How often does one science team prove that another science team is completely wrong?
How often does that proof possibly affect the lives of tens of thousands of people for their entire lifespan?
University of North Carolina, Northern Kentucky University, and WPLab, a private, non-profit research organization, have just published a study which shows that we do NOT know if acetaminophen use during pregnancy causes autism spectrum disorder (autism) or other neurodevelopmental problems such as ADHD. This contradicts a widely publicized study from Drexel University and published in JAMA in April of 2024 claiming that no real association between acetaminophen use during pregnancy and autism exists.
A summary of the JAMA study
The JAMA study, like many studies before it, found a strong association between acetaminophen use during pregnancy and autism. For example, with a “medium” dose of acetaminophen of between 166 and 429 mg/day during pregnancy, they saw more than a 50% greater risk of autism that was highly significant. With heavier use of acetaminophen, the risk was almost 90% greater, or close to double the risk of autism.
Unfortunately, most studies looking at the connection of acetaminophen exposure and developmental disorders have made a common mistake in confusing confounding factors with cofactors.
An important clarification of terms before continuing:
Confounding factor: a variable that interferes with accurately assessing a connection between two different things (For example: if we think factor A might cause outcome B, but in reality factor C causes both A and B, then factor C is a confounding factor)
Cofactor: a variable that plays a key role in the connection between two different things (For example: if factor A causes outcome B, but only if C is also present, then factor C is a cofactor)
The Drexel group adjusted their analysis for numerous factors that they believed were confounding. In these types of investigations, teams should try to identify and compensate for confounding factors when conducting their statistical analyses. But the Drexel group, like most groups before it studying the connection between acetaminophen and autism, confused confounding factors with cofactors.
After the adjustments, the risk of autism associated with acetaminophen exposure was reduced to a much lower level. For example, after adjustment, a medium level of exposure to acetaminophen was associated with only an 11% greater risk of autism compared to no exposure.
Next, the investigators at Drexel took their analysis one step further than anybody else had ever done before: they adjusted the analysis for sibling pairs, which adjusts for genetics and other factors shared by siblings. This caused the association between acetaminophen and autism to disappear completely.
Because of this additional step, the conclusion they reached was that the connection between autism and acetaminophen exposure during pregnancy is not real, but rather due strictly to confounding factors. The conclusion was published in a prestigious journal called the Journal of the American Medical Association (JAMA) and brought to the media’s attention through a press release by the National Institute of Health (NIH), which hit the general public like a shockwave, now dominating search engine results about acetaminophen safety during pregnancy.
However, in a recent turn of events, the claims made in the JAMA paper were refuted by a new study published in Life. The investigative group, led by scientist William Parker, claims that the Drexel University analysis shows only what is already known: if acetaminophen use during pregnancy causes autism, it does not do so alone. Acetaminophen acts together with numerous cofactors to cause neurodevelopmental disorders.
What we know about the connection between acetaminophen and autism
The prevailing theory in the current scientific literature is that a host of genetic and environmental factors lead to autism. Parker and colleagues have previously concluded that this host of factors makes some babies and young children sensitive to acetaminophen-induced neurological injury, leading to many if not most cases of autism. However, it remains unknown if the problem can occur during pregnancy.
Parker and colleagues have previously published several papers on the connection between acetaminophen and autism in peer-reviewed journals.
The most current research points to literally two dozen lines of evidence allowing Parker and colleagues to conclude that acetaminophen exposure, together with a range of genetic and environmental factors, leads to many if not most cases of autism.
For more information on what the current research says, please visit preventautism.org.
How the new study published in Life proved the conclusions of the JAMA paper wrong
The Life study takes an unusual approach in highlighting a critical error in the Drexel University study. Parker’s group started by creating a computer-generated population of 120,000 children in which 50% of all cases of autism were caused by acetaminophen exposure plus a combination of genetic and environmental factors. The other 50% of autism was caused by random, unknown factors.
Then, Parker’s group analyzed their simulated population of children using the same analysis method run by the Drexel University group.
The computer told Parker’s group that acetaminophen was not important in the induction of autism. In fact, in one case, when the percent of individuals using acetaminophen approached 60% in the simulated population, the computer told them that acetaminophen actually protected people from autism.
What happened? How could the computer say that the drug that caused the problem was protecting people from the problem?
This conundrum is caused by a simple and well-known mistake in statistical analysis. If the computer is told that cofactors are confounding factors, the computer can provide a very misleading answer. The computer works just fine, but it is not designed to detect bad assumptions, and it’s going to do exactly what it’s told. Most of the factors that were treated as confounding factors in the JAMA paper are known to be either cofactors for acetaminophen-mediated injury or associated with cofactors for acetaminophen mediated injury.
Let’s break that down even further. Basically, it comes down to telling the computer to look at “or” instead of “and.” The Drexel group asked the computer to determine if it was multiple factors OR acetaminophen that caused neurodevelopmental disorders. Instead, they should have asked if it was multiple factors AND acetaminophen that caused the problem. This is where the JAMA study went wrong.
What is known about pregnancy and acetaminophen
As previously stated, Parker’s and colleagues have concluded without reasonable doubt that acetaminophen use in babies and small children combined with numerous susceptibility factors is responsible for many if not most cases of autism. One part that remains uncertain, however, is whether acetaminophen exposure during pregnancy carries any risks.
Labor and delivery are the exceptions to the uncertainty regarding risk during pregnancy. Based on published science, if the mother is given acetaminophen during labor and delivery (which is essentially useless anyway since acetaminophen does not provide adequate relief for the pain and discomfort of labor), and if the umbilical cord is clamped before the drug has cleared her body, then the baby has to process the rest of that drug on its own.
The initial days after birth are almost certainly the time of greatest risk for acetaminophen-induced chemical injury, and that risk extends out to about 5 years of age.
Unfortunately, whether it’s dangerous during pregnancy remains unknown.
A note from WPLab: With all of this uncertainty, doctors and experts are faced with a dilemma: what do we recommend for use during pregnancy? Sadly, there is no clear answer regarding what the standard of care should be. We at WPLab believe that it's important to inform parents and guardians of what we know and don't know. We also trust women to decide what is best for them during their pregnancies. We have additional resources available for current and soon-to-be parents and guardian available on our website, and would encourage anyone to explore further.
The study by Parker’s group was published in Life (July 23, 2024) and conducted in collaboration with neuroscientists Kate Reissner at the University of North Carolina and Lauren Williamson at Northern Kentucky University. WPLab members on the study were John P Jones III, Zacharoula Konsoula, and Rachel Anderson.
As a scientist with almost 30 years of experience, William Parker is the CEO of WPLab Inc., a private, non-profit research and education organization, and works as a Visiting Scholar at the University of North Carolina in the Department of Psychology and Neuroscience. He is most well-known for discovering the function of the human appendix as a “safe house” for beneficial bacteria, for pioneering work probing immune function in wild animals, and for making discoveries about the benefits of intestinal worms for mental health.
For more information, please contact William.parker@williamparkerlab.org