HealthScienceSpecial Needs

The Weakest Link

Recently one of our bloggers asked for input on a new article, which promoted a link between acetaminophen (Tylenol) use and autism. In that post, Dr. Margulis argues that prenatal or juvenile use of acetaminophen may be linked to autism spectrum disorder, pointing to various scientific sources. Unfortunately, she greatly over-interprets many of these studies and fails to support her core argument.

The case is built upon a mixture of in vitro tests, retrospective clinical studies, and anecdotal statements from various sources. It generally dismisses the potential benefits of acetaminophen use, focusing only on the risk side of the risk/reward consideration. This unbalanced approach is, to my mind, unfair and misleading.

Acetaminophen causes cell death. Yes, absolutely. In a dish, in the laboratory, with artificially manipulated cells. But so do myriad other compounds at sufficient dosage, including coffee and bread, various berry extracts, and numerous important medications (thanks to the research I did for this post, I now have an expensive new book on my wishlist). In vitro studies have their place, but are often poor substitutes for in vivo data. In particular, neural cell cultures (Dr. Margulis’s linked study used liver cells, even further from the claimed autism focus of the article) are extremely difficult to interpret, because neurons do not grow easily in the lab. Moreover, the cultures lack the complex barrier and filter systems of the brain, nor the protective mechanisms within intact tissue. Very few molecules travel intact from the gut to the brain, and so one must use extreme caution in interpreting in vitro model studies.

[tl;dr sidebar for those seeking more detail: in vitro studies are conducted using isolated cells in a dish, maintained by an artificial nutrient bath. In vivo studies are conducted in a living animal, either human or otherwise. Both are useful for different types of investigation, but both need to be carefully interpreted when trying to build an understanding of human clinical conditions. For example: genetic manipulation is much easier in vitro, as is monitoring of specific molecules. However looking at individual cell types in isolation ignores interactions between different cells and tissues, and can over-emphasize small effects which get damped out in a large organism. A treatment that affects a few cells in isolation might have no meaningful effect when the animal’s complex metabolism, immune system, regulatory hormones, and other factors come into play.]

Dr. Margulis over-interprets the Bauer & Kriebel study, which correlated reported acetaminophen use and reported autism across eight countries, rather than within individuals. That is, if a nation used prenatal acetaminophen more, the study data shows a correlating increase in reported autism rates–far from a definitively causal relationship. Moreover, maternal acetaminophen use wasn’t directly measured: the data were obtained by a meta-analysis (a study of studies) to dig out reported use rates. Autism rates were drawn from CDC data, but autism is a challenging condition to research retrospectively, in part because the clinical definitions have changed frequently and significantly. See for example NPR’s recent article or the book it discusses for a comprehensive story on the subject. Suffice to say that autism reporting and autism occurrence are not well linked, and so any time-dependent analysis should be taken with a very large grain of salt.

The Danish and Norwegian studies cited are more compelling, in that they are very large cohort studies tracking individuals. However the Danish study relies upon reported and retrospective (after the fact) maternal use of acetaminophen, so we don’t know how dosing varied among the mothers. Neither study actually measured autism rates; the Danish study measured ADHD and Hyperkinetic disorders, while the Norwegian study measured various psychomotor and behavioral conditions. Certainly these are of concern to parents, but they’re not the same as autism.

In the end, Dr. Margulis provides three discrete recommendations: avoid acetaminophen for pediatric use, delay use of pharmacotherapy, and explore alternative medicines. Let’s consider each of these in turn.

Firstly, a blanket statement against acetaminophen is probably ill-advised. Of the analgesic compounds that are approved for over the counter sale in the US (principally aspirin, acetaminophen, ibuprofen, and naproxen), it has one of the best safety and efficacy profiles and a large body of scientific data. Certainly we shouldn’t be applying drugs indiscriminately–it is always critical to investigate the root cause of a symptom before applying a therapy that could mask important indicators. However denying pain relief altogether seems a poor choice.

So perhaps we should use acetaminophen less and more carefully. There is considerable evidence that we over medicate ourselves and our children (antibiotic misuse being among the greatest examples), but there are certainly cases where pharmacotherapy is fully justified, and delaying it can be harmful. Fevers can be dangerous if left unchecked (note that the linked article also points out many parents administer an incorrect dose of medication). Not all pain is muscular, so stretching isn’t always a solution, and frankly I’m not sure how one could get a screaming 2-year old to sit down and do yoga poses. Again, a blanket statement to delay use of a drug without input from a qualified healthcare professional is ill-advised.

Lastly, we have yet another promotion of unproven and untested ‘alternative cures’. Curiously, the author advises use of quercetin, based upon her husband’s experience. Yet the University of Maryland link provided explicitly states that, “There isn’t enough evidence to recommend quercetin for children.” I find this a curious approach in an article that objects to a drug with considerable scientific evidence behind it.

Is there a need for better pediatric studies to evaluate acetaminophen and other drugs? Absolutely. In fact I wrote an extensive post here a few months ago, encouraging support for the 21st Century Cures Act, which would promote such studies. Unfortunately it now appears that bill is in trouble in the Senate, despite overwhelming support in the House.

Is acetaminophen categorically safe and effective? Nope, nothing is. Water can kill you, but you can’t live without it. Oxygen is a destructive molecule, but try giving it up for an hour. Chocolates is delicious to me, but would kill my dog. A good friend of mine who studied pharmacology once taught me, “every drug has two effects. There’s the one you want, then there’s the other one.” Bottom line, use caution with all things. Use moderation. Listen to healthcare practitioners and have a dialog with them about medical decisions.

Finally, it’s worth noting that neither Dr. Margulis nor myself are licensed physicians. We both hold Ph.D. degrees, and neither of us is qualified or legally permitted to give medical advice. We are qualified to review the underlying science, but whenever in doubt about medical care, you should always seek the input of a licensed medical professional.


Topher is a nerd, scientist, bioengineer, photographer, costumer, dog trainer, beer brewer, and a lot of other nouns. He has too many degrees and too many hobbies because it’s all fun and interesting, and … squirrel!

Related Articles


    1. I didn’t realize until recently how dangerous an overdose of acetaminophen can be, and how there’s such a small gap between “safe dose” and “lethal dose.” Not that I take it cavalierly but now I pay much more attention.

      Connecting it with autism though, when the evidence is just not there, that’s just wrong. Especially when you have the authority of a Ph.D. behind your name, so more people are likely to trust you. What a disgrace.

Leave a Reply