Today, my colleague Dr. Michael Barnett and I published new research in The Lancet, in which we measured changes in prescribing patterns in the United States after President Trump and members of his administration made bold and unsupported claims regarding the causes and treatment of autism at a White House press event on September 22, 2025.
We went live this evening to discuss our paper. We got a lot of media coverage here, with some unusually thoughtful pieces that I was so pleased to see and contribute to.
The New York Times by Azeen Ghorayshi.
The Washington Post by Lauren Weber and Ariana Eunjung Cha.
CNN by Brenda Goodman.
The Boston Globe by Lauren Albano.
NPR by Sydney Lupkin (featured on All Things Considered on Thursday and Morning Edition on Friday).
Plus, Scientific American, Katie Couric Media, Dr. Céline Gounder’s Substack, STAT News, and more.
The process is the power.
Before I get into our findings, I’m going to break the fourth wall here and say something about our scientific process. Our research looks simple enough. A couple of graphs. A couple of pages of text. But that’s the tip of the iceberg. Getting this work down to such a compact manuscript took many months, requiring countless hours of detailed, careful work. We started this back in November. As recently as a few days ago, we were still quadruple checking everything, making sure it was ready for the public. The most important thing we did was look at the scientific questions in a number of ways. We wanted to make sure we could not possibly be wrong. That is what matters: the process. While we didn’t show this in the final paper, we “asked” the data the same questions in several ways to make sure that the answers would always come out the same. (The peer reviewers wanted to see some of that, which is a sign that they were paying attention.)
That, I think, is what distinguishes good science from bad: a willingness to be your own biggest skeptic. We are not high on our own supply. When we get an idea, we look for reliable data that might help us gain novel insights. Once we have the data, we interrogate it rigorously. We don’t stop until we know for sure that we can be confident in the findings. That’s what makes our “side” (if sides can be said to exist) more likely to be correct. In all honesty, I simply don’t think that this essential part of the scientific process is important to the physicians and scientists—and whatever Secretary Robert F. Kennedy Jr. is—who President Trump has appointed to various leadership positions this term. They only seem to care about making a claim and declaring victory. They are allergic, it seems, to any interrogation of their methods or process, such as it is. They talk about transparency, but they do not engage in it. They must believe that this reads as confident. It does not. It reads as amateurish bluster, which ultimately reads as flimsy. While it must be nice to just sorta feel like you know all the answers, deep down, they must know they are profoundly uncertain. If they held themselves to anything like high standards, they would never have gone out in public and made such claims. Holding them to account is far more difficult for us than it was for them to make their claims—but we believe we spent our time wisely here.
The science.
So….let’s get to the study.1
When the White House talks about medicine, doctors and patients listen. Should they?
On September 22, 2025, President Trump, HHS Secretary Robert F. Kennedy Jr., and other officials held a White House event where they claimed that taking Tylenol (generic name: acetaminophen in the US, paracetamol in the UK) during pregnancy increased the risk of autism in children. At the same event, a drug called leucovorin was highlighted as a potential breakthrough in the treatment of autism.
The problem is that the best science does not suggest this linkage between acetaminophen use and later rates of autism. President Trump and his team cherry-picked science purporting to show that link, while ignoring stronger evidence that refutes it. In the case of leucovorin, a drug that can treat a condition called cerebral folate deficiency, there is simply no good evidence to suggest that it will help most children with autism or related conditions.
My colleague, Michael Barnett, and I wanted to know if this event might have single-handedly influenced how medicine is practiced in the United States. So, we accessed the Epic database, which now covers a majority of American healthcare. We looked at how often pregnant patients received Tylenol in emergency departments before and after the White House press event. The results were clear, as we just reported in a new manuscript published in the medical journal The Lancet.
Here’s what we found. There was an immediate decrease in Tylenol use in ERs among pregnant women (but not among non-pregnant women). We also looked at two other treatments that are offered in ERs: opioids and IV fluids. There were no changes in how often those were given. This means that thousands of women did not have their pain or fever treated in ERs, likely because they were needlessly afraid. We think that’s unfortunate because, among the options for pain control and fever reduction, Tylenol is the safest option. It’s sad that thousands of people had to suffer for no good reason. At one point, Tylenol orders decreased by 20% compared to usual rates (overall, we found a 10% decrease during the study, which stretched from late September to early December).
Here’s an animated version of our findings on acetaminophen (Tylenol). Note that when the lines split (after September 22, 2025, the date of the White House press event), the turquoise lines represent “expected” rates of the treatments we studied. You can see that acetaminophen orders fell (rather immediately), while the others had no change.
We also did a control experiment, looking at non-pregnant females of the same ages. Here, there was no change, indicating that the sudden drop in acetaminophen orders in ERs was limited to the patient population targeted by President Trump’s press event.
We then looked at millions of pediatric clinic visits to see if rates of leucovorin prescriptions changed among children ages 5 to 17. The effect here was massive. Only leucovorin showed a major change. At one point, new leucovorin prescriptions more than doubled, and over the entire study period, the rate increased by 71% compared to what would have normally been expected during the same time.
If you look carefully, you’ll note that in this part of the study, we had three controls: aripiprazole, risperidone, and folic acid. We chose the first two because they’re the only other medications that have FDA-approved indications for treating symptoms related to autism. Folic acid was chosen because leucovorin is folinic acid, so they are somewhat related compounds. As in the first graphic, when lines split on September 22, 2025, the turquoise lines represent prescription rates (per 100,000 visits) that our model predicted would have occured, absent the White House press event.
You also might notice that towards the end of the study, aripiprazole and risperidone prescriptions increased a bit. We’re not sure why, but we think it is because a new leucovorin shortage occurred around that time.
In a time in which public trust in health, medicine, and science is under attack, it’s regrettable that so many families will have been misled into thinking that this medication could somehow miraculously change the lives of children with autism. At the same time, a leucovorin shortage has now occurred, which is problematic because leucovorin is a critical component of several chemotherapy regimens.
The implications.
It’s important for people to know that it often takes years and even decades for very high-quality research to reach patients. But President Trump and his team were able to change medical practice virtually overnight, based on misunderstandings of the science around this topic.
Nor is this harmless. Pregnant women need not suffer for want of safe pain medication. Letting fevers go untreated in pregnancy has its own risks, as mentioned. Meanwhile, families need not be misled into thinking that there’s some miracle cure for autism. Moreover, there’s no medium- or long-term data on the safety of the dose of leucovorin being suggested for children with autism. Not to mention, the possibility of delayed chemotherapy.
Indeed, at the end of the day, people still look to the White House for guidance. The administration could have used that opportunity to promote the highest quality science—perhaps using that lectern to promote vaccinations against measles and other preventable diseases—rather than what happened here.
Humility demands due diligence, not hasty policymaking via a White House press conference.
One more thing. This was an unfunded study. That means that, in a real sense, Inside Medicine made this work possible. Thank you for reading this!
