Video: https://www.tiktok.com/@distilledscience/video/7559712264022953246
Transcript
I got to call out Lane Norton for citing fake studies in a supposedly evidence-based post about Tylenol and autism. I have a lot of respect for Lane in general, and if you want generally high-quality health and fitness information combined with a lot of sass and fire, then I suggest you go follow it. But in this case, it looks like he had a preformed opinion, i.e. acetamapin use during pregnancy is unlikely to cause autism or ADHD. And then he had an AI come up with a bunch of citations to support his existing belief. But he did not do his homework and actually read the studies before. citing them. He writes that fever itself is a much stronger predictor of ASD risk and cites these two studies. But I looked up those PMIDs. This first one is a study on symptom-guided heart failure therapy, while this second one is a material science paper about designing better food packaging. I'm guessing he meant to cite this paper instead, which does show a more than doubling of the incidence of autism in a child when the mother gets a fever during pregnancy, a risk that drops to only 1.3x when they take a fever lowering medication. So at least here his point was correct. But he also starts by saying that some studies show an association between Tylenol use and autism. For instance, the cord blood biomarker study found a modest increase in risk with this PMID, modest. But the odds ratios here are 2.14 and 3.62, i.e. double to triple the risk. That's not modest. Again, this doesn't change his main point because this study had a lot of confounders and you can't really use it to determine any sort of causation. And better evidence, like the sibling study he mentioned, shows no increased autism incidence during one pregnancy compared to her own baseline of pregnancies where she did not use it. And as Lane says, don't forget the genetics. Twin studies show autism concordance is 77% in identical twins, versus 31% in fraternal. So if we know that genetics accounts for most of the autism incidents, no drug could be the prime driver. Except that study only said that 77% number for male twins. For females, it was only 50%, albeit with a much smaller sample size. So, is this post problematic? At least this time, the mistakes didn't impact the accuracy of the overall message and argument. But as evidence-based science communicators, we need to hold ourselves to a very high standard, so that our followers trust that we're coming at these issues with an open mind, and not just looking for studies to support our pre-existing beliefs. As a man I still look up to once said, Now, sadly, people with good credentials, honestly, there should be a higher standard on them to actually like tell the truth. have read the study they are citing, especially when they are broadcasting this out to millions of people on these major media platforms. The onus is on the scientist who is there, who is supposed to be reputable, to actually represent the data correctly. Yep. But if you want some tips for how to use AI and get better citations, let's talk.
Additional notes
His post got 341k views, but not a single comment about these fake references. The last time I caught a mistake he made in interpreting a paper I sent him a message about it but didn’t get any response, so I figured i’d do this for some more public accountability. This does NOT mean that you shouldn’t trust him in general! I’ll reiterate here: I have a tremendous amount of respect for everything he does, and I bet that he did read the papers that were more central to his argument. But just as I’ll readily admit that many big sources of misinformation also sometimes provide excellent health advice, so too sometimes the more reliable folks mess up and should be held accountable for it. I get the pressure to publish commentary on these topics as they’re coming off the press. I feel it myself, and often get a bit stressed out when i end up waiting a week or two before publishing a video because I've taken the time in between to read a hundred or so studies, and then by the time I do publish, nobody cares anymore. But I’d rather be slow than wrong, even in a way that most people won’t care about. @Layne Norton, PhD #science #evidencebased #tylenol
References
- Layne Norton post about Tylenol/autism discussed in caption and transcript; direct URL not listed in workbook.
- Fake/mismatched PMID examples, fever/autism paper, cord-blood biomarker study, sibling study, and twin autism concordance study mentioned in transcript; exact IDs/source links not listed in workbook.