Vaccines, Tylenol, and Circumcision?!
JUST THE TIP of what’s wrong with these autism claims
We’ve been working on this article for weeks. Poring over studies on the Tylenol-autism topic. Just before we got it scheduled to post... another soundbite.
At today’s Cabinet meeting, Health Secretary RFK Jr. claimed that “children who are circumcised early have double the rate of autism,” adding “it’s highly likely because they’re given Tylenol.” President Trump backed him up, saying there’s “a tremendous amount of proof or evidence, I would say as a non-doctor.”
Here’s just the tip of what’s wrong with this claim: Kennedy mentioned “two studies,” but didn’t specify which. A quick search turns up two likely candidates. One is a 2015 Danish registry study that found a modest statistical bump in autism diagnoses among circumcised boys, especially under age five. The authors themselves highlighted big limitations—such as incomplete circumcision data and the inability to prove causation. (By the way, the study never mentioned Tylenol, acetaminophen, or any pain medications. The researchers were investigating whether the circumcision procedure itself - the pain and stress - might be linked to autism.) The other is a 2013 ecological paper that used circumcision rates as a stand-in for acetaminophen exposure. Even its authors described it as “hypothesis-generating,” not proof. And just to underline how shaky the circumcision link was, a 2015 published critique of the Danish study pointed out that its autism findings rested on very small numbers and questionable assumptions.
But RFK Jr.’s real logical leap is collapsing these into one neat story: some circumcised boys get Tylenol for pain, and some are later diagnosed with autism—therefore, Tylenol causes autism. That’s like saying umbrellas cause car accidents because both are more common on rainy days. The studies don’t actually connect those dots. The Danish analysis wasn’t about Tylenol at all, and the ecological paper was little more than circumcision-as-proxy guesswork.
Kennedy’s stance? “None of this is positive, but all of it is stuff we should be paying attention to.” This immediately brought to mind that scene in Dumb and Dumber—when Lloyd Christmas hears his chances with Mary are “one out of a million” and responds: “So you’re telling me there’s a chance!”
Now, on to what we actually came here to discuss: the broader claims about Tylenol and autism that have pregnant women wondering if they really need to “tough it out”...
At a press conference on September 22, the Trump administration urged pregnant women to “tough it out” and avoid Tylenol, citing a link to autism. The claim surprised many, including the researchers whose work is being cited. Scientists say that’s not what the research shows. The strongest evidence to date shows that if there is any link, it’s not a cause-and-effect relationship.
Here’s what you should know…
For starters, acetaminophen is the active ingredient in Tylenol and many other over-the-counter medications, making this warning broader than just one brand.
The Review That Doesn’t Say What They Claim
The centerpiece of the government’s claim is a review of 46 studies published in August. Yes, 27 studies reported some association between acetaminophen and neurodevelopmental issues. In science and public health, “association” means that two things are related or connected in some way, but it does not mean that one causes the other. A third thing might be influencing the relationship and causing both of the first two things to happen at the same time. The review’s own lead author, Dr. Didier Prada, confirmed: “We cannot answer the question about causation—that is very important to clarify.”
He even used a perfect analogy: ice cream sales and violent crime both rise in summer, but nobody thinks Rocky Road causes assault. (Association, not causation.)
They also relied on the Navigation Guide, a framework built for environmental toxins, not medications. With medications, we can measure exactly what people took and when. But with environmental toxins—think lead paint from childhood homes or industrial chemicals from old factories—researchers have to guess based on decades-old records. The Navigation Guide was designed for those messy situations. Applied to medications—where we have prescriptions and clear records—it treats weak hints like strong evidence. Medical standards like GRADE correctly start with more skepticism about observational studies; the Navigation Guide doesn’t. Small, questionable associations get inflated into findings that seem more certain than they are.
The review gave high ratings to studies with fundamental flaws—including one where every single baby had detectable acetaminophen in their cord blood, making a true unexposed comparison impossible—while penalizing a massive Swedish sibling study for not having trimester-specific timing data. Reviews like this are extremely sensitive to how studies are weighted—small changes in weighting can flip the overall conclusion from positive to negative. When the methodology lacks transparency about these crucial decisions, readers can’t evaluate whether the conclusion reflects the data or the authors’ choices.
Finally, the review leaned heavily on ‘dose-response’ as a sign of causation. If a drug is harmful, higher doses should mean higher risk. But most dose-response patterns came from small or poorly adjusted studies. In the large Swedish sibling study—our best evidence—the gradient vanished once family factors were controlled (as indicated by the purple line below). Low, medium, or high use all showed essentially the same risk which hovered around no effect (as indicated by the black dashed line at HR=1), suggesting the apparent dose-response patterns indicated in red, blue, and green came from confounding by genetics, maternal health, or the underlying condition being treated rather than the medicine itself.
Credibility matters, too. The senior author, Harvard Dean Andrea Baccarelli, disclosed he was a paid expert witness for plaintiffs suing acetaminophen manufacturers. When a federal judge reviewed his testimony last December, she ruled it “lacked scientific evidence“ and dismissed every case.
What The Swedish Study Shows
The administration is overlooking the most rigorous evidence available: Sweden’s sibling study, which followed essentially every child born over two decades—2.5 million in total, including 186,000 exposed to acetaminophen during pregnancy.
Looking at the population as a whole, researchers initially saw a slight association: about a 5% increased autism risk with acetaminophen use. But when they compared siblings—one exposed, one not—the link vanished. Same parents, same household, shared genes… but different exposure. The risk dropped to zero.
If acetaminophen truly caused autism, exposed siblings would show higher rates. They didn’t. As neuroscientist Sam Wang explains, that likely demonstrates the crude link was reflecting family factors—genetics, socioeconomic status, maternal illness—not the drug itself.
Critics like NIH Director Dr. Jay Bhattacharya argue that sibling studies can “wash out” real effects by over-controlling. But that misunderstands their purpose. When siblings share genes and family environment but differ in exposure, comparing them isolates the drug’s effect. If acetaminophen truly caused autism, exposed siblings would have higher rates than unexposed siblings. When that difference vanishes, it reveals the association was driven by shared family factors, not the medication itself. Moreover, suppose sibling designs truly “washed out” real effects. In that case, we’d still see them in other well-controlled studies—but the best prospective cohorts with biomarker validation show the same null results once properly adjusted.
Out of all the studies included in the review, the Swedish study did the best job of including additional variables in its statistical analysis. These variables, especially parental diagnosis of autism, could affect acetaminophen usage and risk for autism in their children. Even after accounting for more than 25 additional variables, the study found a slight increase in risk of autism with acetaminophen use during pregnancy. However, this effect disappeared when comparing siblings to each other, which suggests that unmeasured household and genetic factors—not captured by any of the studies—are what actually drive the apparent association between acetaminophen and autism.
Those 27 Studies: What They Actually Show
The government continues to cite the “27 studies” that showed a positive association in the Baccarelli study. But many weren’t about autism at all, instead looking at general behavior problems. Others relied on mothers trying to remember medications years after their child was diagnosed—an obvious setup for recall bias (mothers of children with autism naturally scrutinize their pregnancy memories more intensely, potentially “remembering” more medication use). In contrast, the Swedish study recorded acetaminophen use prospectively during pregnancy and captured prescription data through pharmacy records.
This diagram illustrates recall bias in research. When studies ask about medication use years after pregnancy, the results can be skewed because memory works differently depending on context. Some parents may review their pregnancy choices in detail (e.g., those with children diagnosed with autism), while others may be less likely to do so. This natural difference in how much people have reflected on past events can create patterns in the data that don’t represent actual cause and effect. What researchers can measure through surveys (blue) may not match the real relationship they’re trying to study (black).
The authors couldn’t actually combine the studies into a traditional meta-analysis (a statistical method that pools data from multiple studies to get a clearer picture of whether there’s a real effect). The exposure and outcome measures were too different—some asked about ‘any use,’ others counted days, some split by trimester, and one measured drug metabolites in cord blood. So they fell back on vote-counting: tallying how many studies showed an association versus no association, without considering their size, strength, or reliability. That meant a Spanish cohort of 2,600 children based on parental recall was put on the same footing as Sweden’s national sibling study of 2.5 million births with genetic controls.
Patterns are telling: when mothers reported acetaminophen use prospectively, before knowing outcomes, the associations were weaker or absent. When asked years later, after a diagnosis, the associations were stronger. That’s not causation—that’s recall bias.
And even the positive findings were modest, typically 20-30 percent increased risk. At first glance, that sounds substantial. But these numbers come from observational studies, which are notoriously prone to bias. When researchers account for family genetics, maternal health, or the reasons Tylenol was taken in the first place, the apparent increase often disappears. In other words, what looks like a sizeable association on paper may actually be a mirage created by confounding and recall bias. The few stronger signals came from tiny, self-reported studies. In contrast, the largest, best-controlled evidence—the Swedish sibling study—found no effect once family factors were considered. Most crucially, none of the studies could separate the drug itself from the reason it was taken. As Yale researcher Zeyan Liew points out, fever in pregnancy is a known risk factor for developmental delays.
The Real Dangers of “Toughing It Out”
While the president tells women to tough it out, let’s be clear about what he’s asking them to endure. Untreated fever during early pregnancy nearly doubles the risk of neural tube defects and is linked to heart defects and oral clefts. These aren’t theoretical risks—they’re established medical facts.
Many pregnant women experience fever and pain during pregnancy. What alternatives exist? NSAIDs like ibuprofen are contraindicated in pregnancy, especially after 20 weeks. Aspirin carries bleeding risks. Opioids bring obvious concerns. As one analysis warns, creating panic about the only safe option will cause real, preventable harm.
The Timeline That Doesn’t Add Up
Acetaminophen (Tylenol) entered U.S. markets in 1955, more than ten years after autism was first described by Psychiatrist Leo Kanner in 1943. If acetaminophen were a major causal driver, we’d expect a surge in the 1960s–70s as use became widespread. Instead, recorded autism rose mainly in the 1990s as diagnostic criteria broadened and diagnostic substitution occurred, especially in California. (Doctors did more strongly discourage aspirin in pregnancy by the 1990s, likely nudging some women toward Tylenol, but acetaminophen was already widely used and autism rates didn’t spike then—the increase in autism diagnoses aligns with changing diagnostic criteria.)
At the press briefing, officials pointed to a California MIND Institute study claiming those diagnostic shifts couldn’t explain the rise. But the study made a basic error: it retroactively applied 2002 criteria to children from the 1980s and 90s without re-examining them. That’s like lowering the passing score on a driving test and insisting the test hasn’t gotten easier.
California’s own Department of Health later reanalyzed the same population and found what the MIND study missed: between 1987 and 1994, autism diagnoses rose by 9.1 per 10,000 while diagnoses of intellectual disability fell by 9.3 per 10,000. The increase and decrease tracked almost exactly—classic diagnostic substitution, not a new epidemic.
The president’s claim that countries without Tylenol don’t have autism? Cuba has both acetaminophen (sold as paracetamol) and documented autism services. Lower recorded rates likely reflect underdiagnosis and reporting constraints.
The “Consensus” That Isn’t
The administration points to a 2021 statement signed by 91 scientists calling for “precautionary action.” But they recommended exactly what doctors have always said: use acetaminophen only when medically indicated, at the lowest effective dose, for the shortest time. The American College of Obstetricians and Gynecologists (ACOG) immediately responded at that time, warning against frightening women away from necessary treatment. Notably, that consensus was not based on any new research findings—it was a reinterpretation of existing observational studies that major medical organizations found unconvincing. ACOG reaffirmed in September 2025 that “acetaminophen remains the analgesic and antipyretic of choice during pregnancy” when used at the lowest effective dose for the shortest duration. Of course, the ‘lowest effective dose’ language isn’t unique to acetaminophen—it’s standard medical and legal boilerplate for virtually all medications during pregnancy. It’s good practice for everyone, pregnant or not, to take only what’s needed.
We did take note that the Swedish study reports only 7.5% of mothers used acetaminophen versus 50-65% in the U.S. While this might partly reflect measurement issues (the study asked about ‘medications’ generally without specifically prompting about acetaminophen), the sibling analyses still found no increased risk among those identified as users. The stark difference between Swedish and American usage rates—whether real or artifactual—does raise questions about pain management during pregnancy. Are Americans quicker to reach for medication? That’s a conversation worth having about judicious use. But it’s separate from whether acetaminophen causes autism. The Swedish sibling analyses found no increased risk, suggesting the answer to that specific question is clear.
What We Can Actually Say With Confidence
After examining all the evidence: If acetaminophen has any effect on autism risk—and that’s an enormous if—it would be so small we can’t detect it in millions of children. It would be one tiny factor among many in a condition that’s 80-90% heritable, with over 100 genes identified as contributors.
The government’s transformation of these uncertainties into absolute declarations isn’t just bad science—it’s dangerous. Pregnant women will suffer through treatable fevers and pain. Some will experience preventable complications. All based on a certainty that doesn’t exist in the data.
A Message to Mothers
If you’re pregnant and scared, or if you’re a mother wondering if that Tylenol you took somehow “caused” your child’s autism: The premise of this entire debate—that autism is something to be prevented or blamed on someone—is fundamentally wrong.
The best evidence shows no causal link. Can I promise with 100% certainty that acetaminophen never affects neurodevelopment? No scientist can promise that about anything.
Scientists use careful language like “no evidence of harm,” not because we haven’t studied it properly, but because we can’t ethically do the randomized experiments that would prove causation. We must rely on observational data, which is inherently less definitive. And we can’t study every possible combination of dose, timing, genetics, and environmental factors. While we can’t scientifically declare “Tylenol definitely doesn’t cause autism” (proving a negative is nearly impossible), the evidence strongly suggests that for the vast majority of people, acetaminophen use during pregnancy doesn’t meaningfully increase autism risk.
The desperate search for environmental “causes” of autism distracts from what really matters—ensuring autistic individuals get the support, accommodation, and acceptance they need. Your child’s neurology is not your fault, not a tragedy, and not something that needs to be fixed.
The truth is less dramatic than the president’s announcement: there’s no credible evidence linking acetaminophen to autism. The real harm isn’t in taking necessary medication. It’s in the suffering that will result from transforming uncertainty into false certainty, correlation into causation, and weak associations into federal mandates—leaving pregnant women to endure treatable pain and fever based on a scientific conclusion that doesn’t exist.
Stay Curious,
Unbiased Science
Want to support our work? The best way is to subscribe to our Substack and share our content. While all our articles are always completely free to read, paid subscriptions help sustain our in-depth reporting on public health and science topics. Thank you for considering it!
 | A guest post by
|
 | A guest post by
|
 | A guest post by
|
Brilliant article, thank you for elaborating on these points for us. Especially in regards to the studies that were cited and the overall evidence! It's so important to note that when you do look at the studies that had findings of an increased risk, the numbers are incredibly small. Even the Swedish study - before sibling controls, the numbers were something like an increase of 1 per 1000 for autism and 1 per 10,000 for ADHD. Hardly a contributor to the (supposed) "explosion" in rates.
Like...what? I just...WHAT? I guess they're not blaming vaccines anymore? I don't understand anything anymore. Except if RFK.says something, do.the opposite.