Thinking About Ob/Gyn
A fresh and evidence-based perspective of all things related to obstetrics and gynecology. Follow us on Instagram @thinkingaboutobgyn or visit thinkingaboutobgyn.com for show notes and more.
Thinking About Ob/Gyn
10.7 Tylenol and Autism
We push back on claims that Tylenol or vaccines cause autism and explain how weak methods, conflicts of interest, and cherry-picked data fuel public panic. We also unpack why diagnoses have risen—broad criteria, screening, and access—not because of a new environmental villain.
• Summary of claims made at the press event and why they fail
• What the cited acetaminophen paper did and didn’t show
• Conflicts of interest, pay-to-publish venues, and bias
• Why correlation isn’t causation; confounding by indication
• Bradford Hill criteria applied to acetaminophen and autism
• Sibling-controlled studies as the strongest current evidence
• Amish and Cuba myths; diagnosis versus true prevalence
• DSM-5 changes driving higher autism diagnoses
• State-by-state variation explained by services and funding
• Vaccine safety evidence contrasted with myths
• Practical counseling: treat fever; use clear, strong evidence
Be sure to check out thinking about obgyn.com for more information and be sure to follow us on Instagram
0:00 Setting The Record Straight
2:30 The Press Conference Claims
5:30 Tylenol, Vaccines, And Autism
9:30 The Study Behind The Hype
14:30 Conflicts, Bias, And Bad Methods
19:30 Correlation Isn’t Causation
23:00 Bradford Hill 101
28:30 Amish, Cuba, And Diagnosis Rates
33:30 Screening Tools And Subjectivity
37:30 Sibling Studies: The Strongest Signal
42:00 Why Meta-Analyses Can Mislead
46:00 What The “Navigation Guide” Misses
51:00 Vaccine Myths In Perspective
54:00 Why Autism Diagnoses Rise
59:00 DSM-5 And Access To Services
Follow us on Instagram @thinkingaboutobgyn.
Welcome to Thinking About OBGYN. Today's episode features Howard Harrell and Antonia Roberts discussing Tylenol and Autism.
SPEAKER_01:Howard?
SPEAKER_02:Antonia.
SPEAKER_01:What are we thinking about on today's episode?
SPEAKER_02:Well, I don't know how much news you watch, but I think we have to talk about autism and Tylenol and maybe vaccines.
SPEAKER_01:Yeah, I try to stay away, but I think this is an unavoidable news item, and I'm very ready to discuss it. So yeah, let's just get straight in. We normally start with a little segment, the thing we do without evidence on this podcast. But I think in a way, this whole Tylenol announcement by our president is that telling women that Tylenol is not good and tough it out when you have a fever is definitely a thing without evidence. And also saying that places without Tylenol or vaccines don't have autism is also a blatant falsehood. That was part of that same announcement. And I think we'll get into there is a reference to Amish people not having autism, which is not true. I won't get into how culture and school attendance and stuff play into autism diagnosis in a way that might not be able to capture all cases in all kinds of communities. There was also a reference to Cuba not having autism because they don't have Tylenol apparently, but that's also not true. They have over 3,000 confirmed registered cases. Anyway, we know historically before vaccines and Tylenol ever existed or certainly ever became mainstream, autism was around. It was just called many other things. Schizophrenia was often used at one point, and then different types of developmental disorder and some other more outdated terms. Well I think we'll get into this later, but yeah, that's why it wasn't called autism before. So that's why there were so few cases of autism in the past. Although these were these other things that they called schizophrenia were very well described and very well matched up with what we would call autism today. So anyway, I'm probably getting ahead of myself here. There is a lot potentially to discuss, and a lot of patients now, especially pregnant patients, are confused and concerned. And a lot of our colleagues are wondering how to combat this misinformation about this and many other topics that are now coming from so high up the chain. So we'll discuss that and much more today.
SPEAKER_02:Well, you set the picture here. So speaking at the White House, what is now two Mondays ago, I guess, Mr. Trump claimed that the use of acetaminophen, the active ingredient in Tylenol by pregnant women, has contributed to the increase in autism cases. He urged expectant mothers, as you said, to fight like hell to avoid the medication. And he reiterated the long discredited theory that vaccines and their ingredients, such as mercury and aluminum, are a cause of autism or a contributor at least. He specifically advocated for administering the measles, mump and rebella vaccine as three separate injections rather than a single combined shot, a practice which is certainly not supported by any medical evidence. And he made the exaggerated claim that children receive a VAT of eighty different vaccines. Mr. Kennedy, who has a history of promoting anti-vaccine theories, echoed and expanded upon these claims. He attributed the rise in autism to environmental toxins and asserted that the National Institute of Health has been engaging in fruitless research into the causes of the condition, especially highlighting potential genetic causes, which we've learned a lot about over the last couple of decades. He stated that his joint effort with Mr. Trump would investigate the quote ideology of the autism outbreak, with a focus on vaccines as a potential factor. And together, Mr. Trump and Kennedy also announced a plan to investigate the drug Leucovorin as a possible treatment for autism. The statements made during this press conference, they've of course drawn swift and strong condemnation from a wide range of medical organizations and experts. They emphasize that numerous large-scale scientific studies have found no link between vaccines and autism. And the theory that the MMR vaccine in particular causes autism originated from a fraudulent and now retracted 1998 study, which we probably don't have time to talk about today.
SPEAKER_01:Yeah, that's a good classic. This whole press conference about Tylenol and autism honestly was surreal. It's also interesting how many of the people there quickly backtracked from the tone of the conference. So Dr. Oz, who's now a government official, but also known as the TV doctor, who has a history of making false medical claims for money, went on a podcast after this press conference and emphasized that really there's no evidence that Tylenol caused autism, only that some studies showed an association. And then Marty McCary, the FDA commissioner, sent out a letter on September 22nd, 2025 to physicians about acetaminophen. And here he was careful to point out that a causal relationship has not been established, and there are contrary studies in the scientific literature. And it also points out that it's important to treat fevers in pregnancy and that acetaminophen is still the safest over-the-counter alternative in pregnancy among all analgesics and antiparetics. So that's just a dramatically different statement and different tone than what was presented in the press conference. It seems maybe like Dr. Oz and Dr. McCary are trying to maintain some credibility with physicians by writing statements that are technically true and therefore allowing themselves some plausible deniability in the medical community and perhaps in the legal community too. But Mr. Trump and Mr. Kennedy went all in with this message that we have discovered the cause of autism, and that is Tylenol andor the MMR vaccine andor vaccines in general. And of course, we've already addressed this Tylenol question on this very podcast here before. But the question is, does Tylenol cause autism or ADHD? In episode 9.8, we highlighted a Green Journal article from February 2025 that concluded that there was not data to support a causal relationship between Tylenol use in pregnancy and the kid having ADHD or autism spectrum disorder. But there are these retrospective low-quality studies that come out occasionally which show potential correlations. And it seems like everyone continuously forgets that correlate correlation does not equal causation. As much as everyone knows that is true, we all make this mistake every day in our lives, and our literature is full of it, and our medical practices are full of it too, and we need to clean this up. It's really not that meaningful that two things are correlated statistically with each other. And in fact, correlation almost never implies causation. But there was a recent study we'll bring up here that was cited during this press conference, and it's been making the rounds, and some people out there are excited that this is the smoking gun for those who want to blame Tylenol for autism, as best I can tell. So why don't you get into that study?
SPEAKER_02:Yeah, so that's the heart of this press conference. So there was a paper published this year in a low-quality, open access, pay-to-publish journal called Environmental Health. And it was titled Evaluation of the Evidence on Acetaminophane Use and Neurodevelopmental Disorders Using the Navigation Guide Methodology. What you've probably heard about this, if you've seen follow-up coverage, is that the senior author of this paper is a man named Andrea Baccarelli, who is the dean of the Harvard School of Public Health.
SPEAKER_01:Yes, and I think obviously the association with Harvard with this paper is what gives it some credence in a lot of people's eyes.
SPEAKER_02:Well, a lot of folks like him, this dean, are in the business of providing paid testimony for product liability lawsuits. And I've written, and we've talked before, about some of these in the epidemiology business, I'll call it, who look for get rich quick cash through product liability or drug liability lawsuits. So in fact, that describes the whole career of Bobby Kennedy Jr. That's what he does for a living. He's an attorney who does this for a living. But this Baccarelli was paid$150,000 as part of a lawsuit against acetaminophon. And for that$150,000, he claimed that it can cause offspring to develop neurodevelopmental disorders such as autism and ADHD. He said this in a 2023 and written testimony, whereas the year before, in 2022, he co-authored a study on acetaminophil that warned against any change in clinical practice and that there really wasn't any evidence of causation.
SPEAKER_01:In other words, he directly contradicted himself in a short amount of time.
SPEAKER_02:Well, yes, and it seems like he was given an opportunity to make$150,000 and he said, forget the facts, forget this paper I just wrote last year, and he went for it. I'm sure there are many life situations and financial woes where you'd do some surprising things to suddenly get$150,000. And who knows, maybe he even believes it, but of course, even more money would have followed had these cases succeeded in court, and he was paid in a variety of lawsuits to testify in in prolonged court cases. That was just for written testimony. But the judge in that particular case looked at all the contradictory statements from back before and the fact that he had never published or performed any study saying that there was a causal link with acetaminophant with neurodevelopmental disorders and any other issues. And ultimately he was actually excluded as an expert witness in this case. He wasn't qualified by the judge. Then in 2024, he became the dean of this college at Harvard. And now this year, this paper comes out, perhaps I would assume, in an attempt to get himself reinstated as an expert witness, so he can get more of that expert witness money. But even this paper that supposedly blames Tylenol for autism explicitly states that they have not discovered a causal relationship.
SPEAKER_01:Yeah, we've talked a lot about the role of bias in scientific publications, and it's no accident that this paper was published in a low-quality pay-to-publish journal, not JAMA or ACOG or The Lancet, for example. Dr. Baccarelli does list in the very back of this paper on page 40 that he has a conflict of interest because he served as an expert witness for a plaintiff team trying to prove causation of acetaminophen. But he holds out that despite the$150,000 he got, he can still remain objective. But it's interesting again because this paper and his other publications say they have not established causation yet in court. He took 150k to say that there was causation. So there seems to be a significant ethical breach here. This paper, by the way, was published in the middle of August 2025 after what appeared to be a fairly close working relationship between Mr. Baccarelli and Mr. RFK Jr. And he's part of this group of plaintiffs' attorneys who, like Kennedy, are trying to sue Acetaminifin and various vaccine makers. So if that's not a serious conflict of interest, I don't know what is.
SPEAKER_02:Right. And to the point who he was even excluded by the judge because of these very obvious issues and by his peers for this paper in particular, he's been heavily criticized for methodologic issues. But let's emphasize that in this paper, even despite all that, they conclude that the data they analyzed was not robust enough to be subject to meta-analysis. They also decided not to give weight to appropriately controlled sibling studies, sibling pair studies, which we'll talk about in a second. So in other words, they excluded important studies, or at least gave less weight to them or reanalyzed them, the studies that have the best methodology for answering this question. And then they used statistical manipulation of these other low quality studies. And despite all of that, in the end, they conclude that they could not establish causation, but perhaps an association between exposure prenatally to acetaminophen and offspring with autism spectrum disorder or ADHD. So I'll go so far as to say that this is one of the worst papers I've ever seen and is clearly dishonest in its efforts to exclude gold standard evidence.
SPEAKER_01:There's two issues at play in this question. So the first one is there an actual association, even between using acetaminophen in pregnancy and the offspring getting ADHD or autism. When you hear that the studies are mixed, and some studies found no association, and other studies maybe did find an association, that's in reference to this issue of whether or not there even is an a statistical association. There already is no scientific studies implying causation. So we're debating whether or not there's even a true association. Now, if there were, then you would still have to do the work to establish causation. And recall that the vast majority of all things that are statistically correlated don't have any kind of causal relationship. The literature really is full of associations that turn out to be nothing coincidence or maybe an underlying confounder when we attempt to establish causation.
SPEAKER_02:And everybody knows correlation doesn't equal causation. But again, like you said before, no one lives their life that way. People want it to mean something and it doesn't, and we have to be rigorous. Association can be due to all sorts of lurking or unknown or common variables between the two issues, some confounder. It could be that there is no association at all between acetominophine use and autism, and that seems to be what's actually true. We'll get into that. Or if there were a small association, well, it could be due to some common etiology, something that's associated both with acetaminophine use and with autism. Usually these issues are just statistical fallacies, they're just noise in the data and not true signal, especially when you have to go to so much work for such a common condition to establish that. Tylenol is probably the most commonly used medication during pregnancy, other than like a prenatal vitamin. And so if you're looking in the charts of children with autism, you may see a very detailed history for that pregnancy that includes all medication exposures, but your comparator group for children without autism doesn't include all these detailed exposures. Nobody cared enough to put all that history in there. So the things that are very commonly present in our society or in the environment or on the med list of a pregnant woman are going to be the ones that are subject to many of these false positive associations. But in order to establish causation, you've got to do a lot more. Typically, we'll apply the Bradford Hill criteria as a starting point, really, to try to understand the likelihood that there is a true causation relationship between associated factors. So these Bradford Hill criteria are a set of nine factors that epidemiologists use as a framework to try to understand if an observed association, once you've decided that there really is one, between some exposure and some outcome has a causal link. So this includes the strength of the association. And in this case, we're not even sure there is an association. And if there is one, it's a very weak, hard-to-determine association.
SPEAKER_01:Yeah, and that's important because in the press conference, Trump did say that autism is present in up to one in twelve boys, which is so pre prevalent that it should be mathematically very possible to determine a strong positive association with a causative agent and an outcome that that happens so commonly. It's harder to do that with the more rare diseases like congenital heart defects, for example. But we're talking about one in twelve. So the fact that most studies say there's no association, few studies find the subtle possibility of one isn't in any way proof of about there being an association.
SPEAKER_02:That's right. And we'll get back to this sibling studies in a second. But first let's talk about the strength of the association and the consistency. Meaning, do we see this association in different populations using different study designs? And the answer here is no. So that's the first of the Braffertil criteria. And then the next we look at specificity, meaning that we should see it with a very specific outcome, not something with a wide range of outcomes. And that's an inherent problem, honestly, with autism spectrum disorder, because there's so much variation. It's a narrow, nonspecific association, even if it is one, because you know, autism means so many different things. So it's a nonspecific association, even if there is one. Next is what's called temporality, and that's harder here because we're talking about a medicine taken while the woman's pregnant, and then a diagnosis in the child made after the age of two. And of course, there's so many other things that happen in that time period as well. That's why people commonly look at vaccines because of this sort of lazy temporal association. But as you already said, autism itself has been around way longer than acetaminophen. And so that's another argument actually against temporality. The next criteria is dose response, and the data here too is lacking. This would mean that Tylenol, if it was used very commonly throughout the pregnancy, should be associated with a very strong signal of autism compared to someone who maybe used it once or twice, even if there was a causal relationship. Some of the studies make this claim, but there's not an actual dose response gradient, so you don't get credit for that either when you're looking for causality. The next criteria is plausibility, and our current understanding of the causes of autism don't afford any plausibility for prenatal acetominophyln exposure as being a potential cause. What possible mechanism would be involved here? Nobody can answer that, at least intelligently. The next is what's called coherence, which basically says that the causal relationship shouldn't conflict with established knowledge and our understanding of the natural history of the disease. And this is not a coherent finding because it's not plausible or coherent with our current understanding of autism spectrum disorders. Experiment is the eighth criteria of this Bradford Hill list, and this usually comes in the form of an experiment to prove causality, like a randomized controlled trial. But we almost never get this particular one when we look at things related to pregnancy because we just don't do many randomized trials and pregnancies where we deliberately expose a group of pregnant women to something versus placebo. So we don't randomize some women to take Tylenol every day and other women to take placebo and then look five years later and see whose kids got autism. But we can construct quasi experiments with large retrospective data, or again, as I'll discuss in a minute, with these sibling studies where the sibling pairs are used as their own internal controls. The prevalence of autism in sibling pairs, where one sibling is exposed to tylenol prenatally and the other isn't, is the best we can do to create sort of a type of study that controls for lurking variables and is much better than looking at unmatched large populations, like say the Amish versus a U.S. population, where there are no sort of internal controls for some of the genetic or cultural or educational opportunities, things like that that lead to diagnoses. You need the sibling pairs as internal controls. Okay, and the last of the Bradford Hill is analogy, which states that a similar causal relationship in another area can strengthen the argument for a new proposed causal relationship. And here again, we don't have any analogous understandings of prenatal drug exposures leading to this sort of neurodevelopmental disorder. Even for all of the neuropsychiatric type medications that we have out there, we don't see anything like this, let alone for something that reduces fever. So on the strength of the Bradford Hill criteria, acetamenophine exposure as a causal agent of ADHD or autism fails horribly. It fails, in fact, on every single point, particularly if you understand that the idea that there's even an association hasn't actually been established in legitimate studies.
SPEAKER_01:But let's dig more into this claim by the president that Amish kids don't have autism.
SPEAKER_02:Well, of course they do. And this but this gets to an important issue that we need to discuss about why autism rates are rising and why autism is reported as different at different levels in different communities around the country and around the world. But this is a great starting point to understand this. So as a community, the Amish don't access the type of care that children in suburban America do with their private practice pediatrician. When we look at the differences in diagnosis of most diseases, it has mostly to do with how we've implemented screening programs and how objective the criteria for a particular disease is. When we diagnose anemia, for example, we have a relatively objective diagnostic criteria, and so there's some general agreement at least about who has anemia and who doesn't, we can agree. We might debate the edges a little bit, but we can agree. And there there still may be a difference, though, in who gets those blood tests done. So some people get CBCs every year at a doctor's office, and some people, in some communities, like among the Amish community or say Cuba, they don't go to the doctor that often if they can help it, and when they don't get testing that much because they're paying cash or it's expensive. So even though anemia has a rather objective diagnostic definition, it goes underdiagnosed in communities like the Amish, and among others who don't utilize the healthcare system in the same way that many other people do. That doesn't necessarily mean they have less anemia. So to answer this question specifically about Amish children in autism, we do have data. There was some research presented at the International Society for Autism Research in 2010 from a group of autism professionals, some of whom worked here in Tennessee at the Vanderbilt University, and they went and screened for autism spectrum disorder in a population of Amish children. Now, at that time, so this is 2010, they when they did their research, the reporter rate of autism in the United States was one in ninety-one children. They sent screeners door to door in an Amish community and they administered standardized evaluation tools to eighteen hundred and ninety-nine children. Twenty-five children of that group screened positive for autism spectrum disorder, which would have been one in seventy-five children, a rate that's actually higher than what we reported in the general population in the United States at that time. Now, half the children that underwent more in-depth investigations afterwards were diagnosed with an autism spectrum diagnosis. So this came out to be one in about 146 children. So the true rate is somewhere in that range, and you can even see how sensitive this is to which kids got follow-up and how the tests are administered and things like that. But that range, somewhere between one in 75 children and one in 146 children, is consistent with the range of autism reports in the general U.S. population at the time, particularly when you look at state level data and compare that to the states where these Amish kids actually lived, because it varies widely from state to state. Now, most autism diagnoses are made by pediatricians who are seeing children for routine screening exams over multiple visits, and we know that the propensity of some clinicians to diagnose autism is higher than it is among others, and Amish children just don't have these routine and regular visits, particularly with clinicians who are perhaps more prone to label children as having autism. But this study shows that the rate of autism among an Amish population is consistent, in 2010 at least, with the rate of autism in the general population, though it could be a little higher or a little lower. But again, autism in the general population is not diagnosed as strictly or as it could be, or with the same tools that they use in the study, and we all know it tends to be at least a little bit overdiagnosed. So our best evidence is that the rate of autism in Amish children is exact same as it is in the general U.S. population.
SPEAKER_01:On the flip side of that extreme, I think that would explain the other sort of backhanded comment he made about how California seems to have a bigger problem with autism because it has more diagnoses. Well, we know that in general in California, there's a lot better funding compared to some other states to assess for autism early and broadly, and then of course to treat it. So they are catching really as many cases as they can. And who knows, maybe they could be overdiagnosing some cases too, which is in complete contrast to the places that don't utilize health resources. And these sort of simple canards are repeated so often online. I expect that the idea that Amish children don't get autism will be talked about over and over again after that press conference. But again, it's simply not true. You always talk about the difference in the true rate of a disease versus the rate of diagnosis of a disease, and they are not the same thing. The rate of diagnosis is basically what we l label it as, and it's subject to a lot of variables, including the type of screening programs implemented and the threshold for diagnosis, and even the access to get assessed for diagnosis. And Amish communities don't necessarily access healthcare the same way for these sorts of normal well-child preventative visits. And culturally they may not be seeking out autism diagnoses in their children or other other types of behavioral or neuropsychiatric diagnoses. So the rates of labeling autism are naturally going to be lower in that case.
SPEAKER_02:Yeah, it's also not like anemia. We don't have a gold standard objective diagnostic set of criteria to compare one group to another with. So when you're when we if we all agree that a hemoglobin of whatever below a certain level is anemia, that's easy. But there's so much variation and subjectivity. Even the screening tools, the two widely used screening tools, don't necessarily agree. We do have these standardized tools for the diagnosis of autism. But many patients who've received a diagnosis with one of the tools will not with another, and vice versa. So there's disagreement among even the sort of gold standard tools. So if you've received, for example, a diagnosis with a tool like the Autism Diagnostic Observation Schedule or the Autism Diagnostic Interview revised, then you're likely to continue to test positive for these tools over time. They tend to be internally consistent, even though they vary each other a little bit. So that's our best stuff. But there's still problems with it. And in this group of Amish children, those are the tools that they use, but then the final diagnosis that was assigned to the kids was only there with half the kids at screen positive. So even those tools don't tell the whole story, and there's just wide variation in practice among those who are potentially diagnosing children with autism about how they use these tools and which ones they use and how they use that information to arrive at a final diagnosis.
SPEAKER_01:Yeah, and it's interesting again, you already pointed this out, but if you assume that these tools are the gold standard diagnostic test, then the rate of at least screening positive for autism in that study being one in 75 was higher than the rate in the general US population at the time of one in 91. And of course, those are children who are typically not vaccinated, and for the most part their mothers had unmedicated home births and avoided medications and vaccines and Tylenol specifically during their pregnancies.
SPEAKER_02:And epidurals and everything else, yeah. Well, epidemiologic studies are severely limited, and epidemiologists themselves are limited by their complex problems with data to ever really finding what we might call scientific truth in their data. And they don't like this either. They do apparently, some of them do apparently like making$150,000 as a plaintiff's witness, but there are just some epistemiologic limitations in the way epidemiologic science is conducted and the methods used. And one that's important here is this idea of confounding by indication, which states that you can't tell whether the observed health outcome you're seeing is due to The medication or the exposure, or due to the underlying medical reason that the medication was taken to begin with. So is it the blood pressure medicine causing the outcome or the hypertension? Is it the anti-seizure medicine causing the outcome or the epilepsy? Is it the depression or anxiety or mood disorder causing the outcome or the antidepressant medication, etc.?
SPEAKER_01:Yeah, so with Tylenol, you'd be saying maybe having a febrile illness during pregnancy could predispose children to autism. And so mothers who take the Tylenol for fevers are still going to have that association, whether or not it's causal. Or then maybe mothers who have frequent headaches and are are frequently trying to take Tylenol for that, maybe certain types of headaches also run together with autism, just as a couple examples of this. And we know already that maternal fever and infection are independent risk factors for adverse neurodevelop developmental outcomes in the children. So just that fact alone really could confound Tylenol exposure.
SPEAKER_02:That's right. And some of the founders of the science of epidemiology in the 1950s argued that the association between cigarette smoking and lung cancer was due to the fact that people with lung cancer were soothed by smoking cigarettes. And so they already had lung cancer, and then somehow they intuitively figured out that smoking felt good for them and they did it, and that the lung cancer preceded the smoking. And that's obviously absurd, but that's exactly what you could prove using epidemiologic methods. They're just limited in a fundamental way that we're not limited to with interventional trials.
SPEAKER_01:Okay, well, let's talk about the sibling studies, and I'll start with my own single case anecdote of a sibling study in with my two kids, and I could honestly draw the opposite conclusion about Tylenol from that. So while I was pregnant with my oldest kid who does have autism, I never had a fever or a headache, never took Tylenol, had a pretty healthy pregnancy. We walked a lot, drank a lot of green tea. I got all the vaccines in pregnancy, including the COVID shot. We got all the pediatric vaccines after birth. It was a normal birth, and I got an epidural. So then the younger sibling who does not have autism during that pregnancy, I ended up taking Dayquil and NyQuil quite a lot. It was a different time of year, and now I was getting exposed to all these germs from the first kid, and I needed to get through the work day and take care of the kid and all of that. And then I also I don't know that it matters, but I exercised less, I ate worse. I technically failed the diabetes test that time. I had become moved to a different area where you can't walk anywhere, you have to go by car, all the food is fried, so just overall a lot less healthy. And this time the green tea just made me throw up. I just had one little sip of it and it was disgusting for that pregnancy. And I also got an epidural during that birth, got all the vaccines, and that time I also got the RSV maternal vaccine, in addition. So that second kid is not autistic.
SPEAKER_02:Yeah, I'm still waiting on someone to jump on the green tea as the cause of autism because of your story. That'll be out next week.
SPEAKER_01:Yeah. That's just as strong of an evidence basis as what we have against Tylenol or vaccines right now with autism. And I've talked to some autism professionals, especially since I interact with them for my autistic kid. And especially lately after this press conference, a lot of them have told me that the parents of their clients, their little kids who have autism, are saying, I never took Tylenol, but my kids still got autism. So what's going on here? So on the front end, at least the general public, or I'd say maybe the general public who knows about autism intimately, including myself, of course, already we have a lot of baseline skepticism about this Tylenol claim. And if you think about it, it's basically a false promise that if you don't take it, you'll save your kid from autism. You definitely won't. But I think that talking about real sibling studies on a larger scale than just my isolated experience is really going to help underscore that.
SPEAKER_02:Right. So if you did a randomized control trial, you would be controlling for many variables that we just don't understand, like genetics or environmental exposures, dietary variation, a whole lot of things that we can imagine or can't even imagine, or combinations of things, and interplay between various factors that we don't understand. So you take two large groups of subjects, and the only thing that's different between them is whether or not they were exposed to, in this case, Tylenol prenatally, and then you measure an outcome, like the incidence of autism or ADHD. And that's our most powerful tool for determining causation. But again, we can't do that with pregnancy. So this is why we have so many poor epidemiologic data and things, claims about medicines in pregnancy, because we don't have these sorts of studies. And we're not going to do them. We consider them unethical, right? So a sibling control study is one of our best surrogates for this because you find groups of siblings and you find out which ones were exposed to Tylenol prenatally and which ones weren't. And many of them will be concordant and some will be disconcordant, and those are the powerful ones in that. And then you find out, of course, which ones have autism or ADHD and which ones don't. And basically they end up being their own internal control for all these things like genetics and environmental exposures and other variables we don't know about. So there was a large nationwide study, a Swedish study, published in JAMA in 2024 that utilized this methodology. And when they looked at the data without sibling controls and just the raw data, they did find a significant association between acetaminophine use during pregnancy and the risk of autism, ADHD, and intellectual disabilities. But when they applied the sibling control analysis, that association, all those associations, completely disappeared. There was no evidence of an increased risk for any of the conditions. So this suggests that the associations seen in some of the smaller, less rigorous studies are due to some unknown or uncontrolled confounder that goes away when you control for siblings. We already know that if you have a sibling with autism, you're at more risk of that. But in this case, they'll see that trend. But in this case, you'll be able to control for the tylenol exposure in these discordant pairs. Some pregnancies, the mother didn't expose the pregnancy to tylenol, and the child did or did not get autism, and another's the other way around. And so that's the control that they applied. And yeah, when you're applying something like a sibling control, you're probably controlling for genetic factors related to autism or ADHD, but you also might be controlling for things that require pain relief or fever. Or you might be controlling for other unknown environmental things, or even what school opportunities and pediatric surveillance programs were available in those communities. So that's why some sort of methodology that controls for lurking variables is important. And this is our best evidence. And the best evidence says that there's no association between acetaminophen and autism or ADHD.
SPEAKER_01:Yeah, so a sibling study or randomized controlled trial is really the only way to control for genetics. And when you adjust for sibling pairs, you essentially are adjusting for genetics. In Baccarelli's paper that we talked about earlier, the one that seems to be bought and paid for by the plaintiff industry, they included the unadjusted data from that study in their data set, but they didn't like the sibling adjusted results that showed no association, so they just included the unadjusted results. That's just blatant dishonesty right there.
SPEAKER_02:Yeah, you have a keen eye for detail. That is a salient problem with this paper, and the question of why they did that gets to the heart of their bias and dishonesty. And others have noted this too. I'll put a link that's been reported on in Stat News, and Baccharelli has been heavily criticized for this by people who've spent more than 10 seconds looking at the paper. This study, it the sibling twin study, or the Swedish study, is the best study we have for answering this type of question. And the type of analysis that helps to control for these lurking variables is the best analysis we have, and he chose to exclude it.
SPEAKER_01:Yeah, the this one should have trumped all the other smaller, poorly controlled studies, pun intended.
SPEAKER_02:Trump, yeah.
SPEAKER_01:The scientific method asks us to try to disprove our hypotheses because a true association should stand up to that. And that means don't omit data like what they just did here. Now, a person who's just been paid$150,000 to testify in a court case with many more hundreds of thousands of dollars expected if the case proceeds, is gonna try to manipulate and cherry-pook and exclude data in a way that supports the intended conclusion, the desired conclusion, but still looks passable because that's what they're being paid to do. And so that's really the biggest bias of all. But a true scientist would try to control for the unknown variables. And when you do that here, we find no association at all between the prenatal acetaminophen use and autism or ADHD or intellectual disabilities in the offspring.
SPEAKER_02:Yeah, and we live very much in this age where low quality academics publish meta-analyses all the time of all the studies, and these meta-analysis of heterogeneous study methodologies and heterogeneous groups are just really of no value, and publishers need to stop publishing them. And we should remember that one good study with appropriate methodology cancels all the other low quality studies, and it's not fair to lump together bad studies with poor methodologies into a meta-analysis with a good study and then wash out essentially the findings of that good study. Meta-analyses are really only worthwhile when the data is homogeneous and you're meta-analyzing high-quality studies. But since there's no association here using our best methodology, then we can't even begin to ask all those other questions about causation that I already went through because there's no evidence for correlation, let alone causation.
SPEAKER_01:So, what is this navigation guide methodology? They thought it was so important, they put it in their title of their paper.
SPEAKER_02:Yeah, yeah, and it was mentioned in the press conference too. And this re- this is research methodology that was used in this corrupt study that we've been discussing. So, again, one thing people need to understand is that fields like sociology and environmental health are among what we call the soft sciences. They rely on a lot of expert opinion and qualitative reviews, and they really don't have the tools of normal, objective, rigorous science at their disposal, like the randomized controlled trial. In fact, this was much how much the same case for clinical medicine prior to the early 1990s. There was a landmark paper in 1992 in JAMA by Antman et al. that heavily criticized our expert reviews of literature. If you go beget a journal article or a journal in some medicine subject from the 1980s, you mostly just see a bunch of reviews of topics written by some supposed expert with his opinions about things, almost like the level of case series kinds of information, not trials. As we started to see randomized controlled trials come out and show that evidence from those data were often the exact opposite of some of these experts and what people had thought was true, well, there was a call to change and to not utilize this sort of expert opinion systematic review anymore that was so popular in medicine previously. That's what worries me about some of these low quality meta-analysis we're seeing today is we seem to be going back to a form of that. They just found that these opinions that people were publishing lag significantly behind real scientific evidence and were very often wrong or opposite of the truth when we had a randomized controlled trial come about. So in the mid-1990s, after that paper by Antman, we moved away from these sorts of expert-based reviews in medicine towards what we now call the age of evidence-based medicine, I guess, that relies more on data from RCTs. But in areas where randomized controlled trials are not always available or appropriate to do, such as in environmental health and sociology and other soft sciences, we still rely mostly on expert review articles. They are fundamentally limited and if not flawed and subject to bias and corruption that randomized control trials simply aren't. So environmental health has tried to improve the quality of these reviews, not get rid of the reviews, but improve the quality of these reviews. And that's where the navigation guide methodologies come onto the scene. This started in 2009 and was an attempt to infuse some more rigor into the way these reviews were conducted by demanding that they have things like pre-specified protocols, along with comprehensive and documented search strategies and some assessment of the risk of bias, along with standardized and transparent documentation of the findings. So essentially, this was just a toolkit or a checklist about how to more fairly do an expert opinion review, and you could look at it and go through and look at those questions and try to understand if you were reading it or reviewing it what the author was doing. But it doesn't actually change the underlying lack of quality and reliability of the data that's being analyzed, and it's still subject to significant mistakes and errors and bias. It may be a little better than just allowing some ivory tower expert or plaintiff's expert to say whatever he or she thinks in a monograph using whatever methodology and including whatever papers they want to include, but it's not much better, and as you pointed out, even in this paper they chose to discount and not utilize the best available evidence because they knew it it didn't agree with them. So they knew that before they designed the study. So they designed a methodology using the checklist that would avoid looking at the best evidence available. So this is not some giant leap forward like the randomized control trial was for medicine that essentially displaced these expert opinions. It's just something to gild the Lily a bit of what these expert opinion review articles are. You can design a process and a study to find almost anything you want because you're still just manipulating poorly controlled, mostly retrospective data. If they were looking at good high quality data, then a meta-analysis would be appropriate, most of the time, at least. So this methodology is being used in the first place because the data is mostly retrospective, heterogeneous, and low quality, that it's really just an expert opinion piece.
SPEAKER_01:So it's just another framework for assessment that still allows for cherry-picking by the authors, and it's still only as good as the data that's in the analysis. And we keep saying here, these authors have significant bias, and they designed their analysis in a way to exclude the highest quality data that was contrary to what they wanted it to say.
SPEAKER_02:And that's why it's a soft science. If it was a hard science, we'd have rigorous prospective data and trials.
SPEAKER_01:Should we talk about vaccines?
SPEAKER_02:I don't know that we have time to do that today. We're we're into this, but we can do that later. It should suffice to say that we have over 50,000 trials that demonstrate vaccine safety and efficacy, and we know definitively, and without any question at all in the scientific community or in the scientific literature, that neither vaccines nor any of their individual ingredients or components, as people like to say, are associated or causal in relationship to autism spectrum disorder. And I do think we should address this later, but to put it in perspective, we have so much overwhelming data that the b the belief that vaccines might cause autism at this point is akin to the belief that the earth is flat.
SPEAKER_01:But it looks flat.
SPEAKER_02:Well, it's an oblate spheroid, so it's not flat.
SPEAKER_01:Okay, fine.
SPEAKER_02:You thought I was gonna say it was round, but I know better.
SPEAKER_01:Okay.
SPEAKER_02:We'll get a listener comment.
SPEAKER_01:Can we come back to the claims again that the rates of autism diagnoses are going up? Because in the press conference that was heavily emphasized, and it was clearly confused with thinking that the rates of actual disease have gone up, and we know that diagnosis and true disease are not the same thing, but it does make it appear very concerning and the accessibility bias will lead parents to consider things like like common exposures, vaccines or Tylenol, for example. When they do that, they're not usually disciplined about the correl correlation not equaling causation or things like the Bradford Hill criteria you just went through. And really those criteria are probably beyond the immediate recall and understanding of most physicians, let alone most patients. So what are we to tell patients when they ask us about the increasing rates of autism, like they highlighted in this in this press conference? What's the cause of that?
SPEAKER_02:Well, I will say that the increasing rates of autism are also strongly correlated with a ton of other things, like the increased use of organic foods or the increased use of health supplements in the US. Whole bunches, and there's plenty of examples of these online if you if you go look for them. They're all very strongly correlated with autism, and so the exact same science that would lead to vaccines or Tylenol as being related should make you pause with health supplements and organic food. The truth is anything that has increased in utilization over the last 20 or 30 years can probably be strongly correlated with the rise of autism spectrum disorder. So that could be cell phones or electric cars or almost anything, any food item that's popular today that wasn't popular 30 years ago, for example. But yeah, the fundamental question is why are the rates of diagnosis of autism spectrum disorder increasing and is the actual rate of autism itself increasing?
SPEAKER_01:Yeah, yeah. So that's why I brought it up. So tell us.
SPEAKER_02:Okay. Well, we're almost out of time, but let me give an outline of a few things to consider. Many of the rates that you see mentioned all the time come from the CDC's Autism and Developmental Disabilities Monitoring Network, which was established in the year 2000. They abstract data from health and special education records of children and usually focus on a cohort of eight-year-olds at each surveillance interval. So they're taking a snapshot of eight-year-olds at any given time, and that snapshot largely is capturing children who are receiving treatment or special education services for autism, which is a different thing than receiving a diagnosis of autism or truly having autism, let's say.
SPEAKER_01:Right. There's all sorts of resources and educational opportunities available for kids with autism today that didn't exist even 25 years ago when I was that age. So increased utilization of those resources doesn't necessarily signify that an increased percentage or amount of kids have it. And remember, kids with autism spectrum disordered didn't they didn't have access to these things 25 or 30 years ago. So there was no way to count it through that method of resource utilization.
SPEAKER_02:Yeah, so that's a big part of this. So the data often cited doesn't again reflect our rates of true disease as much as people would think they would. But this is where you get the often cited figures of a rate of autism of one in 150 kids in the year 2000 and then one in 31 in 2022. So we're starting with bad data, apples to oranges comparisons. We also, as we mentioned a little bit earlier, see huge variations in the United States from state to state. So the rate of autism in Laredo, Texas, for example, is 9.7 per 1,000 children, but it's 53.1 for 1,000 per 1,000 children in California.
SPEAKER_01:And that's probably a good example just of the availability of treatment programs and educational models. So I'd expect a more densely populated area that also has good funding for autism services to have more resources available for the kids than maybe an underfunded or rural area would. And in this example, their method their methodology may be over-identifying the rates in California while maybe under-identifying the rates in Texas.
SPEAKER_02:Yeah. So let me give you a few other factors to think about that might be leading to these increased rates of diagnosis. So the first one that is just broadened diagnostic criteria. The DSM IV ended in 2013 when we adopted the DSM-5. In DSM 4, autism was classified within a broader category of pervasive developmental disorders, which included autistic disorder, Asperger's disorder, and pervasive developmental disorder not otherwise specified. And that was a catch-all for individuals who had some social impairments but didn't meet the full criteria for autism or Asperger's. Then in 2013, the DSM V shifted this radically. It put all three of these under the umbrella of autism spectrum disorder. It also changed the criteria. The DSM IV had a triad of social impairments, communication impairments, and restricted repetitive behaviors, but this became a domain of social communication deficits or social interaction deficits, along with repetitive patterns of behaviors or interest. A triad went away. That second domain was expanded as well to include hyper or hyporeactivity to sensory input or unusual interest in sensory aspects of the environment. So not only did DSM5 combine three separate diagnoses into one, essentially, but it made it much easier to be diagnosed with a disorder by simplifying and broadening the qualifications and criteria to include more behaviors.
SPEAKER_01:I'm sure when they changed those criteria in 2013, it didn't all happen at once. Because I doubt every clinic out there brought in every kid they had ever seen to reassess them immediately and change their diagnoses overnight. But we do see that roughly prior to 2012, the diagnosis rate averaged around one in 90. And after that, and after everything settled in with the DSM 5, the rate was closer to about 1 in 35.
SPEAKER_02:Right. And the difference in that criteria alone is likely the reason for that change over the last 15 years or so. What looks like an increasing rate of autism is almost certainly just related to DSM V over the last decade or a decade and a half. Remember that Amish study that we looked at was done in 2010 before this change occurred. If it was done after the change, with broadened criteria, et cetera, we might expect to see an increased rate of autism due to the diagnostic accretion and broadening of the criteria. So understanding all that, I think it would be hard to argue that the rate of autism has increased at all since 2012. It almost all seems to be an effect of DSM 5. A second factor is just increased awareness and screening. The rate of autism is not just higher in California than it is in Texas due to the reporting methodology that does heavily rely upon the availability of educational programs to treat autism. It's also higher because we've had a significantly increased awareness and screening for the disease, in some areas more so than others. The American Academy of Pediatrics now recommends universal screening for autism spectrum disorder during well child visits 18 to 24 months of age. And we know that increased screening for any disease leads to increased diagnosis, even overdiagnosis. And then a third factor is this improved access to services. So it's not just that the CDC ADDM network partly relies on educational systems and identification. It's also that those educational systems are in place to treat children with autism. And so the parents are more likely then to seek a diagnosis in the first place to make sure that their child is enrolled in these opportunities that are now available. Getting the diagnosis means something for their children now. This is especially true in comparison to the general public 20 or 30 years ago, due to more awareness and decreased stigma and acceptance of treatment and things like that. In 2022, 67%, two-thirds of the children with autism spectrum disorder were identified through their eligibility for a special education program. So the absence of those programs contributes significantly to lower rates. Two-thirds of the diagnoses are there because these programs exist today.
SPEAKER_01:Right. So we're more likely to find the diagnosis as more educational programs are developed, and parents are more likely to seek out the diagnosis so they can take advantage of those programs. So it's a double whammy. Like in our case, I hadn't suspected I hadn't suspected autism. Maybe I should have, but our but it was through the school system that we got prompted to seek the diagnosis. The pediatricians had set my kid up with speech therapy for a speech delay and feeding therapy and occupational therapy, but they hadn't labeled those things together as autism. But then when the school system suggested get assessed, we got him assessed. And then through that process, I learned that a lot of behaviors I thought were really neurotypical weren't, and specifically would have predicted a lot of difficulties adapting to a normal school environment. And now I'm obviously really glad that we did that because the services we have for him are so great that I think everyone should want those services if they can get them. And so I'm definitely a big advocate for getting assessed. If there's even some kind of suspicion that someone may have autism, because it can only help to know that.
SPEAKER_02:And those attitudes are so different than they were 25 years ago and those resources available. So that's a big factor in the increased rates of diagnosis. There have also been disparities or gaps that have been closed in the last 20 years. We used to think that autism primarily affected white boys, but the truth is white boys, well, they had more access to care, and girls in some cases they tend to present with different symptoms or present differently. In the current data, the 2022 data, white children actually have the lowest incidence of autism spectrum disorders, while the highest incidence is among multiracial children, Hispanic children, and well, the highest of all is South Asian Pacific Islander children, but the highest major category is among black children today. This shift started happening in the mid to late 2010s, and it's because access has been improved for those populations, and not just available in more affluent and more white communities. But as you start to expand access to care, then suddenly a whole bunch of people who previously were undiagnosed get diagnosed. So again, the rate of diagnosis has been increasing, but it's unlikely that the rate of autism itself has significantly increased, at least in the last 25 years.
SPEAKER_01:And we're still learning a lot about genetics, because it's been observed that the concordance rate of autism spectrum disorder among identical monozygotic twins is up to 92%. Whereas for fraternal twins it ranges anywhere from zero to just 10% in some studies. So this is very powerful evidence that there must be a good genetic component, not just environmental, at least as the main cause. But we know really it's multifactorial. Unfortunately, they haven't found just one single gene that either you have it or you don't. But they have identified certain genes that are associated more strongly with autism spectrum disorder. We also know that advanced maternal or paternal age are more consistently associated with higher risks of autism. And men and women are older now at the time of parentage than they were on average 20 years ago. And there's also an association between gestational diabetes and autism spectrum disorder. Not saying that implies causation, obviously, but our patients are becoming more obese and more with higher rates of diabetes than they've ever been in the past.
SPEAKER_02:Well, and all those are just reasons why, as the demographics change, essentially, we would expect to see a slight, modest increase in actual rates of the disorder, but almost all of the reported increased rate of diagnosis is likely due to the factors we've just discussed, and they're they're artifactual. They don't represent a true increased rate of disease. And that difference is important. And but because we've seen this increased reported rate, it confounds attempts to understand environmental or epidemiologic causes. We're taking it as a preset that autism is more common today than it was 20 years ago in studies like Baccarelli's and others, and that's simply not the case.
SPEAKER_01:Well, there's so much more we could say, but I think we're going over time at this point.
SPEAKER_02:Yeah. Well, I would just like to quote myself from my book, Clinical Reasoning. Let me read one paragraph about why we keep making these sorts of mistakes cognitively. And I know that humans have a tendency towards first order thinking, which is what I'm discussing and we all do. And so this was a quote from that about autism. Quote, it's simply easier for us to imagine and understand simple solutions to our problems. Humans search for cause and effect paradigms to explain observable phenomena. Thus it's easier to say and understand that the increased rate of autism diagnosis is caused by vaccines, and I'll have to add in ortylinol today, instead of the increased rate of autism is related to a combination of changing diagnostic criteria, diagnostic accretion, greater awareness that can sometimes lead to overdiagnosis and misdiagnosis, rising parental age, spatial clustering, disuse of intellectual disability diagnoses, as well as increased prevalence of other known and Unknown polygenetic and environmental causes. We tend to reject complexity for simplicity, and we tend to favor first order thinking rather than the higher order thinking. And that's my description of that press conference last week.
SPEAKER_01:Yeah, we want l an open and shut case. We want the simple answer.
SPEAKER_02:And a lot of well-intentioned people do too. They would love for there to be a smoking gun or one thing we could change that would fix everything. So I know we need to wrap this up, but I'd also just like to say that we've got to hold the medical profession to a higher degree of scientific rigor. So one of my favorite examples, if you're routinely ordering vitamin D levels at new OB visits or Well Woman visits, and you're having everyone supplement vitamin D because it's below 30, or if you're doing a lot of the other quasi-scientific things that we discuss on this podcast, then you're can you're practicing the same quasi-science that led to this press conference last Monday. You're relying on non-rigorous statistics that don't have causal relationships established, therefore misinterpreting causation and correlation. And it's time to stop and let's be rigorous scientists.
SPEAKER_01:I think I've been seeing vitamin D in patients' charts more than ever, so we should talk about that another time.
SPEAKER_02:Yeah. Well, physicians are responding to the broader Maha movement, and you see this in ordering hormone levels inappropriately and all sorts of vitamin levels. And we've had a spillover of what we've called functional medicine, and it's non-rigorous, it's non-scientific, and it uses the same science that would have us believe that Tylenol causes autism. So more to come.
SPEAKER_01:Well, let's wrap up for the day.
SPEAKER_02:Alrighty. Well, we'll have links to everything we discussed on the eventually on our website, and we'll see you guys in a couple of weeks.
SPEAKER_00:Sounds good. Thanks for listening. Be sure to check out thinking about obgyn.com for more information and be sure to follow us on Instagram. We'll be back in two weeks.