Last month, the Trump administration advised pregnant mothers and the parents of young children to avoid paracetamol – known as Tylenol or acetaminophen in the US – stating that its use is associated with the development of autism. Branded as Calpol in the UK, paracetamol is every parent’s go-to when kids are unwell; it is also one of the only pain and fever medicines considered safe in pregnancy. Despite rebuttals by the US Food and Drug Administration, the World Health Organisation and the European Medicines Agency, the White House continues to cite multiple published studies that appear to vindicate its position. Unpicking this conundrum takes us to the heart of the scientific method – how it works, and how it can be misconstrued or misused.
Trump and his health secretary, Robert F Kennedy, are right: there does appear to be a small association between paracetamol use in pregnancy and rates of autism in children. Their error is to equate correlation with causation – to assume the association is a smoking gun. All such medical research is susceptible to confounding, where hidden factors affect both variables, making them appear causally connected. Consider the sales of sunglasses and cases of heatstroke. Both will rise at certain times of year: they are correlated. But we would be mightily surprised to be warned against wearing shades.
Autism is strongly genetic, so a child born to a mother with autistic traits is more likely than average to be diagnosed. Autistic traits are also associated with chronic pain syndromes such as fibromyalgia, so affected mothers may be more likely to use paracetamol as a painkiller in pregnancy. One way to resolve such issues is to analyse pairs of autistic and non-autistic siblings: being raised in the same family will level out many potential confounders. Over the past two years, large studies in Sweden and Japan have shown that the marginal association between paracetamol use and autism vanished when sibling pairs were matched.
Any half-decent scientist would have alerted Kennedy and Trump to the pitfall of confusing correlation with causation, so why the wilful ignorance? America, like the UK, has seen a recent surge in diagnoses of autistic spectrum disorder (ASD). For politicians and voters of a certain persuasion, this has become something of a moral panic. Normalising neurodiversity is seen as a facet of the reviled “woke” agenda, and it feels as though cases of autism are spiralling upwards. In fact, were ASD becoming more common we would see increasing cases of every severity, but we don’t. The surge in diagnoses is at the mildest end of the spectrum – a product of greater public awareness coupled with dramatically loosening diagnostic criteria.
In April, Kennedy pledged to identify the cause of what he termed the “autism epidemic” within five months. Bang on target, September saw paracetamol thrust squarely into the frame. Undaunted by the chorus of scientific criticism his initial assertion provoked, Kennedy has marshalled ever more tenuous research to shore up his position. An intriguing Danish study found a small correlation between male circumcision and autism. This almost certainly reflects genetic variations in some populations that favour circumcision for cultural reasons – one example: the rate of one ASD-related mutation is 250 times more common in Ashkenazi Jews. But for Kennedy, the correlation must be because paracetamol had been used for post-operative pain relief – an entirely unfounded assertion not evidenced by the study at all.
Kennedy is unwittingly demonstrating one of the greatest problems to affect the practice of science: confirmation bias – the tendency to search out and favour only information that confirms one’s prior beliefs. The reality is that he is squandering an opportunity to examine and properly understand a significant social phenomenon because he doesn’t like what the science is telling us.
[Further reading: Wes Streeting grants families full inquiry into Leeds maternity care]
Content from our partners
Related
This article appears in the 23 Oct 2025 issue of the New Statesman, Doom Loop