News · Brain & Mental Health
Sibling study found no autism or ADHD link to paracetamol
Comparing siblings whose mothers used paracetamol in one pregnancy but not the other, a JAMA Internal Medicine study of 124,333 children found the association vanished.
Based on a peer-reviewed population cohort study in JAMA Internal Medicine
- Researchers drew on 708,020 mother-child pairs in Hong Kong from 2001 to 2023, then built sibling-matched cohorts from families where paracetamol use differed between pregnancies, publishing in JAMA Internal Medicine.
- The final samples were 124,333 children for the autism analysis and 97,285 for ADHD.
- In the sibling comparison, prenatal paracetamol was not associated with autism (aHR, 1.00; 95% CI, 0.91-1.11) or ADHD (aHR, 1.01; 95% CI, 0.93-1.08).
- Nothing appeared by timing of exposure, cumulative dose, or usage pattern.
- The conventional approach, ignoring family structure, did show a positive association, matching earlier studies.
- The decisive check: use before pregnancy also appeared to raise risk (autism HR, 1.12; ADHD HR, 1.24). Exposure that ended before conception cannot cause either condition, so that signal has to be confounding.
- Exposure came from dispensing records rather than self-report, which is more reliable but misses over-the-counter purchases.
- Observational. The sibling design controls for shared family factors, not for everything that differs between two pregnancies.
Few questions in pregnancy have generated more anxiety in recent years than whether paracetamol harms the developing brain. A study in JAMA Internal Medicine went at it with a design built specifically to answer it, and the answer changed depending on how the same data were analyzed - which turns out to be the finding.
The authors set out the problem plainly. Paracetamol (acetaminophen) is the first-line analgesic and antipyretic recommended globally during pregnancy, and observational studies have reported associations with increased risks of autism spectrum disorder (ASD) and attention-deficit/hyperactivity disorder (ADHD) in offspring, raising public and clinical concern; however, these findings may be substantially confounded by unmeasured familial factors.
Comparing brothers and sisters
The usual approach compares exposed children with unexposed children and adjusts for whatever the researchers thought to measure. The trouble is that women who take more painkillers differ from those who do not, in ways that run in families and resist measurement.
So this team compared siblings instead. Starting from an initial cohort of 708,020 mother-child pairs, of whom roughly 43.3% had prenatal paracetamol exposure, they constructed sibling-matched cohorts of children from families with discordant prenatal paracetamol exposure - families where the mother used the drug in one pregnancy and not another.
That comparison holds constant the mother, the household, the income, the neighborhood and half the genome. The final cohorts comprised 124,333 children for the autism analysis and 97,285 for the ADHD analysis.
Exposure was not self-reported. It came from electronic dispensing records with data on drug name, strength, dosage, and prescription dates, which removes the recall problem that troubles much of this literature.
The result
Once siblings were compared with siblings, the association disappeared.
Comparing sibling with sibling, prenatal paracetamol exposure was not associated with the risk of ASD, nor with the risk of ADHD. Both estimates landed essentially on 1.00, meaning no difference, with narrow ranges of uncertainty around them.
The authors then went looking for a signal in the places it might hide. Null associations were consistent across exposure timing, cumulative dose, and usage patterns - sporadic, intermittent or persistent. No trimester mattered. More did not mean worse.
The check that settles it
Here is the part that makes this study more than one more entry in a contested literature.
The researchers ran the same analysis on paracetamol taken before pregnancy. That exposure ends before the child exists, so it cannot possibly affect that child’s brain.
It produced a positive association anyway. Running the same method on prepregnancy exposure for ASD and ADHD returned raised risk regardless - about 12% and 24% higher respectively, both with tight enough uncertainty to look convincing on their own.
An impossible cause produced a confident-looking result. That is not a subtle statistical point; it is a demonstration that the conventional method manufactures associations in this dataset. And the conventional method is what produced the earlier alarming findings, which this study also reproduced: positive associations were observed in conventional cohort modeling that ignored family structure.
Same data. Same drug. The answer depended entirely on whether family was accounted for.
What it still cannot tell you
This is observational, and observational studies cannot prove that something has no effect. They can only fail to find one, which is a weaker claim, though a well-powered failure in 124,333 children is a meaningful one.
The dispensing records are a strength and a limit. They capture what was prescribed, not what was swallowed, and in most places paracetamol is also sold over the counter. Some genuinely exposed pregnancies were probably counted as unexposed, which would tend to blur a real effect rather than create a false null.
The sibling design controls for what two siblings share, not for what differs between two pregnancies in the same family. Maternal age changes. So does the illness that prompted the painkiller in the first place.
And this is one health system, in Hong Kong, with its own prescribing culture.
What the authors conclude
Their summary is direct: a sibling-matched analysis found no evidence of an association between prenatal paracetamol exposure and the risk of ASD or ADHD in offspring, and the positive signals observed in conventional studies are likely attributable to residual familial confounding.
They frame the practical upshot carefully as reassurance regarding the safety of indicated paracetamol use during pregnancy. Indicated is the operative word. This is not a case for taking more of anything. It is evidence that the specific fear attached to this specific drug rested on a design that also flagged an exposure that could not have mattered.
People also ask
What is a sibling-matched design and why does it matter here?
It compares brothers and sisters from the same family where the mother used the drug in one pregnancy but not another. Siblings share half their genes, the same mother, usually the same household, income and neighborhood. Comparing them cancels out the family-level factors that a standard study can only try to adjust for statistically. If an association survives that comparison it is much harder to explain away. Here it did not survive.
Why does use before pregnancy matter so much?
It is what researchers call a negative control, and it is the sharpest thing in this paper. Paracetamol taken before a child is conceived cannot possibly affect that child's brain development. So if before-pregnancy use appears to raise autism and ADHD risk, and here it did, the method is producing an association where no causal path exists. That is direct evidence the signal comes from something about the families who use paracetamol, not from the drug.
Does this contradict the earlier studies?
It explains them rather than contradicting them. This study reproduced the positive association when it analyzed the data the conventional way, ignoring family structure. The same dataset gave a signal one way and no signal the other way. That pattern points at the method, not the medicine. Mothers who take more painkillers differ from those who do not in pain, illness, mental health and genetics, and those differences run in families.
So is paracetamol safe in pregnancy?
This is general information, not medical advice. The authors describe their findings as reassurance regarding indicated use, and the word indicated is doing work: the study looked at prescribed use for a reason, not unlimited use. Untreated fever and pain in pregnancy carry their own risks. Anyone pregnant and weighing pain relief should follow the guidance of their own clinician, who can weigh the specific situation.
What are the limits of this study?
It is observational, so it cannot prove absence of an effect, only fail to find one. Exposure came from dispensing records in Hong Kong, which capture prescriptions but not paracetamol bought over the counter, so some exposed pregnancies were probably classed as unexposed. The sibling design also cannot control for what differs between two pregnancies in the same family, such as maternal age or a specific illness. And a Hong Kong population may differ from others in prescribing patterns.
References
- Luo S, Gong Q, Ai Y, et al. Prenatal Acetaminophen (Paracetamol) Use and the Risk of Autism and/or Attention-Deficit/Hyperactivity Disorder Among Sibling-Matched Cohorts. JAMA Internal Medicine (2026).
- National Institute of Child Health and Human Development. Autism Spectrum Disorder.
- Centers for Disease Control and Prevention. Treating for Two: Medicine and Pregnancy.