Why Tylenol Is (Still) Fine to Take During Pregnancy

Back in September, President Trump and Health Secretary RFK Jr. held a press conference on autism. At this conference, they claimed, among other things, that consumption of Tylenol (drug name: acetaminophen) during pregnancy was a significant cause of the increases in autism rates.

As I (and many others) discussed at the time, this claim is not well supported in the best data. But the ramifications of the administration’s position have been significant. A recent paper in The Lancet showed that prescriptions for acetaminophen in emergency rooms dropped about 10% in the immediate aftermath of this press conference.

Although the administration has, to some extent, walked back its claims about Tylenol, a lot of uncertainty remains for many pregnant people, and this is an active research area. Last week, a new paper in JAMA Pediatrics weighed in with some slightly confusing results. You’re likely to hear more about this paper, if you haven’t already, so I wanted to take today to unpack it. The bottom line is that this new paper strengthens our conclusions about the safety of Tylenol in pregnancy. The interesting thing is what this tells us about statistics.

Take me to the bottom line

What did we know already about Tylenol in pregnancy?

There is a temptation when a new paper comes out to act as if it supersedes anything we had before. This is generally a mistake. It’s rare that a new study leads us to ignore prior work, and this is no exception. We already have a lot of data on Tylenol in pregnancy.

To briefly summarize, when we compare children born to moms who took Tylenol (especially many times) during pregnancy to those whose moms did not, we see higher rates of ADHD and autism in the children who were exposed during pregnancy. However, it is difficult to learn from this type of data. The women who take Tylenol differ from those who do not in many ways other than just their Tylenol consumption, and in this case, we especially worry that this correlation might be driven by the reasons they took the medication.

Fully controlling for all differences across families is very hard. More recently, researchers have gotten around this by looking within families. Instead of comparing across families, they compare sibling pairs in which one child is exposed and the other is not. In these studies, like a large one from Sweden published in 2024, we do not see a relationship between Tylenol exposure and either autism or ADHD. In other words, the correlation that some studies observe seems to be driven entirely by differences across moms. When we are able to hold the mom constant, there is no relationship.

This is the best data we have, and it is the set of findings that experts are relying on to say that there is no relationship between Tylenol and these diagnoses.

What does the new paper show?

The new JAMA Pediatrics paper uses a dataset of 2 million births in Taiwan between 2004 and 2015. The authors observe whether the mother had a prescription for acetaminophen during pregnancy and whether there were ADHD or autism diagnoses in their children. The basic analysis is the same as existing work: look at whether children exposed to Tylenol in utero are more likely to have these diagnoses. I’m going to focus on ADHD in this discussion, as that is what the researchers primarily focus on as well, but their results on autism are similar.

(Note: this paper uses prescription data to measure exposure. It is also possible that people accessed these medications over the counter, although that is a smaller part of the market in Taiwan.)

The first set of results in the paper (see the graph below) is exactly what we saw in the 2024 paper from Sweden. If you look at the relationship across families, there is a strong positive correlation between the extent of Tylenol exposure and ADHD diagnoses. However, when you look within families, the relationship disappears.

So far, this is simply what we have seen before, with another huge sample. It’s reassuring evidence that when you hold constant enough factors, Tylenol during pregnancy does not have negative impacts.

The authors then present a second set of results. Here, they analyze the siblings in two groups, focusing only on families with exactly two children. First, they run the analysis with only the cases where the first, older sibling was exposed and the second, younger one was not. They find that, in these cases, exposure during pregnancy seems to increase ADHD diagnoses. Second, they run the analysis with only the cases where the older sibling was not exposed and the younger one was. There, they find that exposure during pregnancy seems to decrease diagnoses.

You can see this in the graph below. The data here is a little murkier since there is no clear “dose-response” relationship — the data makes it look like any exposure changes risk, but the risk doesn’t scale with the amount of exposure (e.g., number of prescriptions).

This is confusing. The authors say as much. So … what does it mean? I’ve been thinking about this for days, and I don’t have a full answer but I think we can make a little progress.

What do these results mean?

We should start with what they do not mean. One possible (wrong) takeaway from this would be that Tylenol exposure in a first pregnancy increases ADHD risk, but in a second one, decreases it. However, this is not biologically plausible, and the fact that the risk doesn’t scale with the amount of exposure makes this an even worse take. Moreover, the one other study we have that does this type of analysis — the one from Sweden — shows the opposite pattern, with a positive effect showing up only for later children.

Instead, what must be going on here — as the authors note — is some other form of bias. There must be some other difference, something which varies over time within a family.

One possible explanation relates to how people respond to a first child’s diagnosis. Consider this: imagine some underlying genetic factor impacts the mom as well as her children’s risk of ADHD or autism. Let’s say this factor also increases the chance that the mom takes Tylenol in a first pregnancy, perhaps to address symptoms related to this. (To be clear, we don’t know which genetic factors contribute to autism, but it is possible they would also impact the frequency of taking Tylenol.)

Then imagine that if the mom has a first child diagnosed with ADHD or autism, she may start to wonder whether her behavior, perhaps taking Tylenol, was responsible for this. This could mean that in a later pregnancy, she would change that behavior and be less likely to take Tylenol than other people.

With this pattern, you would get what you see in the data: higher risk for the first child, lower risk for the second, compared with their sibling. Unfortunately, this is not a theory that I (or the authors) could test in these data, so it remains speculative.

What should you take from this?

I suspect you’ll be hearing about this study from all sides — people may claim, for example, that this does show a causal link with Tylenol for some children. I want you to have the context for when you see those claims. (Also, I love, more than anything, trying to understand biases in observational data — it’s my great passion and I strongly believe we must do what we love.)

What you should take from this is two things.

First, with their first analysis, we now have yet more data showing that Tylenol during pregnancy is not associated with autism in children. With the new data, combined with existing work, this has been demonstrated in over 4 million births across three countries. If you need pain or fever relief during pregnancy, please take Tylenol.

Second, sibling analyses can be complicated. I believe these can tell us a lot more than analyses across families, but any kind of nonrandomized data, which these are, is subject to biases. There is no substitute for careful thinking about statistics.

The bottom line

Despite initial claims by the Trump administration linking Tylenol use in pregnancy to autism, the strongest available evidence thus far has shown no causal relationship between acetaminophen and autism or ADHD.
A new study out of Taiwan confirmed prior findings that Tylenol’s apparent link to ADHD disappears when comparing siblings, though a puzzling secondary analysis produced contradictory results that even the authors couldn’t explain.
It’s confusing, but the contradictory birth-order results likely reflect some unmeasured bias within families rather than any real biological effect of the drug.
Despite the complicated sibling portion, this study ultimately adds to a now-massive body of evidence confirming Tylenol’s safety in pregnancy.