24/09/2025
***UPDATED to try to simplify explanations at the request of readership. If you have any other questions about what is written in this post, please send me a message.***
It has been a while since I have posted about literature misinterpretations, however, after Donald Trump stated that paracetamol is linked to autism, I felt I needed to. This statement comes after a study funded by the US National Institute of Health (NIH) was published in the journal 'Environmental Health' in August 2025. Rightfully, RANZCOG and the AMA were quick to condemn this statement, but did not go into detail about why this study was flawed.
This study 'reviewed' several studies that have assessed any link between paracetamol (acetaminophen/Panadol/Tylenol) and autism and ADHD. It used a method described as the 'Navigation Guide', a method used in environmental exposure studies to combine data from several studies. This method is supposed to minimise differences between studies (heterogeneity) by individually assessing each study against a standardised rating assessment for bias. This is, perhaps, the only 'strength' I could identify in this study.
Now for the issues. Apart from the fact that this study was funded by the NIH, was published in an Environmental health journal, one author publishes widely on the risks of paracetamol and neurodevelopmental conditions and the authors even state that 'no datasets were generated or analysed during the current study', there are other concerns also.
A metaanalysis is a study that combines other, smaller, studies to increase the total number of participants. A metanalysis of well-designed randomised controlled trials is widely considered the 'top level' of evidence. This study was not a metaanalysis and did not follow worldwide standards of metaanalyses (PRISMA). This meant that the study itself does not present any new data. In fact, it does not even quantify associations. Using the ‘Navigation Guide’ it has merely ranked studies based on what the authors have assessed as being high or low risk of bias. It does not address issues with included studies apart from bias. While bias is important, it is not the only important thing about a study. This study does not weight results based on sample size (the number of included subjects) or study design (the way a study is conducted) and does not account for heterogeneity (differences between included studies) which is what a proper metanalysis would do. In pooled studies (where lots of smaller studies are combined) such as this, the study outcomes are only as rigorous as the studies included. In other words, if you combine lots of poor-quality studies, your final outcome will be of poor quality. Indeed, some of the included studies relied on parental reports of ADHD and ASD diagnoses (which may or may not have been professional diagnoses!). This analysis does not assess this any further than risk of bias, and even the assessment of risk of bias is fraught with subjectivity. Far from being ‘transparent’, which is what the authors state, this is anything but.
As such, causation can definitely NOT be shown in this study.
If you would like further in-depth comments about other concerns I noted around the authors justifications for limitations let me know. This includes: opting against a metaanalysis because of ‘significant heterogeneity...outcome measures, and confounder adjustments’; ignoring the highly likely confounding effects of other exposures (eg. Preterm birth) because they go ‘beyond the current analysis’; and using the relative number of positive studies to make inferences about causation (ignoring the risk of publication bias - where studies that find a difference are more likely to be published than negative studies). The authors also down-ranked the largest population based study of almost 2.5 million children from Sweden (Alqvist et al 2024) which found no association between paracetamol and autism. This would sway their conclusion more in favour of a positive association.
SUMMARY
The article on which Trump based his comments was funded by the NIH. It was not a proper metaanalysis and did not provide any new data, risk estimates or confidence intervals. Its methods are not as transparent or objective as stated by the authors. The lack of new data, at the authors’ own admission, highlights this. Finally, a previous very large study has shown that there is no association between paracetamol and autism.
There is no evidence, none whatsoever, that shows a causative link between paracetamol and autism. If you are pregnant at any gestation and you need paracetamol for pain relief or to reduce your fever, please take it!
