The Peter Attia Drive - #367 - Tylenol, pregnancy, and autism： What recent studies show and how to interpret the data

00:00:00.000 Hey, everyone. Welcome to the Drive podcast. I'm your host, Peter Atiyah. This podcast,

00:00:16.540 my website, and my weekly newsletter all focus on the goal of translating the science of longevity

00:00:21.520 into something accessible for everyone. Our goal is to provide the best content in health and

00:00:26.720 wellness, and we've established a great team of analysts to make this happen. It is extremely

00:00:31.660 important to me to provide all of this content without relying on paid ads. To do this, our work

00:00:36.960 is made entirely possible by our members, and in return, we offer exclusive member-only content

00:00:42.700 and benefits above and beyond what is available for free. If you want to take your knowledge of

00:00:47.940 this space to the next level, it's our goal to ensure members get back much more than the price

00:00:53.200 of the subscription. If you want to learn more about the benefits of our premium membership,

00:00:58.020 head over to peteratiyahmd.com forward slash subscribe. Welcome to a special episode of

00:01:06.000 The Drive, everyone. If you've been following the headlines recently, you may have seen,

00:01:10.300 of course, stories linking acetaminophen or Tylenol use during pregnancy to autism. Not surprisingly,

00:01:16.920 those headlines have generated a lot of questions, a lot of controversy, and a lot of confusion.

00:01:21.260 I've heard about this a lot from every direction. My patients, listeners of the podcast, friends,

00:01:28.760 family members, people writing in through the website. Basically, it's like I'm sure many people

00:01:33.240 in the space, we've all been inundated by it. And the more I thought about it, the more I realized

00:01:37.060 this was a great opportunity to, I think, maybe put forth a framework for how to think about these

00:01:43.240 things critically. While we initially thought we would just do this in the newsletter last week,

00:01:47.100 once we got into it, we realized, no, this doesn't really lend itself to an article or even a short

00:01:52.460 video. It really commands effectively the discipline of what we do in the AMAs, the Ask Me Anythings.

00:01:58.060 Of course, unlike the normal AMAs, this is going to be made available to everybody.

00:02:02.440 So before we dive in, though, I want to kind of lay out some groundwork. We're going to unpack some of

00:02:06.780 the points in more detail that I'm going to lay out below. But I also want to make sure we're

00:02:12.440 starting from a place of reference. I want to start out with a few important observations.

00:02:16.780 Okay. So the first is autism rates have risen dramatically over the past generation. Now,

00:02:22.780 we're going to talk about why that might be, but it's very important to state up front that there

00:02:26.540 is unlikely to be a single cause. Why? Because complex conditions usually don't have simple

00:02:32.540 explanations. This is true of obesity, despite what some people would have you believe, that it's

00:02:38.260 just this one thing or just this one thing or whatever. But the reality of it is complex

00:02:42.480 conditions require multiple things typically. So anytime we look at a possible contributing factor,

00:02:48.680 we need to kind of resist the temptation to assume it's the sole cause. Now that doesn't

00:02:52.620 diminish the interest in identifying a bunch of potential causes. Okay. Second point I want to

00:02:57.860 make here, and it's kind of weird that I have to make it, but I do. Science is supposed to be

00:03:02.480 apolitical. Unfortunately, that's not the case. And for reasons that I don't think I'm smart enough

00:03:07.440 to understand, autism happens to be one of those examples. But so are many other topics we've

00:03:12.260 discussed, like nutrition or protein, which has become remarkably political. My goal here is not

00:03:18.180 to have a political debate, but rather to examine the evidence as carefully and objectively as I can.

00:03:25.240 Third, we do need to realize something that I think is very hard to accept, and that is that as humans,

00:03:31.220 we are not wired to think scientifically. I want to restate that because it sounds condescending,

00:03:38.760 but it's simply an observation of how we have evolved. We are not wired for critical and scientific

00:03:45.300 thought. This is something I've written about, and we're going to actually link to a piece I wrote

00:03:50.900 over 10 years ago that I think synthesizes that point really well. But again, it really comes down

00:03:57.500 to the fact that we should understand that the scientific method and critical thought are human

00:04:02.700 inventions. They're wonderful inventions, and I would argue they are the single most important

00:04:08.700 invention our species has ever put forth. And without this, nothing else would exist. We'd still

00:04:14.700 be living in caves. But that doesn't mean that it comes naturally, and it doesn't mean we're wired

00:04:18.940 to do it. So just keep that in mind as you catch yourself, as I catch myself, falling into

00:04:25.020 non-scientific thought. We're going to rely on a framework at some point during this discussion,

00:04:30.960 which is very helpful when considering epidemiology, which is the branch of science we're going to be

00:04:36.100 talking primarily about today, and it's called the Bradford Hill Criteria. These are nine principles

00:04:40.600 that were laid out in the mid-60s to help us determine whether an observed observation is likely

00:04:46.480 to be causal. So this framework looks at things like strength of association, consistency across multiple

00:04:53.300 studies, biologic plausibility, temporality, and things like that. They're not a checklist per se,

00:04:58.580 but they provide a disciplined way to try to make sense of correlations and interpret which ones have

00:05:05.120 a higher probability of being causal from those that don't. Another thing I want to point out is we're

00:05:10.460 going to be talking about medications. We're going to be talking about pregnancy. And I think it should

00:05:14.660 be obvious, and I'm sure anyone listening to this or watching this who has gone through pregnancy

00:05:19.460 will understand that the bar is very high when we are talking about medications to be used during

00:05:26.440 pregnancy. Most physicians, myself included, though I don't treat very many pregnant women,

00:05:31.380 think about drugs and supplements very differently in the setting of pregnancy. Of course, because we are

00:05:37.980 typically not treating patients with life-threatening conditions, our mantra is during pregnancy women

00:05:43.500 should basically stop all medications and supplements beyond the obvious ones, such as prenatal vitamins,

00:05:49.460 or hormones such as thyroid hormone, which can be essential. But anything that's even in a gray area

00:05:55.320 or probably okay, we tend to just avoid. Now, since the late 70s, the FDA has used a very simple

00:06:02.960 letter system to classify drugs by their risk during pregnancy. These categories go by A, B, C, D, and X,

00:06:11.180 and basically each letter refers to a level of evidence, mostly from animal and human studies,

00:06:17.520 about the potential harm of the drug to the fetus. So for more than, I don't know, 35 years or so,

00:06:25.300 this was the framework physicians relied on. About 10 years ago, the FDA replaced it by a framework that

00:06:31.020 is called the Pregnancy and Lactation Labeling Rule, or the PLLR. The idea was to move away from single

00:06:37.080 letters and instead provide a more descriptive guidance. And in theory, that's an improvement, but in practice,

00:06:43.460 it's been kind of slow to roll out. And frankly, I'm a little guilty of generally thinking about it in

00:06:49.260 the A, B, C, D, X category. And that's what I'm going to refer to a little bit. So I'm going to stick

00:06:55.700 with that older category. And while it's imperfect, it's widely understood. It is still a clear framework.

00:07:02.440 And I just want to share with you as we begin this, so you have a broad sense of how drugs fit into

00:07:08.620 this. Okay. So category A means that there is no demonstrated risk in well-controlled human studies.

00:07:17.200 Again, that's pretty unusual because that's a hard thing to do. And that's reflected in the

00:07:22.320 proportion of total drugs and supplements out there that fit in this category. And it's somewhere

00:07:27.320 between two and 5%. Okay. So what does that mean? That means that is completely safe. We have

00:07:32.420 definitive evidence that women can take these things during pregnancy. And obviously as reflected

00:07:36.360 by the numbers, virtually nothing fits in that category. By the way, the examples I gave earlier

00:07:40.300 of thyroid hormone and prenatal vitamins do fit in that category. Then you have category B,

00:07:45.780 which says there's no evidence of risk in humans, but animal data might show signals in some studies.

00:07:52.280 And so these are generally thought of as safe, but exercise caution basically. And this is 15 to 25%.

00:07:58.260 Then you have category C. This is the biggest one. This says risk can't be ruled out. We don't have

00:08:03.020 evidence that there's risk, but we don't have evidence that it's safe. And most drugs,

00:08:06.360 sit here. It's a big range, somewhere between 60 and 75%. Category D says, actually, we do have

00:08:11.680 some positive evidence of human fetal risk. And the drugs that sit in this category,

00:08:16.620 and I'm going to come back to this and give you examples of each of the drugs,

00:08:19.580 the drugs that sit in this category, you might say, well, why would a woman ever take a drug if

00:08:23.160 there's some evidence of risk to the fetus? And that is only if the risk to the mother not taking it

00:08:29.460 is greater. So the classic example here are seizure medications. So if you have a woman who is

00:08:34.480 going to be debilitated by seizures, and this is the only treatment she can have,

00:08:38.920 then a physician will typically make that decision. And again, very few drugs fit in this category.

00:08:42.980 It's typically 5% to 8%. And you have category X, which are drugs that have definitively been proven,

00:08:49.000 as much as you can prove anything in biology, to cause significant harm to the fetus, regardless of

00:08:54.000 benefit to the mother. And again, these are pretty rare, and this is 1% to 3%. So again, just keep in mind.

00:08:58.500 Now, where does Tylenol or acetaminophen fit into that? Well, it fits into category B. And to be

00:09:04.720 clear, for the last 10 years, there has been some concern about does Tylenol belong in category B?

00:09:10.100 Should it belong in category C? The other thing to keep in mind with Tylenol is you always have to

00:09:15.640 ask yourself about the switching cost or the alternative choices. And of course, a very common

00:09:21.580 alternative choice for Tylenol would be something like ibuprofen or an NSAID. Now, ibuprofen, Advil,

00:09:29.380 Aleve, for example, are considered category B in the first two trimesters, but bump to category D

00:09:36.400 in the third trimester for reasons I don't necessarily need to get into. But for those who are interested,

00:09:42.780 it has to do with the premature closure of a very small blood vessel that connects the aorta

00:09:47.700 and the pulmonary artery. And if that closes prematurely, it leads to premature delivery and

00:09:53.060 all sorts of things like that. So anyway, I just want you to kind of keep in mind why these categories

00:09:59.220 exist. And the reason we're walking through all of this now is to just sort of set the stage

00:10:03.320 for this discussion. So the goal today is not just to look at the potential link between acetaminophen

00:10:08.300 and autism, but also to put it in context so that we can hopefully end with, I think, the question that

00:10:14.100 at least some of you are asking, which is, okay, science aside, Peter, what's the bottom line?

00:10:18.920 If a woman is pregnant, should she be taking Tylenol? And I'm going to resist the urge to just give you

00:10:23.460 that answer right now, because I think it undermines the process of thought. So of course, to answer

00:10:28.480 that question, you have to not only take into account the possible effect of Tylenol exposure on

00:10:32.900 the baby, but also the health and the well-being of the mother, and also the possible effects of going

00:10:37.900 without Tylenol in the case of fever or inflammation, which is also associated with negative health

00:10:43.380 outcomes for babies that are exposed to those conditions in utero. We're going to take the same

00:10:48.480 structured approach that I basically try to recommend and utilize any time I've confronted

00:10:54.320 with an association between exposure X and condition Y. And that's true if condition Y is

00:11:00.740 positive or negative, good or bad. So I want to lay this out right now so that you know what we're

00:11:07.060 going to do and how we're going to land this plane. The first thing you want to be able to do is

00:11:11.340 confirm that there is indeed an association, statistically, okay? So a lot of times people

00:11:15.880 say there's an association, but there might actually not be. So you actually want to document

00:11:19.220 that statistically there is an exposure. So you want to verify that. The second question you're

00:11:24.520 asking is, of course, this is the hardest one, is the first one's pretty easy. If there is a

00:11:29.280 statistical association, you want to determine the likelihood that the association is causal.

00:11:35.020 Now included in this would be sensitivity analyses, falsification tests, and things of that nature.

00:11:40.100 Now notice I said, you're not trying to prove if the association is causal. Why? Because as I'm

00:11:46.100 sure many of you heard me say before, there are no proofs in biology. It's not like mathematics. You

00:11:50.960 don't get to write QED at the end of your work here. What we're really dealing with here is

00:11:55.220 probabilities. And we're trying to determine the likelihood of causality. Now, if the association

00:12:02.120 is believed to be more likely causal than not, then we have to ask the final question,

00:12:08.000 which factors into what do you do, which is we have to understand the effect size.

00:12:12.800 So you could have things that are causal, but the effect size is so small that it doesn't matter,

00:12:18.140 in which case your behavior is going to be quite different. So final point before kind of jumping

00:12:23.200 into this, it's very important to remember that we're discussing the state of science today. And

00:12:27.300 science is not about being right or wrong in an absolute sense. It's really about constantly updating

00:12:31.220 our priors, understanding the probability of something as new evidence becomes emergent.

00:12:36.780 And that's how we should really work. So as more and more data come online, we might have to revise

00:12:41.480 whatever views and conclusions I've come to here. Doing so is not a weakness, although tragically it

00:12:47.280 has become viewed as a weakness. Certainly if politicians change their mind about things,

00:12:51.920 that's viewed as waffling. But as scientists or as communicators of science, we shouldn't be afraid

00:12:57.220 of that. We should be open and acknowledge that as of today, this might be how we view things. And

00:13:02.180 in the presence of new information, we should be very receptive to changing that mind. So I know

00:13:06.640 that was a lot of background, normally far more than I would at the outset of a podcast, but I think

00:13:11.800 it's really important to have a shared foundation of knowledge and an understanding of the framework

00:13:16.920 before we wade into a topic that is not only scientifically complex, but obviously very emotionally

00:13:22.340 and politically charged. That's the lens I'm hoping to use for this discussion.

00:13:26.260 Now, with all of that said, I thought rather than just continue a monologue, it would be great to have

00:13:31.940 my co-host Nick Stenson from the AMAs join me and basically lay out this discussion with me in the

00:13:40.380 form of a Q&A such that it really reflects questions we're hearing and creates a bit of a storyline

00:13:48.020 through this. So Nick, thanks very much for joining me on very short notice.

00:13:52.000 Peter, I think as we get started, it would be helpful first to just even look at and lay the

00:13:56.940 foundation of what exactly are the claims being made about acetaminophen and autism?

00:14:02.960 The basic gist of the scientific claim is that maternal use of acetaminophen during pregnancy

00:14:09.160 is associated with an increased risk of autism in the exposed child. And this has prompted the

00:14:15.020 government to respond by asking the FDA to issue warnings to physicians and change the labels on

00:14:21.280 acetaminophen products with obviously Tylenol being the most common and the most familiar

00:14:25.180 to reflect the possible risk during pregnancy. But it's important to note that both the FDA and

00:14:30.020 the scientific community agree that we don't yet have evidence to assert that the apparent

00:14:35.220 correlations between prenatal acetaminophen exposure and autism risk reflect a causal relationship.

00:14:41.720 In other words, no authoritative sources are claiming that we can conclude from the existing

00:14:46.960 body of evidence that acetaminophen actually causes an increase in autism risk. Though some argue that

00:14:53.360 a causal relationship is plausible and others argue that a causal relationship is very likely and that

00:14:58.960 acetamin should therefore be avoided during pregnancy or used at most with strong precaution.

00:15:04.240 You mentioned at the offset that there could be a lot of reasons why we're seeing autism rates increase and not just a single thing.

00:15:12.400 What do we know about why there are so many things being linked to autism these days?

00:15:17.680 So actually there's two parts to your question, Nick, because on the one hand you're asking me

00:15:21.600 why are autism rates going up? We can't deny that. That's sort of like saying, is the sun coming up every morning?

00:15:28.520 But your second question is, why are there so many things being linked to it these days?

00:15:32.960 So I'm going to answer that question first because I think that's the more jugular question at the moment.

00:15:37.440 I promise I will get to that second question.

00:15:39.080 So why are so many things being linked to the enormous uptick in autism these days?

00:15:44.440 And it really comes down to a very, very understandable, rational, and logical desire,

00:15:51.000 which is a very strong motivation to look for the triggers of autism. We are looking for culprits, okay?

00:15:58.280 So autism rates have risen dramatically both nationally and even globally over the past few decades.

00:16:04.600 So according to the CDC, the prevalence of autism increased from 6.7 cases per thousand children

00:16:12.120 in the year 2000, just 25 years ago, to 32.2 cases per thousand children just three years ago.

00:16:20.360 That's a five-fold increase. Lots of explanations for this, which we will get to, but understandably,

00:16:27.160 that's not a subtle increase. And again, while some of that increase is due to an expanding diagnostic

00:16:32.960 definition and increased awareness, which we'll get into more later, as I said, there is no doubt that

00:16:38.560 some residual increase, even after accounting for these changes, is out there. And therefore,

00:16:43.680 in an effort to find these potential causes, a lot of research has been done to find potential

00:16:49.520 associations between autism and countless other variables. Now, all of that sounds great,

00:16:55.440 and all of that makes sense, but it poses a significant statistical problem. And this is

00:17:00.880 known as the multiple comparisons problem. If you look at enough variables, you are bound to find

00:17:09.200 statistically significant associations. This is the first example that I'm going to pull forth from

00:17:14.960 what I stated at the outset, which is we are not wired to think scientifically. If anybody out there

00:17:20.080 thinks they are smart enough that they can understand p-values out of the womb, more power to you.

00:17:26.160 And I majored in mathematics. I spent my life doing math and stats. This idea is not that intuitive

00:17:32.960 until it is explained to you. So it's understandable why what I'm about to say doesn't necessarily jump

00:17:39.440 to your mind as the explanation for this. Now, let me use an example. Imagine if you're trying to detect

00:17:45.920 if someone has psychic powers by having them guess the outcome of coin flips. You create the rule such

00:17:53.600 that if they guess correctly on at least seven out of 10 flips, which by the way, that's a 5% chance

00:18:02.480 somebody would do that based on pure luck with a fair coin. So if I had a fair coin and I flipped it

00:18:08.480 10 times, each of those has a 50-50 shot of heads or tails. But if you can guess correctly,

00:18:14.000 7 out of 10, 8 out of 10, 9 out of 10, or 10 out of 10, there's only a 5% chance of doing that,

00:18:18.480 I'm going to declare you a psychic. So that game is basically like a single hypothesis test with a

00:18:24.880 significance level of 0.5. So you see where I'm going with this. Now I'm setting this up like a

00:18:30.240 single hypothesis test with a significance level of 0.05 or a p-value of 0.05. Now suppose instead of

00:18:39.040 just testing one person, I'm going to test a hundred random people, all with fair coins.

00:18:44.560 And let's of course just assume for the purpose of illustration, there are no real psychics.

00:18:48.480 I realize there's going to be a little hatred directed toward me from people who believe they're

00:18:51.680 psychics. Well, by chance alone, I'm going to identify five people out of a hundred who are

00:18:57.680 going to pass the psychic test. And they're going to look like psychics. But the probability that at

00:19:02.880 least one person passes the test isn't just 5%, it's much higher because randomness can hit

00:19:09.600 anywhere across that group. So if you keep scaling this up to thousands of tests, like scanning genes

00:19:16.560 for diseases and links and running marketing experiments, I mean, anytime you're running

00:19:21.680 massive amounts of experiments, the odds of finding at least one false hit approach near certainty.

00:19:27.680 So you're essentially trolling through noise until patterns emerge by accident, like seeing faces

00:19:34.640 in clouds or winning a lottery. If you buy enough tickets, there's a great website called spurious

00:19:41.600 correlations. I've been playing with this website for a long time and went back to it recently that shares

00:19:47.600 examples of how easy it is to find significant correlations, even very strong correlations between

00:19:54.720 variables that clearly have nothing to do with each other, provided you're willing to look at

00:20:00.400 enough different combinations of variables. So for example, one very silly one is a 98.5% correlation

00:20:09.040 between the per capita consumption of margarine and the divorce rate in Maine. One of my personal

00:20:15.200 favorites, if you look at the number of physicists in the state of California and the ranking of Michael

00:20:22.080 Schumacher, when he was driving in F1 between the year 2003 and 2012, the correlation was 0.971, 97.1%.

00:20:33.120 But what's most interesting about this is that the site also demonstrates how easy it is to come up

00:20:39.360 with plausible sounding stories for why two clearly unrelated variables might be related. So they ask AI to

00:20:47.920 come up with a train of logic linking the variables. If you consider the example of the California

00:20:53.840 physicists and Michael Schumacher's success, AI explains that by saying that the rising number

00:21:00.320 of physicists in California drive innovation in the automotive industry, which leads to faster and more

00:21:07.680 effective race cars that propelled Michael Schumacher to higher rankings, which of course is ridiculous.

00:21:13.200 Peter, what do we know about why these ideas about associations with autism tend to persist,

00:21:19.760 even if the evidence can be shaky? I think it comes down to the fact that it is literally impossible

00:21:25.520 to disprove the link between any variable and autism, the way that you can disprove other things,

00:21:32.800 such as the earth being flat, or even things that are really complicated, like resveratrol extending

00:21:37.840 mammalian life where you have the luxury of doing randomized controlled experiment after randomized

00:21:43.280 controlled experiment after randomized controlled experiment, all of which fail. You have such a

00:21:47.360 high degree of probability that you've effectively disproved it, but we can't do that in epidemiology.

00:21:54.160 And I think that's why these ideas persist.

00:21:56.560 Going back to the recent news, what do we know about, was there anything in particular

00:22:01.920 that triggered the recent concern around acetaminophen and autism?

00:22:06.720 Not really, other than a publication that we'll talk about, but the idea that autism might be linked

00:22:12.800 to prenatal acetaminophen exposure isn't new. A handful of studies have reported

00:22:18.240 very small associations between that exposure and outcome over the past decade, more or less.

00:22:24.560 But the recent alarm was triggered by a systematic review of earlier research, which was published

00:22:31.520 in late August in a journal called BMC Environmental Health. Now, importantly, and I was a little surprised

00:22:37.920 to see this, this publication was not a meta-analysis, so they didn't pool the data from the studies to

00:22:44.480 re-evaluate the overall association or perform any new statistical tests. The authors of this paper just

00:22:50.400 collected all the relevant studies they could find on the relationship between prenatal acetaminophen

00:22:55.520 exposure and the risk of autism. They also looked at ADHD and some other neurodevelopmental disorders

00:23:00.800 in non-overlapping human cohorts, and they shared the basic study details and results in one place

00:23:06.720 and added some additional commentary. So basically, you can think of it as sort of a review article.

00:23:11.680 And so I think now it would be just helpful to just break down this paper in more detail for people.

00:23:16.160 So let's start with, what did the paper show?

00:23:18.400 Yeah. So in the case of autism, there were six observational studies that met

00:23:22.640 their criteria for inclusion. And the authors reported that these six studies, quote,

00:23:27.680 consistently reported a positive association between prenatal acetaminophen use and ASD,

00:23:34.080 autism spectrum disorder, with an exposure response relationship observed in four of the five studies

00:23:41.280 that evaluated the relationship. But this isn't actually true. In actuality, two of the six studies

00:23:48.000 showed no significant association between use of acetaminophen use during pregnancy and the risk

00:23:53.280 of autism in the offspring. And only three of the included studies directly examined the potential

00:23:57.920 dose-response relationship, while the fourth, by authors Xi and others, attempted to assess dose,

00:24:05.360 dividing the participants into tertiles, groups of thirds based on acetaminophen detected in a single blood

00:24:11.360 test. This method kind of had the advantage of using a quantitative biomarker instead of potentially

00:24:17.680 biased patient questionnaires that try to get at recall. But since the measurement was based on just

00:24:22.880 one sample taken during birth, it's a very poor indicator of overall exposure during pregnancy.

00:24:29.840 Acetaminophen is almost completely eliminated from the body within 24 hours.

00:24:34.640 So all the blood tests from the Xi study really tell us is whether or not a woman happened to take

00:24:40.480 Tylenol in the 24 hours leading up to delivery. Of course, a woman who had none before delivery might

00:24:47.040 have taken Tylenol for weeks on end earlier in her pregnancy, or a woman could have taken Tylenol with

00:24:52.880 delivery could have had none up until that point. So again, it's an interesting study, the Xi study,

00:24:58.800 which we're going to look at all of them in a moment. But I just want to point out that most of these

00:25:02.640 studies rely on questionnaires. And this study attempted to look at this biomarker, but obviously

00:25:08.240 it has significant limitations. Another point I'd say is that in the largest study examining a dose

00:25:14.480 response relationship, this was a study by lead author Alkfest, who the senior author on that was

00:25:21.040 Lee. And I'm going to come back to that in a second because Dr. Lee was interviewed this week by JAMA.

00:25:26.800 The dose response was only present in a partially adjusted statistical model.

00:25:32.640 Where it disappeared in what was called the fully adjusted model. And I'll talk about this in a

00:25:37.840 moment so you understand what I mean. But this suggests that the dose dependent association

00:25:43.040 between acetaminophen and autism, which is actually very important, was actually due to confounding

00:25:48.560 variables that weren't accounted for in the partially adjusted model, but were accounted for

00:25:54.400 in the fully adjusted model. So to take a look at this, I want you to look at the figure sitting next

00:26:00.400 to my head here, which is an analysis that my team pulled together by plotting the risk ratios

00:26:07.200 from the studies included in this analysis. I was surprised that this figure was not in the paper

00:26:13.600 because almost all review articles would do this and certainly a meta-analysis would have. But

00:26:17.600 nevertheless, they didn't. And so we've done this and you can feel free to check us if you like,

00:26:22.160 but we've taken all of the data out of their tables and simply put them into a pooled

00:26:26.800 table. And then the one thing we did at the end was pooled. So that's what's shown in red here.

00:26:31.440 So let me just orient you to this figure. So you've got all the names of the studies.

00:26:37.200 By the way, the Alka study is referred to as the Swedish study. So you've got at the very top,

00:26:41.840 you've got the sibling controlled version of the Swedish study, followed by the full cohort of the

00:26:46.400 Swedish study, followed by a set of other studies. And then the G, you're seeing two versions of this.

00:26:52.320 You're seeing the third tertile compared to the first tertile and the second tertile compared to

00:26:58.560 the first tertile. You then have a couple of these other studies and then you can see the summary in

00:27:02.880 red is where we're showing the pooled data. Now, let me remind you how to interpret these lines and

00:27:07.200 bars and things like that. The unity line represents absolutely no risk. Anything to the right of the

00:27:15.360 unity line would represent an increase in risk. Anything to the left of the unity line would represent a

00:27:20.720 decrease in risk and the bars represent the 95% confidence intervals. In an ideal world,

00:27:27.440 you're looking to see dots that would be quite far from the unity line one way or the other.

00:27:32.240 Peter, can you walk people through what this chart is showing us? Because one,

00:27:36.640 a lot of people haven't looked at or interpreted charts like this before. And two,

00:27:40.720 we also have people listening, not watching. So for those people, this will be in the show notes,

00:27:45.040 which will be available to all as well. But can you walk through what the big insights that you and the

00:27:49.920 team had from this chart? In this chart, you can see that the overall association between acetaminophen

00:27:57.280 use and autism is very small, corresponding to just a 5% increase in relative risk between exposed and

00:28:05.200 unexposed children. But there are a few other details that jump out at you when you're looking

00:28:10.400 at this. So the most obvious feature, at least to me, is that there is a very strong association coming

00:28:17.200 from one very small study. In fact, the smallest study here, which is the Xi study from 2020.

00:28:23.120 But as discussed a minute ago, this study was done very different from the others. So instead of

00:28:29.440 comparing the risk between children who were versus were not exposed to acetaminophen during gestation,

00:28:36.880 they then just divided the participants into three groups or tertiles based on the concentration of

00:28:42.720 acetaminophen that was detected in the samples of umbilical cord blood. So then you could compare

00:28:48.000 the risk in the first to the second and the first to the third tertiles, respectively. So again, this

00:28:56.080 might eliminate the issue of recall bias, which of course is a real issue as well, where you have to

00:29:00.960 ask a woman after she has the baby, how much Tylenol did you take during your pregnancy, when, etc. But again,

00:29:06.480 it still has a pretty big issue, which is acetaminophen does not stick around very long

00:29:13.120 in the blood and therefore we don't really know how much the acetaminophen levels identified in cord

00:29:21.200 blood at the time of delivery really tell us anything about the amount of acetaminophen that the woman

00:29:26.800 took during pregnancy. Now, some have pointed out that this isn't a concern. Since a study published

00:29:31.360 this past January by another group reported a positive correlation between three levels of

00:29:38.000 self-reported acetaminophen use throughout pregnancy, so they called it non-use, less than 14 days or more

00:29:43.840 than 14 days, and the level of acetaminophen detected in cord blood at birth. However, the correlation was

00:29:49.760 not that great. It was 72% and involved a completely distinct cohort. So there's certainly a lot of room

00:29:57.040 for error based on that. Additionally, the Xi study mentions that all cord blood samples contained

00:30:04.160 detectable levels of acetaminophen, but they don't actually report how those levels differed across

00:30:09.520 the tertiles, either averages or thresholds. So we have no idea how much acetaminophen we're actually

00:30:15.360 talking about here. However, they do report that 70% of the samples had no detectable levels of acetamin

00:30:22.320 metabolites, which would strongly suggest that the majority of participants had very minimal levels

00:30:28.480 of acetaminophen exposure, such as what you might see through drinking water. This essentially means

00:30:34.240 that the comparison between the second and first tertiles was comparing sub-therapeutic exposure to

00:30:41.200 sub-therapeutic exposure, telling us virtually nothing. And as further evidence of this, that paper I just

00:30:48.320 referenced that looked at comparing cord blood levels to recall, remember it divided them into

00:30:53.840 nothing up to 14 days, more than 14 days, it found acetaminophen in all cord blood samples,

00:31:00.000 yet it showed that tertiles one and two were statistically identical. The suggestion here is

00:31:04.720 there's some low level of hermetic Tylenol or Tylenol metabolite that we're probably all exposed to

00:31:10.640 that doesn't really constitute an exposure. Another issue, and I actually think this is the single

00:31:15.920 biggest issue by the way, seems to be the biased participant inclusion. So the participants in the

00:31:22.160 G study were enrolled in 1998 and followed for another 20 years. And anybody who dropped out before

00:31:28.960 the end of that 20 year mark was excluded. But think about that for a moment. People are more likely to

00:31:34.800 stay in an extended longitudinal study if they have a personal interest in the results. For instance,

00:31:40.400 if they find their child has autism. This is likely to affect studies on this subject, but it seems

00:31:47.200 especially pronounced in the G paper as this study, which was based in the US reported rates of autism in

00:31:55.200 the participants at roughly 11%. So meaning when you look at all the participants who completed that study,

00:32:02.480 11% of the kids had autism. But for context, the general population currently at that 5x increase

00:32:11.120 is 3%. And at the time of enrollment, it was 0.7%. The enrollment was between 98 and 2004. So the cohort

00:32:19.600 used for this, the G study was the Boston birth cohort, which originally enrolled almost 9,000 mother-child

00:32:26.720 pairs. But among them, preterm births were quite overrepresented over 35%. By 2018, approximately 3,000

00:32:37.760 dyads remained in the active follow-up cohort, presumably due to loss on follow-up. And of those

00:32:45.200 eligible dyads, slightly less than a thousand had available umbilical cord plasma samples and complete

00:32:51.760 outcome data. So again, we have kind of a concentration, it would seem, of cases due to

00:33:00.800 all the reasons I just stated. Now, but one of the most important things to take away from this graph

00:33:06.160 is the fact that the risk estimate summarized from all the studies, what we're sort of calling the

00:33:10.880 pooled one here in red, is virtually identical to the risk estimate derived solely from the 2024

00:33:18.160 Alkvist study, which was done in Sweden. And so it's commonly referred to, and you've probably heard

00:33:23.520 of it referred to this way as the Swedish cohort study. And the reason they're essentially identical

00:33:28.720 is that that Swedish study included more than 10 times the number of participants than all other

00:33:34.960 studies combined. I know that's a lot there. I think it's actually worth just summarizing that

00:33:39.440 again before we move on to the Swedish study, because if you looked at this graph that we put together,

00:33:44.000 you could very easily come to the conclusion that the Xi study is indeed the smoking gun.

00:33:48.240 But again, you have to remember the limitations of this. And there are several, right? One is the

00:33:52.480 sample size is incredibly small relative to the others. Two, you have the concentration effect,

00:33:58.560 where based on the nature of the study, you concentrated and disproportionately counted cases

00:34:05.040 of autism versus non. And then of course you have the collection methods, where using this single

00:34:10.880 sample of cord blood, which may have some association with maternal use during pregnancy,

00:34:16.080 but is very unlikely to account for the actual nuanced differences in dose and exposure during

00:34:23.520 pregnancy. Again, think about that through the lens of any other thing. If I could only measure

00:34:28.080 how many donuts you ate on your birthday, it would be very difficult for me to impute how many donuts

00:34:33.520 you eat over the course of a year. Would it have a correlation? Probably. But it wouldn't be strong

00:34:38.720 enough to take to the bank if I was trying to use donut consumption as a marker of predicting heart

00:34:43.840 disease. So again, it's very tempting when you look at these meta-analyses, or in this case,

00:34:49.520 even just a review article, to think more is better. But remember, a thousand sow's ears

00:34:54.880 makes not a pearl necklace, quoting the great James Yang, who used to be one of my mentors in the lab.

00:35:00.560 Given how big the Swedish cohort study was, I do think it's worth spending some time to really break

00:35:05.840 that down. So can you walk people through that study and what it found in more detail?

00:35:10.960 Yeah. The Swedish study was a very large prospective cohort study, and the general results indicate a

00:35:16.320 small correlation between acetaminophen use by the mother during pregnancy and later life ASD in the

00:35:22.960 offspring. So there were just under two and a half million Swedish children included in the full

00:35:28.480 cohort. And the primary exposure metric was ever use of acetaminophen in pregnancy, with dose serving

00:35:36.640 as a secondary metric. So again, the primary outcome is binary, either you ever used acetaminophen or you

00:35:42.560 did not. So acetaminophen use was determined through a combination of prescriptions, because again,

00:35:47.840 single health care system, they have access to all the prescriptions, and also through maternal

00:35:52.720 interview with midwife or physician throughout the pregnancy. They don't specify the number of

00:35:58.080 interviews and it probably varied across participants. So over a median follow-up of about 13 and a half

00:36:04.240 years, the general cohort showed a very small but statistically significant positive association

00:36:11.360 between prenatal acetaminophen exposure and autism. The hazard ratio is 1.05 and the confidence interval was

00:36:19.440 1.02 to 1.08. So what does that mean? That means it showed a 5% increase in relative risk and the

00:36:29.760 confidence interval of 95% confidence was significant because it did not cross the unity line. So anytime the

00:36:37.360 error bars do not cross the unity line, it's statistically significant. Remember, just going back to the

00:36:42.720 studying study stuff we talked about, we always like to calculate an absolute risk exposure if we can.

00:36:47.920 So the relative risk was a 5% increase, the absolute risk increase was 0.09% increase at 10 years. So that's

00:36:57.520 a very small absolute risk increase, less than one-tenth of one percent. So the researchers then examined risk

00:37:05.040 specifically in a cohort subset that was composed of matched sets of full biologic siblings. So the authors

00:37:14.000 examined sibling pairs that were discordant in acetaminophen exposure and found no significant

00:37:20.720 difference in risk for autism between exposure and lack of exposure. So why do this? The logic here is

00:37:28.560 similar to any matched cohort study, but instead of merely matching based on general characteristics like age

00:37:36.960 or sex, each exposed individual is matched to an unexposed sibling. This means that exposed and

00:37:44.720 unexposed groups on the whole should be relatively evenly matched in terms of several confounding

00:37:51.360 variables related to home environment and even many genetic factors. Each person in one group

00:37:57.440 could be mirrored by somebody in the control group. So there shouldn't be any systematic differences

00:38:02.880 between the groups. Can you walk people through what happened when they did the more detailed

00:38:07.360 sibling analysis? Yeah. So when they did that concordant discordant analysis, the correlation was

00:38:13.760 entirely abolished when they compared and controlled for family environment and genetics as best as you

00:38:20.240 could. Remember, this was not an identical twin comparison. It was just siblings. But obviously,

00:38:25.440 this is the best control you could get. And this suggests that the apparent link observed in the full

00:38:31.600 cohort was likely due to confounding factors. Given these results, the authors of the Swedish study

00:38:39.760 came to the same conclusion. They stated, quote, results of this study indicate that the association

00:38:46.080 between acetaminophen use during pregnancy and neurodevelopmental disorders is a non-causal

00:38:52.000 association. Associations observed in models without sibling control may be attributable to confounding.

00:39:00.320 Now, it's important to note that the review article that came out in August, in their analysis,

00:39:07.680 they state that the Swedish study only included siblings that were discordant for both exposure

00:39:15.120 and outcome. But this was not the case according to the Swedish study's senior author. And such a design

00:39:21.760 would introduce what's known as a collider bias, where the selection criteria create a situation

00:39:28.400 where the exposure and outcome are already related in some way. To illustrate why this double discordance

00:39:35.920 selection doesn't work, consider a very extreme example. Imagine autism can only occur with acetaminophen

00:39:43.280 exposure, but that acetaminophen exposure does not guarantee autism. So in biological parlance,

00:39:49.840 we would say acetaminophen is necessary but not sufficient. If you select only pairs that are

00:39:56.800 discordant for both the exposure and the outcome, you would exclude all cases in which acetaminophen

00:40:04.720 exposure did not result in autism, even if those sibling pairs accounted for the majority of sibling

00:40:11.520 pairs discordant for the exposure. In other words, you would falsely conclude 100% risk.

00:40:18.560 Therefore, I believe that Lee, the senior author of the Swedish study, is correct in his assessment,

00:40:26.240 which is that once you correct for genetics and home environmental exposures, the risk of autism

00:40:37.040 as it pertains to acetaminophen exposure is not causal. Stated another way, acetaminophen exposure

00:40:44.080 prenatally in the Swedish cohort does not appear to be causally related to autism.

00:40:50.640 Looking at what you just looked at, is there any reason to question those results and question what that

00:40:55.840 study said? Another way to think about it is, do people who argue in favor of a potential link between

00:41:01.200 acetaminophen and autism have anything to say about those results?

00:41:05.260 Yes, you should question everything. So I would say that one of the criticisms that's been leveled

00:41:11.340 against the Swedish study is that the overall rates of acetaminophen use in that study were much

00:41:17.420 lower than are observed here in the US or elsewhere in the world. Only about seven and a half percent of

00:41:23.980 the participating mothers in the Swedish study were consumers of acetaminophen, whereas some studies

00:41:31.260 have reported up to 50% of mothers using acetaminophen during pregnancy. And again, given this discrepancy,

00:41:38.220 some have argued that the generalizability of the Swedish study is limited, which is interesting.

00:41:44.860 And as it happens, you couldn't make this up. Another large cohort study with a similar nested

00:41:52.380 sibling analysis was just published a couple of weeks ago. This was after the August publication

00:42:00.220 of the review article, and it supports the findings of the Swedish study. The new study was conducted

00:42:06.700 in a nationwide Japanese population and consisted of almost 220,000 children, of which almost 40% were

00:42:16.860 exposed to acetaminophen during gestation. So very similar to the rates we see in the United States and

00:42:23.900 in some of the other high exposure studies. The associations reported from this Japanese cohort

00:42:30.780 were similar to those in the Swedish study. In the general cohort, so unadjusting for siblings,

00:42:37.660 prenatal exposure to acetaminophen was associated with a 6% uptick in autism rates. Recall in the Swedish

00:42:45.500 study, it was 5%. But in the Japanese study, this did not reach statistical significance. So the confidence

00:42:52.380 interval, the 95% confidence interval crossed the unity line. It was 0.98 to 1.15. But when they did

00:42:59.340 the sibling analysis, even this small trend towards an increased risk was completely abolished. So when

00:43:07.100 you take this Japanese study of nearly 220,000 children and pair it with the Swedish study of two and a half

00:43:17.020 million children, and both of them, when done by this method, abolish any causality, it's very difficult

00:43:25.340 to make a strong case for causality. We've talked about this before, but I think anytime we're this

00:43:30.220 deep into science, it's good for people to kind of step back. And so can you walk people through why

00:43:35.180 it's so hard to make assumptions about causality based on just observational data?

00:43:41.020 Yes. Again, I think you can think back to the sort of spurious correlation site that I was talking

00:43:46.460 about earlier. It really comes down to the potential influence of confounding variables that we are

00:43:53.180 blind to. That's basically what it comes down to. Dr. Lee, the senior author of the Swedish paper, was

00:44:00.620 interviewed by JAMA this week. It's a great interview. I think it's worth reading. We'll link to it. He talks

00:44:05.740 about a great example that I'm sure many people have heard. I'd certainly heard it before in a statistics

00:44:10.300 class, but it's worth repeating. It's the example of the strong correlation between ice cream consumption

00:44:17.820 and drowning. So as we see rates of ice cream consumption go up, we see drowning deaths go up,

00:44:23.980 and as one falls, the other falls. Obviously, if you were being cheeky, you would say somehow eating ice

00:44:30.120 cream is causing people to drown. But of course, there's a confounding variable, and the confounding

00:44:35.500 variable is heat. The warmer it gets, the more people are likely to eat ice cream, and separately, the more

00:44:42.940 people are likely to swim. And therefore, it's this confounding variable that isn't immediately obvious

00:44:49.020 that explains both of these things. I think that's really the challenge of epidemiology, and I don't say that as a

00:44:56.140 knock on epidemiology, I say it's the legitimate challenge. It's that you can never, ever, ever identify

00:45:02.780 all of the confounders, and therefore, you are always at the mercy of wondering, is there something

00:45:09.500 I'm not seeing here that is what is actually explaining the causality? The only way to show

00:45:16.100 causation, unfortunately, is through randomized trials. That's the only way you can really be as close

00:45:21.300 to 100% sure that you've established causality by doing a well-controlled randomized control trial. But

00:45:27.680 unfortunately, some questions do not lend themselves to that for either ethical or logistical reasons, and

00:45:32.880 clearly, this question, the use of acetaminophen in autism is one of those tricky questions. So we're not going to

00:45:39.120 get an RCT to do this, and instead, we're going to have to glean what we can as best we can from

00:45:45.540 epidemiology. And that's where I think we get to this set of guidelines that I talked about at the top of the

00:45:51.080 show here called the Bradford Hill criteria. So the Bradford Hill criteria are a set of nine principles

00:45:57.360 used to assess whether an observed association is likely to reflect a true causal relationship. So

00:46:04.180 Sir Austin Bradford Hill in 1965 put forth these criteria to help epidemiologic researchers examine

00:46:13.540 their data when RCTs were not available. So they consider factors like strength, consistency,

00:46:19.560 specificity, temporality, biological gradient, biological plausibility, coherence, experimental

00:46:27.820 evidence, and analogy. So Peter, I think what would be helpful now is let's just go through the state

00:46:33.280 of the evidence for each of those criteria. So looking at acetaminophen and autism, what's first on the

00:46:40.420 list? Yeah. So let's just start with strength. How large is the effect? And the larger the effect,

00:46:46.760 the more likely it is to be causal. Again, we can talk about a few obvious and famous examples. The

00:46:52.120 example of smoking is perhaps most notable. I've never met a person who doesn't understand or disputes

00:46:58.680 the exposure relationship between cigarette smoke and lung cancer. There's nobody out there making the

00:47:04.680 case that we need an RCT to determine that. We don't have an RCT. And why is it? It's because if you run

00:47:10.000 the smoking lung cancer data through the Bradford Hill criteria, it pops on many levels, but effect

00:47:15.800 size is probably the biggest. We're talking about an effect size of 10X. So 10X is just a magnitude

00:47:23.320 beyond what we normally would find in most biologic associations. By comparison, the effect size here is

00:47:29.200 1.05X. That's what a 5% relative risk increase is. So this is smaller than associations that have been

00:47:39.200 reported for many other things that we actually know are probably almost assuredly not causal based on

00:47:45.680 more well-to-do data, such as the association between red meat consumption and type 2 diabetes,

00:47:50.920 which is a 1.10 or a 10% relative risk increase per 100 gram per day increase in red meat. Of course,

00:48:00.040 we've argued ad nauseum that those associations are almost assuredly picking up a confounder,

00:48:06.500 which is healthy user bias, or even poultry consumption and the risk of type 2 diabetes,

00:48:11.420 1.08. Again, both of these are stronger associations here. An even clearer example would be

00:48:18.300 the meta-analysis of observational studies that reported that a higher leisure time physical

00:48:24.780 activity was linked to a 5% higher increase in prostate cancer. Again, we know that that is

00:48:31.960 completely nonsensical, but that is what you get when you go trolling for signal in a sea of noise.

00:48:39.360 You will eventually find it. In other words, for this, the effect size is very weak. In other words,

00:48:45.000 when the effect size is weak, and here we're defining weak as a subset of the type of epidemiology we're

00:48:51.480 looking at, and in this case, we're looking at pharmacoepidemiology versus, say, nutritional

00:48:55.040 epidemiology or toxicology epidemiology, weak is generally regarded as 1.5 for pharmacoepidemiology,

00:49:02.140 and the reason for it is just based on the pervasiveness of bias throughout these studies.

00:49:06.180 So when you're showing up at 1.05 and the threshold for interesting is 1.5, you're well below it, I would

00:49:14.840 say we don't do very well on the strength here. Moving on to the next piece of the criteria,

00:49:20.360 consistency. How consistent are the data linking autism and acetaminophen?

00:49:26.160 I would say actually they're reasonably consistent. So consistency obviously just means how often does

00:49:31.180 this show up across multiple studies and maybe even across multiple populations and different

00:49:35.700 methodologies. I would say a handful of prospective cohort studies have reported this positive

00:49:40.660 association. But remember, these associations tend to go away when you control for family environment

00:49:47.020 or genetics, as we've seen in the two largest studies here that we've talked about. So in general

00:49:52.980 population studies, i.e. no sibling control, the reported associations have varied somewhat in

00:49:58.500 magnitude. Some have shown little to no association, but I would say more often there is some association.

00:50:05.720 Moving on to specificity. What do we know about specificity?

00:50:09.800 Yeah, so specificity is asking the question basically how specific is the cause-effect relationship?

00:50:16.240 So if an exposure is associated with only one outcome or an outcome associated with only one

00:50:22.320 exposure, a causal relationship is more likely. So here it's very non-specific because there are

00:50:28.300 many variables that have been linked to autism risk. Many with much stronger lines of evidence than

00:50:34.920 acetaminophen. For example, advanced paternal age, premature birth, air pollution exposure,

00:50:41.420 and heavy metal exposure. So these are all variables that have much stronger associations with autism,

00:50:48.160 so we're not dealing with a one versus one. And some lack of specificity also exists in the other

00:50:54.040 direction. There's observational studies that have also reported associations between acetaminophen use

00:50:59.820 and ADHD and language development. So again, what you really want to look for if you want to check the

00:51:06.440 specificity box is a one-to-one mapping. It's not a deal breaker not to have it. Cigarette smoking can

00:51:12.080 cause lung cancer and other cancer and heart disease, and that doesn't necessarily by itself at all

00:51:16.780 diminish the fact that it causes lung cancer. Moving next on the list, we have temporality.

00:51:21.460 Does the exposure precede the reported effects in this case?

00:51:25.940 It does, and certainly to a first order it does, meaning acetaminophen use comes before autism.

00:51:31.740 But there really hasn't been a consensus on the impact of timing of exposure or critical

00:51:37.140 windows in which gestational exposure might be more problematic than others. And if you look at the

00:51:43.580 two researchers most known for their belief that acetaminophen exposure raises the risk of ASD,

00:51:51.460 Andrea Baccarelli and William Parker actually have conflicting views on the critical window of

00:51:57.160 acetaminophen exposure. Baccarelli, who is the author, by the way, of the paper that came out in

00:52:02.500 August that we've been talking about, he believes that maternal use during any part of pregnancy

00:52:08.420 increases the risk to the fetus, while Parker believes that prenatal exposure carries relatively

00:52:13.600 little risk provided the mother has a healthy liver to process the drug. Parker instead argues that the

00:52:19.140 greatest risk comes with exposure in the neonatal period or even during birth itself, basically

00:52:25.680 starting from the time the umbilical cord is clamped and onward. What do we know about dose

00:52:30.560 dependency in this case? Yeah, so dose dependency, which you could also think of as biological gradient,

00:52:36.780 says the more you have the exposure, the more you should see the outcome. And some studies have

00:52:42.980 reported modest dose dependency based on the amount of time over which the mother was taking acetaminophen

00:52:49.100 during pregnancy. But the results, again, have been pretty inconsistent. Is there a plausible

00:52:54.400 biological mechanism for exposure that might cause an effect? The mechanism of action for acetaminophen is

00:53:02.380 generally pretty poorly understood. It's kind of amazing that we don't understand how such a ubiquitous

00:53:07.000 drug actually lowers temperature and alleviates pain. So therefore, we don't really have much clarity on

00:53:13.100 how it might ultimately lead to autism. That said, its effects are mediated at least in part through

00:53:19.540 inhibition of the synthesis of prostaglandins, which are molecules that contribute to pain

00:53:25.240 and the inflammatory response. And since prostaglandins also play a role in neurodevelopment,

00:53:31.280 some researchers have argued that acetaminophen leads to autism by disrupting

00:53:35.980 neurodevelopmental pathways. So again, there's no clear evidence of this, but there is at least what

00:53:42.740 we would call biological plausibility, even if at best it might be a little bit hand wavy.

00:53:49.020 This, of course, also leads to another criteria, the next one, which is analogy, where we compare the

00:53:55.440 current body of evidence to another similar intervention with a more established effect. And I think here

00:54:01.300 we can look at the effect of another prostaglandin inhibitor in the CNS, which is aspirin, which is

00:54:08.860 shown to have modest protective effects against autism-like symptoms in animal studies. So this

00:54:14.920 potential protective effect was also seen in the Swedish cohort study that we talked about earlier,

00:54:19.960 in which sibling analyses showed a small but statistically significant reduction in autism risk

00:54:26.620 with prenatal aspirin exposure. This was about a 13% relative risk reduction. In other words,

00:54:34.680 the analogy criteria would actually argue against it based on the dual inhibition of prostaglandins

00:54:42.360 between both of these drugs. And then wrapping this section, looking at the Bradford Hill criteria,

00:54:47.880 do you want to cover the last two? Yeah, the last two don't really help us much here because one of

00:54:53.720 them is on whether or not we have intervention-based evidence to support these conclusions, but

00:54:58.100 obviously we don't. We don't have randomized control trials that can point to sub-analyses here.

00:55:04.460 As an example of where we would be able to use this, if you're looking at exercise epidemiology or

00:55:09.560 nutrition epidemiology, you might not be able to answer the meta question with epidemiology,

00:55:15.240 but you could do short-term well-controlled studies to show that, for example, like six months of

00:55:20.960 exercise improved blood pressure, then you'd be more likely to believe that exercise could reduce

00:55:25.860 the risk of cardiovascular disease if that's what the large epi showed. But again, we can't do the

00:55:29.860 short-term studies here. And then, of course, the final metric here is called coherence, which is

00:55:35.100 how do we tie the observational data with the in vitro and in vivo testing? And while we have some data

00:55:41.520 here, they're very inconsistent to the question. There are some studies that involve pre- or perinatal

00:55:47.540 acetaminophen exposure in mice and rats that have reported a few neurodevelopmental abnormalities,

00:55:52.820 but they've been very inconsistent in the nature of the effect, and many have aligned quite poorly

00:55:57.780 with the characteristics of autism. For example, one study showed minor alterations in spatial learning

00:56:04.420 and locomotor activity, which aren't typically associated with ASD, but not anxiety-like behaviors which

00:56:10.940 often do accompany ASD. Additionally, some of the studies have used extreme doses far exceeding the

00:56:16.780 therapeutic doses used in humans as adults or children at all. What I think might be helpful

00:56:22.000 is if you could just quickly summarize all the information we just talked about as it relates

00:56:27.160 to acetaminophen and autism in looking at the Bradford Hill criteria. Okay, so let's go through

00:56:33.520 them one by one. Strength, definitely weak. Consistency, moderate. Specificity, weak. Temporality,

00:56:41.440 I would say modest and probably even strong. Biological gradient, I would say moderate.

00:56:46.660 Plausibility, I would say weak. Analogy actually provides evidence against this effect. And then

00:56:54.320 obviously experiment and coherence. We don't really have meaningful data, but if we do, I would classify

00:57:00.020 coherence as probably somewhat weak. Based on that, where do you land on looking at autism and

00:57:07.020 acetaminophen? Well, again, to the first question I posed, which is, is there even a statistical

00:57:12.860 association? I would say possibly. Obviously, there is in an uncorrected or unadjusted analysis,

00:57:20.400 but I'm really trying to refer to these adjusted analyses. So I would say, yes, there's probably

00:57:25.280 some association between acetaminophen and Tylenol. It's not particularly large, but let's assume it

00:57:31.600 is there. The important question, and the only question that really matters here is, what is the

00:57:36.040 probability that that association is causal? And based on everything we've just talked about,

00:57:41.220 inclusive of the running of the Bradford Hill criteria, I would say the probability that the

00:57:46.460 association between acetaminophen use by a mother and the development of autism of her child is a very

00:57:53.400 low probability event to be causal. Again, let me restate that. What that means is, I think the

00:57:58.460 probability that if a woman takes Tylenol during pregnancy, it's going to increase the probability

00:58:04.960 probability that her child has autism is very low. And I'm sorry for using the word probability twice

00:58:09.540 in one sentence, but that's the challenge of trying to talk about this thing technically and accurately.

00:58:14.300 I hope that makes sense. I'll clarify it if it doesn't. No, I think it does. And I think what

00:58:18.800 would be helpful now is kind of stepping back. So early on, you mentioned that one of the things we

00:58:23.280 do know is that there is an increase in cases of autism. Let's assume there is causality here.

00:58:30.860 Is it enough to explain what we opened with, which is a five fold increase in the prevalence of autism

00:58:39.660 today? I think the answer is unquestionably no. That's a much more confident thing that we can say

00:58:46.880 that there is essentially zero chance that maternal Tylenol use is the thing, quote unquote,

00:58:55.740 the thing in quotes, responsible for the rise in autism. So if it plays a role, it would be a very

00:59:02.520 small role and it would have to be in the setting of another susceptibility. Again, I still would argue

00:59:08.380 that it is not playing a measurable role based on everything we've discussed.

00:59:13.140 And so Peter, I think now would be a good time to like take a look at some of those

00:59:17.900 important risk factors. So if you look at autism, what are some of the most important risk factors

00:59:23.920 when it comes to that? Well, this is something I started looking into probably three or four years

00:59:30.260 ago. So it was kind of actually nice to kind of go back and brush up on this literature and see what

00:59:35.340 had been updated. But the long and short of it is genetics play a much larger role in autism risk

00:59:41.420 than all other variables combined and account for an estimated 80 to 90 percent of the inter-individual

00:59:52.020 variability in autism risk. The term for that is heritability. So the heritability of a trait

00:59:58.580 can be assessed through studies that compare monozygotic twins, so identical twins, and dizygotic

01:00:06.540 twins. So these are fraternal twins. This can also be done comparing what are called concordant-discordant

01:00:13.900 identical twins or monozygotic twins where you take identical twins that are raised in different

01:00:17.580 environments. So there's lots of elegant ways to do this. What do we know? We know that monozygotic

01:00:21.680 twins are obviously genetically identical, whereas dizygotic twins are genetically no more closely related

01:00:28.220 than any other pair of siblings. However, all twins are exposed to the same in utero environment.

01:00:34.960 And in most cases also raised in the same environment. I mentioned that there are some studies that do

01:00:41.580 look at identical twins raised apart, but let's put that off to the side. This means that you can assume

01:00:48.260 that dizygotic twins differ mostly in genetics, whereas monozygotic twins don't really differ at all.

01:00:58.200 You have very elegant what we call natural experiment. So if we see that a given trait is highly

01:01:04.440 correlated between monozygotic twins, but is often discordant between dizygotic twins, it must have a

01:01:12.920 very significant impact from genes. So I just want to pause before I go any further because so much

01:01:19.760 of what I'm about to say hinges on that. So Nick, did that make sense? Do you want me to explain this

01:01:25.780 beautiful natural experimental tool that we have? Yeah. I think in this case, it is worth

01:01:31.960 double checking and just reconfirming things so people understand because it is such an important

01:01:37.660 point. And it's come up on so many podcasts we've done in the past. I can think off the top of my head

01:01:42.560 of three guests we've had on the drive over the past five or six years where we have talked about

01:01:48.640 the heritability of various things. They're almost always neuropsychiatric. So the heritability of

01:01:55.320 bipolar disorder, schizophrenia, major depressive disorder. Okay. So how are they figuring this stuff

01:02:00.880 out? If you have identical twins, they are in the mother at the same time. Therefore they are exposed

01:02:08.600 to all of the same things while the mother is carrying them. And let's just again, limit this to all

01:02:15.100 twins that are raised together, which most are, then they come out and they're also exposed to the

01:02:20.460 same environment. If you have dizygotic twins, they're just siblings. They're genetically obviously

01:02:26.760 similar, but not identical, but they were exposed to the exact same environment inside the mother.

01:02:32.160 But then once they're born, they're exposed to comparable things outside. So if we see that a trait

01:02:37.980 is highly correlated only in the monozygotic twins, but the correlation is nowhere near as strong in the

01:02:44.920 dizygotic twins, then we know that genetics are playing the role. So let's talk about two things

01:02:51.440 that everyone will appreciate, height and body weight. Height has approximately an 80% heritability.

01:03:00.060 This shouldn't be surprising to people. We understand that on average, tall parents have tall kids and

01:03:06.600 short parents have short kids. Is it perfect? Not at all, but it's 80% heritable. Body weight,

01:03:12.400 also quite heritable, though not as much. It's about 60% heritable. So that's what we mean by

01:03:19.400 heritability. Now, in one of the studies that was included in the August paper, the review paper,

01:03:26.860 and this was the Leppert paper, the primary study actually focused on how acetaminophen used during

01:03:33.020 pregnancy correlated with the mother's genetic predisposition for autism. And they didn't find any

01:03:39.800 significant association, but if they had, it might suggest that a woman's genetic predisposition

01:03:46.720 towards autism might be the real variable behind the apparent association between acetaminophen and

01:03:53.660 autism within the offspring. If the mother is predisposed towards autism, then the child is also

01:03:59.340 likely at a higher than average risk of autism based solely on genetics. But if a genetic predisposition

01:04:06.980 also increases the likelihood that the mother might use Tylenol during pregnancy, which is entirely

01:04:12.460 possible given that autism is related to sensory perception, which is in turn related to pain

01:04:18.720 sensing, then it would appear as if acetaminophen use and the child's risk of autism were related,

01:04:25.960 even though both associations might actually be explained by genetics. And this is what I referred to

01:04:32.400 above when I talked about a sort of middle confounding variable. The example I gave earlier

01:04:38.920 about the temperature being the thing that relates ice cream consumption and drowning. In other words,

01:04:46.780 genetics would constitute a confounding variable that influences both autism risk and acetaminophen use,

01:04:53.220 just as temperature is the confounding variable that influences both ice cream consumption and drowning.

01:05:00.260 If so much of autism risk is genetics, what can we say about genetics explaining the increase in autism

01:05:09.240 rates over the past few decades? They definitely don't because genetics do not shift enough over

01:05:15.480 those kinds of timescales to explain this five, six, or potentially even seven fold increase in autism

01:05:22.020 diagnoses that we've seen over basically, let's just call it two generations if you want to go back

01:05:27.260 enough. Now, some cases of autism do involve denovo mutations, but the majority of this increase

01:05:32.460 seems to be explained by the increased awareness and expanded diagnostic definitions. So let's review a

01:05:40.360 little bit of history here. There has been a progressive expansion of the diagnostic criteria

01:05:45.780 for autism over the last 40 years. In 1987, the DSM-3 made a revision which expanded from a strict

01:05:54.520 infantile autism diagnosis or definition where the symptoms must occur between 30 months of age

01:06:01.960 to something called autistic disorder, which was defined by a checklist of symptoms that could

01:06:07.700 manifest well beyond infancy. Then in the 1990s and into the 2000s, a series of revisions in the DSM-4

01:06:17.220 and the ICD-10 created something called the pervasive developmental disorder family, the PDD

01:06:24.380 family, which encompassed autistic disorder, Asperger's disorder, something called PDD not

01:06:30.440 otherwise specified, which I talked about on the podcast with Trenna, it sort of became the all else

01:06:36.000 bucket, Rett's disorder, and then something called childhood disintegrative disorder where kids actually

01:06:41.640 go on to lose an already acquired skill. So if they acquire a language skill, but then go on to lose it.

01:06:47.220 So further expands this recognition, but with very inconsistent boundaries between the subtypes.

01:06:52.920 The age of onset was typically before three years of age. Then in 2013, the DSM-5 collapsed all the PDD

01:07:05.000 subtypes into a single diagnosis called autism spectrum disorder or ASD. It also relaxed the before age

01:07:15.900 three requirement to symptoms in the early developmental period, and it introduced certain

01:07:22.340 specifiers with or without intellectual or language impairment. And the severity levels were based on

01:07:29.960 needed support. Other changes that were also made to some of the checklist criteria, but the main issue

01:07:35.480 is that an array of disorders are now lumped together under this ASD umbrella, which has vastly increased

01:07:44.100 the number of individuals who fall under that umbrella. Now the estimates for how much this dramatically

01:07:52.260 increasing diagnostic aperture has contributed to the increase in prevalence vary. But the analyses that have

01:08:00.880 looked and attempted to assess this directly report that the expanded criteria account for 40 to 60% of the

01:08:10.020 increase. So a 2009 study found that roughly 26% of the increase in autism diagnoses in California between 92 and 2005

01:08:28.100 were attributable specifically to cases in which children had previously been diagnosed with mental retardation

01:08:35.220 and were then subsequently screened for autism. So again, when I hear people say, oh yes, but even the

01:08:41.560 cases of severe autism are increasing, not necessarily. It could be that kids that we now think have severe

01:08:48.180 autism, for example, being nonverbal, were actually previously diagnosed as something else. Racial and

01:08:54.740 socioeconomic disparities in autism diagnosis have narrowed or reversed over the last 30 years, which the CDC and others

01:09:01.440 suggest is evidence of more widespread awareness and screening. So take those two together, Nick, 40 to 60%,

01:09:09.040 20 to 30%. That's really kind of the lion's share of what explains this increase.

01:09:15.120 But that also still leaves room for other factors. And so do we know what else might be accounting for the

01:09:22.320 increase, not what we just covered?

01:09:24.860 Yeah, I think the next clearest contributor is advancing parental age. Both in mothers and fathers,

01:09:31.440 although the paternal age probably seems to play a greater role. This is seen mostly in the U.S. and other

01:09:37.840 high-income countries. And various studies have put this at about 5 to 15% of the increase in autism prevalence.

01:09:48.960 So paternal age has advanced in the U.S. from 27.6 years when I was born to 31.1 years 10 years ago.

01:10:00.480 That number seems to be going up. The proportion of fathers with more advanced paternal age has also

01:10:06.560 increased. So fathers over 40 at the time of offspring birth has more than doubled and going

01:10:14.240 from 4.1% to 8.9%. And fathers over 50 has also doubled going from 0.5% to 0.9%.

01:10:21.200 Trends in maternal age have also increased during the same average period by about three years. And the

01:10:28.880 CDC reports that between 2016 and 23, the proportion of births in women aged 35 and over has increased

01:10:36.960 from 10 to 12.5%. Furthermore, there are other factors such as maternal obesity, metabolic disease,

01:10:44.480 preterm birth, and air pollution that are also widely recognized to contribute to the remaining

01:10:50.960 15% of unattributable factors. So let's talk about these just briefly, right? Maternal

01:10:57.120 health, including metabolic health, is an important factor and there's no question that obesity rates

01:11:02.320 among women at the time of conception have risen steadily. A meta-analysis of global data reports

01:11:08.240 obesity rates in pregnancy have more than tripled in the last three decades from a pre-1990 rate of

01:11:15.840 4.7% to 16.3% in the decade from 2010 to 2020. And rates in the United States are even higher than

01:11:26.640 those averages. According to a 2024 CDC report, rates of preterm, so under 37 weeks, and early term,

01:11:35.840 37 to 38 week births rose during that period as well. So preterm births rose from 7.74% of all singletons

01:11:46.640 in 2014 to 8.67% in 2020, while early term birth rates rose from 24.31 to 29.07%. Sources that track

01:11:59.600 earlier years indicate a steady rise in preterm birth from at least 1980 to 2005 after which rates dip

01:12:08.000 slightly before beginning to rise again in the 2010s. Some of this of course could be attributed

01:12:14.000 to advanced maternal age, which is in and of itself a risk factor for preterm birth. Finally, I would say

01:12:20.240 globally air pollution has been increasing. We've talked about this a lot on the podcast. We talk a lot

01:12:25.280 about the PM 2.5s. Truthfully, we've always talked about it more through the risk of all-cause mortality

01:12:30.320 and cancer mortality, but here is yet another issue. So we've seen a 38% increase in PM 2.5s. Again,

01:12:36.720 just for folks maybe not familiar with that content, these are particles that are sub 2.5 microns in the

01:12:43.920 air. Obviously you can't see these things, you don't feel these things, but because of how small they are,

01:12:49.280 when inhaled, these particles can go all the way into the bloodstream because of their ability to go

01:12:55.360 straight down into the most distal part of the air sacs of the lung and cross the diffusion barrier

01:13:00.960 where oxygen and CO2 are transmitted. So seeing this enormous increase in pollution driven largely by

01:13:08.240 the industrialization of China and India is another part of this. And while air pollution in the US has been

01:13:16.000 coming down, we've seen in the last decade an uptick in this mostly attributed to wildfires.

01:13:22.640 Looking at what you just covered and those three buckets in that last bucket of environmental factors,

01:13:29.280 is it possible that acetaminophen could be in that bucket as well?

01:13:32.480 Yes, it is possible. Nothing I have discussed today, none of the analysis we've done or anybody has done,

01:13:40.240 has shown dispositively that we can disprove the role, the causal role between acetaminophen and

01:13:49.840 pregnancy and the elevated autism risk. Yes, it is possible. Again, as I've stated a couple of times,

01:13:55.840 it is impossible to disprove anything. We can't disprove anything here, that's the nature of what

01:14:02.800 we're doing epidemiologically. But the point here is, look at how many other variables we have that have

01:14:10.720 either demonstrated, i.e. genetically, or much, much stronger associations. That even if acetaminophen

01:14:19.840 plays some causal role, it is going to be very, very low. Think back to what we talked about on the

01:14:25.680 absolute risk increase. This was a 0.09 absolute risk increase with a 5% relative increase. So

01:14:34.960 even if you assume that to be causal, which again, I don't, because when the twin analysis was done,

01:14:42.320 all of that vanished in addition to everything else we've talked about, this would be a very, very,

01:14:47.680 very small contributor relative to other modifiable things such as maternal obesity,

01:14:55.120 metabolic health, air pollution, paternal and maternal age. So I think there are many things

01:15:01.040 we should be looking at before this. As we finish this podcast, I think one thing

01:15:06.480 you mentioned early on, which I think is really helpful and ultimately what a lot of people are

01:15:10.480 curious about is, based on everything we just talked about, what advice would you give to women

01:15:16.080 who are pregnant about the use of acetaminophen? As a general rule, I would advise women to stop

01:15:21.440 taking medications when they get pregnant, but medications aren't the only potential threat

01:15:26.720 to the unborn child. The health of the mother is also important to the unborn child. The medical

01:15:33.280 conditions that these medications are intended to treat can sometimes also create problems indirectly

01:15:39.760 or directly, but we have to balance that against the use of the medication and what's already being

01:15:46.080 addressed. Let's take an example. If a woman has an elevated ApoB, she should be taking a

01:15:51.440 lipid-lowering medication, but does she need to take that during pregnancy? I would argue no. Why?

01:15:56.880 Because nine more months of additional ApoB exposure are not a meaningful threat to a young woman's life,

01:16:04.320 whereas there may be some downside in suppressing her cholesterol synthesis if we're talking about a statin.

01:16:10.480 Conversely, when we think about something like thyroid hormone, where we've established actually

01:16:16.080 quite safe use during pregnancy, if a woman is requiring thyroid hormone because she has hypothyroid,

01:16:22.640 to withhold that from her during pregnancy would pose enormous risk to her and, by extension,

01:16:28.320 to the child. Now, it gets interesting when we talk about other classes of drugs. So, for example,

01:16:33.440 GLP-1 drugs. They're very common, and of course, the question is, should women stop these during

01:16:39.520 pregnancy? Well, I don't think I have enough data to comment, but I can tell you how one would have

01:16:44.160 to think about this. If a woman's taking a GLP-1 receptor agonist is the difference between her

01:16:50.800 having gestational diabetes and not, maybe it's considered. Of course, we would typically turn to

01:16:55.920 something like metformin as a first-line therapy there, where we have much more ongoing safety data.

01:17:01.680 But the point here is, you have to be able to consider this in a nuanced way, which is the single most

01:17:07.440 important thing for the healthy development of a fetus is a healthy environment in utero. And

01:17:14.480 sometimes that may actually require the mother taking a medication. With that kind of background,

01:17:20.960 might be worth going back to the historical FDA risk categories and just kind of walking through

01:17:26.880 what they are again. And then even highlighting a few different medications that are included in

01:17:32.160 in each category. So people just have a much better idea of how this is done in practice.

01:17:36.720 Yeah. Again, the good news is you don't have to guess here. You should be talking about this with

01:17:40.240 your doctor. So again, that FDA category, category A, which is pretty small. We've only got about two to

01:17:46.400 five percent of drugs here. We have controlled studies in humans that demonstrate no risk to the fetus in

01:17:52.400 any trimester. So again, the two most obvious here are T3, T4, prenatal vitamins, that kind of stuff.

01:17:57.760 Then you have category B. So animal studies that for the most part show no risk or animal risk,

01:18:04.160 not confirmed, adequate human epidemiology that generally shows safety. Again, 15 to 25 percent of

01:18:10.480 risk. We see a number of antibiotics in here, things like Benadryl. As I mentioned, Tylenol is in here,

01:18:16.320 as is Metformin. Then you go to category C. We have animal studies that show some adverse effects,

01:18:22.000 but no real adequate human studies. And here, these are drugs that are supposed to be used,

01:18:26.960 provided there's enough benefit for the mother to justify it. Again, this is most drugs fit in this

01:18:31.600 category, 60 to 70 percent. So you have something like gabapentin, amlodipine, which is a blood

01:18:36.720 pressure medication, trazodone for sleep, GLP-1 agonists are in here, certain SSRIs or antidepressants,

01:18:44.000 and even very short-term use of narcotic pain medication. Then you go to category D. So we have

01:18:51.280 positive human fetal risk data, but in some cases, the benefits might outweigh it. So for example,

01:18:57.600 a couple of seizure medications, valproic acid and phenytoin, also lithium, which would be used to

01:19:02.960 treat bipolar disorder, NSAIDs, which in the third trimester should be discontinued for the reasons I

01:19:08.080 talked about earlier, and even long-term use of narcotics. And then finally, we have category

01:19:14.160 X. We have drugs where there's simply no reason for women to take these during pregnancy. Statins

01:19:20.560 would be in this category, methotrexate, and drugs that also are known to cause teratogenic defects in

01:19:27.680 the child. One of the other things you talked about early on in the beginning was not only do you have

01:19:33.280 to look at the risk of taking the medication, but you also have to balance that in terms of what else

01:19:40.160 could be going on during pregnancy. And so how do you think about the use of acetaminophen in terms

01:19:46.800 of balancing the benefits that it can also cause for people who are pregnant? Yeah, I think we need

01:19:52.160 to look at the other side of the equation. What's the risk of not taking Tylenol during pregnancy?

01:19:56.880 In many cases, maybe the trade-off is just an annoying headache or some other discomfort that

01:20:02.640 the mother can sort of power through. And in those instances, maybe she's just better off skipping

01:20:07.200 the Tylenol and trying to get to bed. But we can't discount the mother's well-being and the

01:20:12.320 importance of that as well, not just for herself, but her well-being in the context of how important

01:20:17.280 it is for the unborn child. And we certainly shouldn't trivialize the likelihood and presence

01:20:22.800 of more intense debilitating pain with pregnancy. If the pain is bad enough that she's unable to get

01:20:28.000 out of bed for several days on end, that in and of itself poses a risk to the child.

01:20:33.680 Let's not forget, Tylenol is also used to reduce fever. And for this purpose,

01:20:38.400 current evidence would suggest that the scales clearly tip in favor of using Tylenol since

01:20:43.680 exposure to fever itself carries a number of known risk factors to a developing fetus.

01:20:49.360 So, for example, children born to mothers who experience fevers during pregnancy,

01:20:54.080 especially during the first trimester, are at a significantly higher risk of certain birth defects

01:20:59.440 than children who weren't exposed to fever in utero. Various analyses have reported anywhere from

01:21:04.720 25 to 200 percent higher risk for cleft palate or neural tube defects. In fact, prenatal exposure to

01:21:12.160 fever and maternal infection are also separate risk factors for autism and other neurodevelopmental

01:21:18.560 disorders. So several studies have reported that exposure to maternal infection is associated with an

01:21:24.000 increase in autism risk by 25 to 40 percent, while exposure to maternal fever is typically associated

01:21:31.120 with an even greater risk, up to 200 percent across most analyses. So, for some, research has reported

01:21:38.480 that these risks are attenuated when the mothers actually take a fever-reducing medicine like Tylenol.

01:21:45.120 All of these associations between infection and fever exposure and autism may actually be contributing to the

01:21:51.120 apparent correlation that we see between autism and autism risk. Given that acetaminophen is by far the

01:21:58.400 safest option for reducing fever and pain relief during pregnancy, because remember, NSAIDs and opioids are

01:22:04.240 category D. If a woman does have an infection during pregnancy, there's a good chance she might try to ease the

01:22:10.080 fever or the aches with Tylenol. And in other words, it could be that it's the infection that is the issue,

01:22:18.000 and the signal we're picking up and measuring is the acetaminophen use.

01:22:22.160 As we finish and wrap this podcast, is there anything else based on what we covered you want

01:22:27.360 to share with listeners and viewers? Look, there are some people who might be wondering,

01:22:31.840 why did you just take so long to explain all this to us? Why don't you just give us the answer? Like,

01:22:35.680 I just want the soundbite, man. And it's like, if you just want soundbites, you're never going to learn.

01:22:40.240 Honestly, if you just want soundbites, this isn't the podcast for you. But if you actually want to be able to

01:22:45.600 learn to think for yourself, then that's what we're here to do. And that's the reason we killed

01:22:49.600 ourselves over the past week to put together the most thorough gathering of all the data we could

01:22:56.880 find and the most intense night weekend analysis possible. It's because we want to help you think

01:23:04.880 about this stuff, because this is not going away. This is going to be a forever game of whack-a-mole.

01:23:10.960 There is always going to be a bad guy. I'm not going to be here every time to do a two-hour podcast

01:23:17.840 on helping you think through why exposure X leads to disease Y. I don't want to sound like a scolding

01:23:25.040 teacher, but the truth of the matter is we live in a world today where people don't want to think.

01:23:29.680 People use stupid vehicles like social media to get their information and they don't want to read the

01:23:36.240 fine print. And even if they do read the fine print, they just want to outsource thinking to

01:23:40.720 somebody else. So I appreciate that those of you who are still watching this have outsourced your

01:23:45.920 thinking for the past couple of hours to me. But honestly, we're never going to get out of this

01:23:50.960 rut of people not knowing how to think critically unless everybody starts taking steps to try practicing

01:23:57.040 this on their own. We're going to include amazing show notes to this podcast like we do for every

01:24:01.760 podcast. Although for this episode, it'll be not behind a paywall. Normally our show notes are only

01:24:06.880 there for our subscribers, but I would encourage you to go through this and follow the logic as I've

01:24:12.000 laid it out here with the help of my team. And when the next thing comes up, because it will come up,

01:24:17.280 whether it's this drug or that drug or this intervention, it's just going to keep happening

01:24:21.040 over and over again. You've got to be able to kind of go through this type of thinking.

01:24:24.880 If you don't want to, that's fine. It is hard, but then I think you've sort of forfeited your

01:24:29.280 right to have an opinion on it. So again, I don't mean to sound like a crotchety old man,

01:24:34.160 but honestly, I think on a day like today, I kind of feel like it. So let me just put a bow on this

01:24:38.960 and let's land this plane. I think the upshot here is that any one potential risk can't be considered

01:24:43.680 an isolation. You have to look at the full picture, the risk of a given intervention like Tylenol,

01:24:49.680 as well as the potential risks of not taking Tylenol, as well as the nature and magnitude of those risks.

01:24:56.560 For minor aches and pains, maybe it's best to just err on the side of caution and skip

01:25:01.280 the acetaminophen. Whereas when the pain becomes really a nuisance and it might interfere with you

01:25:06.400 doing things that are otherwise going to help you provide the best environment for your fetus,

01:25:12.480 then judicious use of acetaminophen can help with the oversight of your physician. I think for maternal

01:25:17.760 fever, the balance is clearly leaning towards the use of acetaminophen. But I want people to understand

01:25:24.000 that the strength of these associations is very small and in many cases vanishes altogether when

01:25:30.000 you apply some rigorous statistical corrections that look at the most important variables that

01:25:36.320 we should be considering here, which is genetic and environmental. So I hope that this exercise

01:25:41.360 has indeed provided benefit to all of you, not just as we consider this particular question,

01:25:46.480 but as we consider the onslaught of questions that we're going to see in the future.

01:25:50.880 Thank you for listening to this week's episode of the drive, head over to peteratiamd.com

01:25:57.040 forward slash show notes. If you want to dig deeper into this episode, you can also find me on YouTube,

01:26:03.520 Instagram, and Twitter, all with the handle peteratiamd. You can also leave us review on Apple

01:26:09.520 podcasts or whatever podcast player you use. This podcast is for general informational purposes only

01:26:16.080 and does not constitute the practice of medicine, nursing, or other professional healthcare services,

01:26:20.560 including the giving of medical advice. No doctor patient relationship is formed. The use

01:26:26.320 of this information and the materials linked to this podcast is at the user's own risk. The content

01:26:32.320 on this podcast is not intended to be a substitute for professional medical advice,

01:26:36.320 diagnosis, or treatment. Users should not disregard or delay in obtaining medical advice

01:26:41.360 from any medical condition they have, and they should seek the assistance of their healthcare

01:26:45.840 professionals for any such conditions. Finally, I take all conflicts of interest very seriously.

01:26:51.760 For all of my disclosures and the companies I invest in or advise, please visit peteratiamd.com

01:26:58.400 forward slash about where I keep an up-to-date and active list of all disclosures.

01:27:06.320 Thank you.

01:27:13.760 Thank you.

The Peter Attia Drive - October 06, 2025

#367 - Tylenol, pregnancy, and autism： What recent studies show and how to interpret the data

Episode Stats

Length

Words per Minute

Word Count

Sentence Count

Misogynist Sentences

Hate Speech Sentences

Summary

Transcript