On 22 September, Donald Trump urged pregnant women to “fight like hell not to take” Tylenol/Paracetamol. This article summarizes the most recent recommendations, evaluates the key studies, and offers relevant background information for counselling patients on paracetamol use in pregnancy. Let’s start.
What do health authorities recommend?
- 22 September, FDA replied: “a causal relationship has not been established”
- 23 September, EMA replied: “paracetamol remains an important option to treat pain or fever in pregnant women”
- 24 September, WHO replied: “research found no consistent association”
What’s the evidence?
Trump said, that “Tylenol during pregnancy can be associated with a very increased risk of autism. So taking Tylenol is not good. All right, I’ll say it: it’s not good.” His critics say he’s wrong. Here are the publications which influenced both sides of this debate.
Trump relies on these three publications in particular:
- In 2018, a a Systematic Review showed a 19% increased risk for autism in offspring when acetaminophen was used in pregnancy.
- In 2021, 91 experts signed a Consensus Statement in a Nature journal, which called for “precautionary action” on paracetamol in pregnancy.
- In August 2025, Andrea Baccarelli, Dean of the Harvard School of Public Health, published a Navigation Guide study, which showed an association in 27 of the 46 included studies.
Critics often cite these two sibling pairs analyses:
- In 2024, a Swedish study observed 2.5 million children. It showed no association at all when sibling pairs were matched and analyzed.
- In September 2025, a similar Japanese study showed no association as well after matching sibling pairs.
Who is right?
- Current evidence cannot prove causality. But proving small causal effects by observational studies is usually not possible. You would need large, long-term RCTs – but nobody pays for them. That’s why it’s an epidemiological debate, being “sure” is not possible.
- The critic’s main argument: small effects disappear in better studies. They argue, that the small associations reported by observational studies are mainly or entirely caused by uncontrolled confounding. That’s why the sibling pair analyses, which reduce such confounding, showed no association. That’s a strong argument.
- The main counter-argument: the Swedish study might be false-negative. Baccarelli’s study questioned the Swedish sibling pairs study, because it reported only a 5% usage during pregnancy, much lower than the 56%-63% previously reported in Sweden. This discrepancy could indeed reduce the statistical power of the study and lead to a false-negative result.
- Baccarelli’s study was highly criticised. The authors applied the GRADE system inappropriately, pooled very different studies together, downgraded sibling control analyses incorrectly due to lack of power, and applied a ‘Navigation Guide‘, which might sound fancy, but is actually quite unproven (I found only 16 such studies in the entire PubMed database). Also, while they acknowledged that „observational limitations preclude definitive causation“, they nevertheless recommended to take „immediate steps“.
- Baccarelli himself was criticised too. According to the New York Times, Dr. Andrea Baccarelli consulted Trump’s top health officials before the press conference and was also paid at least $150.000 as an expert witness in lawsuits against the manufacturer of acetaminophen. Interestingly, the lawsuits were dismissed and the judge agreed that Dr. Baccarelli “cherry-picked and misrepresented study results“ and was therefore “unreliable“.
My conclusion: There is not enough evidence to prove a causal effect, and not enough to rule it out entirely. But if there is a causal effect, it’s most likely very small.
What can we learn from this debate?
- Science is never absolutely sure. I believe the public will trust scientists more, when we communicate honestly. That’s why I believe that both, Trump and some of his critics (those who rejected the possibility of a causal relationship entirely), communicated misleadingly.
- Rare events usually cannot be proven. That’s simply not possible methodologically, or at least not affordable – because you would need very huge RCTs to identify them and to rule out confounders.
- Relative risks are misleading. Even if there would be a causal effect, how often does this happen? Trump spoke of a “very increased risk“. A hypothetical relative risk of 10% sounds a lot, but at a prevalence of around 1% in Europe, this would mean an absolute risk of 0.1% (NNH=1,000). It’s still the same risk, but communicated differently.
- Trump’s team asked the wrong questions. It’s not just about if acetaminophen causes autism, but also about the alternatives. NSAIDs are not recommended after 20 weeks of pregnancy and may slightly increase the risk of miscarriage, although this association is not statistically significant. Acetaminophen is therefore the established treatment option for fever and pain in pregnancy, and maternal hyperthermia in the first trimenon is associated with several diseases, according to Systematic Reviews:
- Neural tube defects like spina bifida or anencephaly (OR=1.9, 95% CI 1.6-2.3). The relative risk nearly doubled, the absolute risk increased from 0.27% to 0.61%.
- Congenital heart defects (OR 1.5, 95% CI 1.4-1.7).
- Oral clefts (OR 1.9, 95% CI 1.4-2.8).
- Autism (OR 1.2, 95% CI 1.0-1.3).
- Being a famous scientist is not a quality criteria. Baccarelli is the Dean of the Harvard School of Public Health. That might be useful politically, but should be ignored scientifically. There are always famous researchers in favor or against something, fame therefore doesn’t mean anything. His financial conflicts of interest and the criticism of his study are more relevant.
What should we tell our patients?
- On the Autism Link: While some studies show a small link between Paracetamol and Autism, the best studies which compare siblings did not confirm that. And even if there would be a link, it would be very small.
- On the Risk of Fever: High, untreated fever, especially in early pregnancy, is linked to birth defects like neural tube or heart conditions. Therefore, treating high fever in pregnancy is still recommended by EMA and others.
- Recommendation: The major health authorities WHO, FDA and EMA still recommend paracetamol as an important option to treat pain and fever in pregnancy. Using a low dose and short duration seems plausible.
