, that bell shaped curve, to represent the unexposed population — our controls.

In toxicology we’re also going to have some exposed group. And again, for the sake of argument, that group will also have a distribution of responses, and let’s say it’s also a bell shaped curve distribution. Now then, in simplistic terms, if those two distributions overlap completely, so that they’re identical, we’d say there is no treatment effect. What needs to happen is for the response in the treated group to be so different from control that we can’t conclude that the treated group’s response came from the control distribution.

Said another way: if I randomly sample from the control distribution and get the same response as I’m seeing in the treated, then there’s no treatment effect. That’s the simplistic way of explaining this (there are some nuances I’m not bringing up).

Well, in today’s episode, we’re going to talk about a case where the “signal” (that is the treatment response that NIEHS’ scientists say is a real treatment effect) is actually just noise from the control group.

Here’s my analysis report on how I came to this judgement, and you can see the analysis code here.

But there are other issues with this particular NIEHS study that says hair relaxers, straighteners, and pressing products are linked to uterine cancer. For starters, NIH’s press release made a causal statement that, “women who used chemical hair straightening products were at higher risk for uterine cancer compared to women who did not report using these products” — that’s simply not true at all because this study didn’t find causality.

But the other big issue — the study isn’t even representative of the US population of women. 85.6% of the women in the study were white only, and not hispanic or LatinX — just white. That’s way more than we have in the US population. The number of Black or African American women is well below the US population. Same goes for other non-white groups. Which means, these results are not even relevant to the US population of women. We can’t translate these results to the US population.

Listen to today’s episode to find out more!

Latest articles

Related articles

Leave a reply

Please enter your comment!
Please enter your name here