select search filters
roundups & rapid reactions
before the headlines
Fiona fox's blog

expert reaction to US study looking at reported paracetamol use during pregnancy and sleep and attentional problems in children at age 3

A study published in PLOS ONE looks at maternal use of paracetamol during pregnancy and neurobehavioral problems in offspring at 3 years.


Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:

“This is an observational study, and like all observational studies, there are unavoidable issues in understanding what actually causes what.  In this case, the question is whether the taking of paracetamol (acetaminophen) during pregnancy causes, or at least plays a role in causing, the sleep and attentional issues in some of the resulting children at age 3.  This is because there are bound to be differences between the mothers that did, or didn’t, take paracetamol, apart from the fact of whether they took paracetamol during the pregnancy.  Those other differences may be what’s known as potential confounders, and the problem is that they may turn out to be actual confounders, that is, they are themselves involved in causing the differences between the children whose mothers took or didn’t take paracetamol, and in principle they may indeed explain all the cause and effect.

“Researchers on observational studies know about this, of course, and it’s possible to make statistical adjustments that aim to allow for the differences between the paracetamol-taking and the non-paracetamol-taking mothers.  The researchers on this study did do that, for a wide range of potential confounders.  This process can’t be perfect – for example, researchers can’t adjust for a potential confounder on which they have no data, and these researchers mention a few potentially confounding factors that they couldn’t include for that reason (in the Strengths and Limitations subsection at the end of the Discussion section of their paper).  Therefore, in general, no single observational study can reliably tie down exactly what’s causing the associations that it discovered.

“There’s one aspect of the adjustments for potential covariates in this study that I do, however, want to mention specifically.  The researchers asked the participants which medications, if any, they had taken when they were pregnant, and also asked the reasons for taking each medication.  These answers were used to adjust the statistical models.  The reason for that, particularly in relation to taking paracetamol, is that a medication is generally taken because the person who takes it wants to deal with some health problem, large or small.  The study found an association between taking paracetamol and some outcomes in the children at age 3, and one possibility is that the differences in outcomes could be caused by the health reason why the mother took the medication, and not by the medication itself.  (This general issue is called “confounding by indication”, meaning that the association between the medication and the outcome is confounded, that is, confused with, the indication for taking the medication.)  Adjusting for the reasons for taking medication helps avoid this particular kind of confounding.

“But for two of the reasons for taking medication during pregnancy, the researchers report that this adjustment could not be made.  For participants who said they had taken a medication (paracetamol or something else) because they had a fever, or because they had a headache or migraine, the adjustment was not done.  The statistical reason was that these reasons for taking medication are too strongly correlated with taking paracetamol.  That is, all the women who said they had taken something for fever, had in fact taken paracetamol, and none of those who did not report taking something for fever, had taken paracetamol.  For headaches and migraines, the vast majority who reported taking something for these reasons had taken paracetamol, and very few of those who did not report a headache or migraine had not taken paracetamol.  Putting it another way, two-thirds of the women who had taken paracetamol during pregnancy also said they had taken some medication (not necessarily paracetamol) during pregnancy for a headache or migraine, and only 1% of those who had not taken paracetamol during pregnancy also said they had taken some medication for headache or migraine.

“In cases like this where there’s a very strong correlation between the outcome (the health of the children) and one or more of the potential confounders, what happens is that the statistical models can’t estimate how strong the association between the outcome and taking paracetamol actually is, having adjusted for the potential confounder, or indeed how strong the association between the outcome and the potential confounder is.  Really, the underlying reason is that knowing that a mother took paracetamol when pregnant is very similar to knowing whether they took something for a headache or migraine during pregnancy.  The statistical model doesn’t know what’s actually going on in the participant’s body, so it can’t distinguish between the association of the outcomes with paracetamol and the association of the outcomes with the mother having taken something for a headache or migraine.  The model can tell you that one, or other, or both, of these factors is associated with the outcomes in children, but it can’t tell you which of the factors is involved, the paracetamol or the potential confounder.

“It can be very awkward to deal with this kind of correlation.  But if it’s done simply by leaving out the potential confounder, as happened here, that makes difficulties in interpreting the results.  Basically, it means there is no way to tell whether the issues in the children could be caused by their mothers having taken paracetamol, or instead by their mothers having had a headache or migraine (or some underlying problem that itself caused the headaches or migraine).  The same applies to fevers, though this is probably less of a concern since only about 1% of the mothers reported having taken some medication for fever, whereas taking some medication for migraine or headache was much more common.  The upshot is that adjustments were successfully made for the other reasons that the participants gave for taking medication during pregnancy, which partly guards against the possibility that the real cause of the children’s issues was some issue of health during pregnancy rather than the paracetamol, but that these adjustments couldn’t be done for the two reasons for taking medication that were most strongly related to taking paracetamol.

“In any observational study, there always remains some doubt about the pattern of cause and effect, but I’d say this issue with headaches and migraines (particularly) does make the pattern of cause and effect even more difficult to decipher than is generally the case.”


Prof Neena Modi, Professor of Neonatal Medicine, Imperial College London, and President-Elect, European Association of Perinatal Medicine, said:

“This paper reports an association between paracetamol use in pregnancy and neurobehavioral problems in offspring.  An association between two factors should not be confused with a causal relationship; i.e. should not be taken to imply that paracetamol exposure causes neurobehavioral problems in offspring.  An association can often be explained by other factors that have not been recognised or measured.  Thus the more people carrying an umbrella is highly associated with the likelihood of rain, but carrying an umbrella does not cause rain.  One among several possible reasons underpinning the association between paracetamol in pregnancy and neurobehavioral disorders in children is maternal stress and anxiety.  In other words the cause of the neurobehavioral problems in children is maternal stress and anxiety (which led to increased paracetamol ingestion), not the paracetamol ingestion itself.  The authors recognise this but used a very weak statistical approach to try and mitigate against this possibility.  There are several stronger statistical approaches aimed at inferring causality.  The most reliable way of determining whether paracetamol causes neurobehavioral problems would be a randomised controlled trial comparing use against another analgesic.  However a major issue in drug development is that pregnant women are all too often excluded from clinical trials.  This has held back improvements in their care and that of their unborn babies, and in the confidence with which practitioners can prescribe or advise medications.  To report this paper in terms that imply a causal relationship would add to this disadvantage in irresponsibly raising unwarranted alarm among them, the public and parents.”


Prof James Dear, Professor of Clinical Pharmacology, Queen’s Medical Research Institute, University of Edinburgh, said:

“Interesting paper showing a correlation however that is not causation.  A key weakness, which the authors acknowledge, is they do not know the dose or duration of paracetamol use.  If paracetamol causes attention and sleep problems in offspring then there should be a dose-response relationship.  This paper could not test this.

“The BNF states that paracetamol in pregnancy is not known to be harmful.  NICE states that ‘paracetamol is the analgesic of choice for women who are trying to conceive or who are pregnant’.  The Royal College of Obstetrics and Gynaecology state that paracetamol is the analgesic of choice in pregnancy.  As with any medication taken during pregnancy, paracetamol should be used only as needed and at the lowest effective dose for the shortest time.  More research is needed before I think these recommendations should change.”


Prof Andrew Shennan, Professor of Obstetrics, King’s College London, said:

“The original study the data comes from was not designed to look at the problem and thus did not have accurate data on dose and timing of paracetamol given.  Although the authors attempt to control for indication, this type of exploratory study cannot rule out that women taking the drugs are different from those that don’t, and it is difficult to blame paracetamol for the findings.  5 of the 7 outcomes are reassuringly not different, including behaviours related to emotions and aggression.  It may be an exaggeration to suggest two findings are important, especially as the statistical significance are borderline and could be by chance (lower confidence levels only just meet significance by 1%).

“Paracetamol remains recommended for use in pregnancy when indicated and benefit still outweighs risks.  Avoiding treatments for fever and pain, and many alternatives to paracetamol, can be harmful.  This data is important to focus future research.”


Dr Graeme Fairchild, Reader in Developmental Psychopathology, University of Bath, said:

“The key point to emphasise is that the effect sizes or odds ratios are very small – as the paper reports, even for the strongest association, 19% of the children whose mothers did not use paracetamol during pregnancy had sleep problems, versus 23% whose mothers did use paracetamol.  So while the difference is statistically significant, the magnitude of the association is very small and would only be seen at the level of very large groups – the small differences would not be evident at the level of the individual.  The authors also find that there is a relationship between stress during pregnancy and paracetamol use which might partly explain some of the associations they attribute to paracetamol use.

“My take on the findings is that expectant mothers need not worry too much about the findings of this paper, and should continue to take paracetamol if they need to (i.e., if they have a bad headache or migraine), because there is such a small difference in rates of sleep problems and attention problems between the two groups of children.  We already know that prenatal exposure to maternal stress increases risk of mental health problems in offspring and has consequences for brain development in utero – and the very small differences between the groups in rates of sleep problems and attention problems could be largely or entirely explained by higher rates of stress in the mothers who took paracetamol during pregnancy (plus it seems they were more likely to take other forms of over the counter medication too).”



‘Maternal use of acetaminophen during pregnancy and neurobehavioral problems in offspring at 3 years: A prospective cohort study’ by Kristin K. Sznajder et al. was published in PLOS ONE at 19:00 UK time on Wednesday 28 September 2022.

DOI: 10.1371/journal.pone.0272593



Declared interests

Prof Kevin McConway: “I am a Trustee of the SMC and a member of its Advisory Committee.  My quote above is in my capacity as an independent professional statistician.”

Prof Neena Modi: “I am a clinical researcher and advocate for the right of women and babies to benefit from research.”

Prof James Dear: “I am Global Chief Investigator on the ALBATROSS Trial sponsored by Egetis Therapeutics which tests calmangafodipir as a treatment for paracetamol overdose. Due to start Jan 2023.”

Prof Andrew Shennan: “No conflicts to declare.”

Dr Graeme Fairchild: “I don’t have any conflicts of interest to declare.”

in this section

filter RoundUps by year

search by tag