Image of an epidural being administered.
Scientific reflection: A seemingly simple epidemiology paper prompted researchers to consider what publishing is for.
ChaNaWiT / Getty Images

A questionable study linked epidurals to autism. Then what?

Researchers and clinicians were quick to point out the flaws in the study, and a flood of work refuted it.

Listen to this story:

The spread of misinformation — and the way fear can burn through a landscape — was forefront in Tor-Arne Hegvik’s mind when, in October 2020, he opened up his web browser and found a new autism paper in JAMA Pediatrics. Hegvik is a gynecologist and researcher at the University of Bergen in Norway, and the study, co-led by biostatistician Anny Xiang of Kaiser Permanente Southern California (KPSC), made what appeared to be an improbable claim: The epidural analgesia given during labor is associated with increased chances of having a child with autism.

Hegvik had never wondered whether epidurals could contribute to autism, and he didn’t think anyone else had seriously considered it either. He assumed the association detailed in the paper was due to confounders — unaccounted-for factors that skew a result. Plus, he had clinical duties to attend to and patients to see. He tried to put it out of his mind.

But the paper continued to plague him. It reminded him of Andrew Wakefield’s infamous fraudulent statement that vaccines cause autism: Though the link was easy to disprove, it was exceedingly difficult to quash in the public consciousness. The epidural paper was not exactly that — Hegvik could see it was honest science — but the potential for harm felt similar. The decision to get an epidural requires clear, straightforward information. The Xiang paper, he worried, would muddy the waters. He couldn’t turn away. So he emailed some of his colleagues and asked: Would anyone be interested in trying to refute this?

Other worried researchers were also talking about the paper, united by the thought that the study’s implied conclusion — that epidurals cause autism — could become a Wakefield-esque fiasco. It is difficult to know exactly what fears will catch fire in the public consciousness, but if it was even a possibility here, they wanted to do something about it.


he first person at JAMA Pediatrics to read the epidural paper was the journal’s chief editor, Dimitri Christakis. Christakis rejects on sight about three-quarters of the papers he receives. The rest go to his team of associate editors, some then go on to peer review, and in the end just 6 percent of submissions are accepted. So it’s safe to say JAMA Pediatrics is a discerning publication. Today, Christakis doesn’t remember anything particularly noteworthy about Xiang’s paper or its peer-review process. (The journal does not make its reviews public.)

What he saw was a study showing that among nearly 148,000 babies born in a KPSC hospital between 2008 and 2015, children whose mothers had an epidural while giving birth were 37 percent more likely to be autistic than those whose mothers did not have one; the longer the epidural exposure, the higher the chance of autism. Christakis felt there were some weaknesses in the data, but the researchers acknowledged them and were cautious in their conclusions. The methodology was not flawed on its face. He felt the question Xiang had introduced deserved an answer.

Christakis’ decision to publish revolved largely around the role of scientific publishing. Part of the job of a journal, he told me, is to present ideas for its community to evaluate. He could find no reason not to present Xiang’s. JAMA Pediatrics accepted the paper on 17 June 2020 and published it on 12 October.

If the Xiang paper presented an intriguing scientific idea, it did not appear to be one the research community had previously cared much about. According to PubMed, before 2020 just one study had ever specifically looked at epidurals and autism together — a 2004 Australian paper that found a small association in about 2,000 people.

In their paper, Xiang and her colleagues were careful not to state that epidurals might cause autism. They had simply observed an association and noted that it should be viewed “with caution” and “cannot be interpreted as a demonstration of a causal link.”

But Hegvik thought the cautioning statements did not go far enough.

“When you read the article, they never directly claim that epidurals might cause autism,” Hegvik said. “But it’s like, everywhere you read between the lines, that’s basically what they’re writing.”

Daniel McIsaac, an anesthesiologist at Ottawa Hospital and associate professor of anesthesiology at the University of Ottawa in Canada, felt much the same after reading the Xiang paper. “I thought, when it first came out, ‘Oh my goodness, this is just going to permeate every discussion I have for the rest of my career,’” he said.

McIsaac, like many others, was concerned that the study failed to account for factors that could make a person more likely to both use an epidural and give birth to an autistic child. Xiang and her colleagues, for example, did not control for the mother’s own psychiatric conditions, even though such conditions in parents are widely known to be linked to autism in children. Xiang and her team also excluded women who had received cesarean sections, which often occur because of complications with labor.

What’s more, it’s hard to see how epidurals — which come at the end of a pregnancy and generally last just a few hours — could lead to autism, a condition thought to originate in the earliest stages of fetal development. Because of this, McIsaac couldn’t understand why the research was even done. Typically, to justify doing a major epidemiological study like this one, a researcher needs good reason to think there’s something to find. “Just because you have data,” McIsaac said, “is never the right reason to do a study.”


eyond citing some preliminary research in people, Xiang and her team also looked at a 1998 study comparing 11 infant rhesus monkeys born to mothers that had received a dose of bupivacaine (one of the drugs in a standard epidural) with eight control rhesus infants. The study showed that “standard clinical doses” of anesthetic drugs can “alter normal behavioral development,” according to Xiang’s paper. But that 1998 study concluded that even though the time course of the exposed rhesus infants’ development was shifted, the drug “does not cause neonatal abnormalities or specific cognitive deficits,” the researchers wrote.

Alexander Butwick, professor of anesthesiology at Stanford University in California, found the rhesus monkey data referenced in Xiang’s paper to be “pretty weak,” he told me. He and a colleague commented on the JAMA Pediatrics website that the dose used in the animal study wasn’t comparable to the one used in the delivery ward and called the researchers’ interpretations of relevant studies in people “at best misleading and at worst erroneous.”

Butwick set out to repeat the Xiang study with fresh data, and to do it quickly. “We really felt like we had a responsibility here to do not just the work but also respond in a timely manner,” Butwick said. “We were worried that the longer this was out there, the more people may feel like there’s a genuine cause-and-effect issue going on.”

He called epidemiologist Elizabeth Wall-Wieler, then assistant professor of community health sciences at the University of Manitoba in Canada (she now works in California), who had access to a dataset that linked clinical data from hospital visits to parent socioeconomic information and long-term outcomes for the child. Butwick and Wall-Wieler pulled numbers on 123,175 babies born in Manitoba from 2005 through 2016, adjusted for more than 30 factors — including maternal anxiety and depression — and found that epidurals were not associated with autism. That finding stood in a secondary analysis of siblings born to the same mother.

The paper was published 19 April 2021, also in JAMA Pediatrics. It was the first time the journal had published conflicting results on the same topic within 12 months — such a rare occurrence that Christakis pointed it out in an editor’s note published the same day.


till, the papers around epidurals and autism kept coming. JAMA published two on the same day, in late September 2021. One, which looked at nearly 400,000 births in British Columbia, Canada, found a small association, though the authors ascribed it to “residual confounding.” The other, in nearly half a million children from a Danish registry, found no link at all. In a commentary accompanying the two papers, Cynthia Wong, chair of the Department of Anesthesia at the University of Iowa, wrote that “current evidence does not justify” considering the chances of autism when making decisions about epidurals.

In December 2021, the British Journal of Anesthesia published yet another null result, this time in almost 625,000 children born in Denmark. It was closely followed by a commentary from Wong and Butwick in the same journal, asserting “no meaningful association” between epidurals and autism. The case, they argued, was closed.

But four months later, JAMA Network Open published a study involving more than 650,000 mother-child pairs in Ontario, Canada, that found a small link between epidurals and autism — though the research team, which included McIsaac, noted the finding “must be interpreted with caution.” There were two meta-analyses, both of which found little reason to continue investigating the question. In fact, in the two and a half years since Xiang’s paper came out, there have been at least 28 papers addressing the purported connection. If the research community had not been interested in a link between autism and epidurals before Xiang’s study, it certainly was now. (Xiang declined to be interviewed by phone for this story, but in an email sent via a publicist affiliated with Kaiser Permanente, she repeated that her paper had only noted an association and added that she and her team welcomed additional studies. Xiang also stressed the benefits of epidurals for mothers and noted that there is not enough evidence to recommend against them.)



There was one more paper coming. Amir Sariaslan, a psychiatric epidemiologist at the University of Oxford in the U.K., was on the receiving end of Hegvik’s October 2020 email about Xiang’s study. He had never studied epidurals and had limited experience studying autism, but he had worked with Hegvik before and, crucially, could tap national registry data from Finland and Sweden, which contained information on more than 3.5 million births. Hegvik had access to data from another million births in Norway, and together they pulled together a game-changing cohort: 4.5 million people born across 28 years, with more than 1 million of them delivered using an epidural.

It took two years to finish the study; the findings were published online in the American Journal of Obstetrics and Gynecology last August. In a crude analysis, Sariaslan and his team found that children exposed to epidurals were 12 percent more likely to be autistic and 20 percent more likely to have attention-deficit/hyperactivity disorder than their unexposed counterparts.

Other studies had made initial steps beyond these preliminary associations. But because of the size of the cohort, Sariaslan had the power to do a new kind of analysis. He and his team pared their sample down to include only full siblings — other studies had confirmed just a shared mother. They then pulled just those siblings who differed both in their exposure to epidurals and whether or not they had autism or ADHD.

That specific group was about 2 percent of the original 4.5 million. But it was still 94,000 people. That was sufficient to find that people with either condition were no more likely than their siblings without a neurodevelopmental condition to have been exposed to an epidural.

In other words, the association between autism and epidurals completely disappeared.

“I personally think we’re done with this,” Wong said after Sariaslan’s study came out. There would be nothing to gain from mining more databases or a larger dataset. “There’s no place to go.”


he point of scientific publishing is to go “wherever science leads,” Christakis told me. Much of the pushback after the Xiang paper was published, he said, was by people worried that “this is going to panic women” about epidurals. “But that’s not a reason to not publish science. It’s just not.”

Science works only when researchers are free to publish their findings, regardless of what societal structures they may upend or whom they might upset. That march toward truth relies on the free flow of information. But the human brain is not always so rational. When encountering a new concept, people tend to focus on and remember the parts that cause negative emotions, like fear and anger. When that happens, collective efforts to debunk new ideas can sometimes just reinforce them.

That could have happened with Xiang’s work. The paper alarmed not only scientists but also doctors, and on the initial day of publication, five U.S. medical societies representing more than 100,000 physicians posted a response saying the study “does not provide credible scientific evidence” for the association and urging women not to be fearful of epidurals. By the end of that month, the Royal College of Anaesthetists, the Royal Australian and New Zealand College of Obstetricians and Gynaecologists and the Canadian Anesthesiologists’ Society had also all pushed back against the paper’s findings.

Their intention was much the same as Hegvik’s: to quickly refute what they saw as a potentially harmful conclusion in a paper based on science that could be challenged. But Dominique Brossard, an expert in the communication of science and risk at the University of Wisconsin-Madison, told me that rushing out a press release was, in her mind, exactly what not to do. The collective statement from physicians on the day of the publication, and those in the days that followed, could have stirred up a long-lasting media buzz — the kind that leaves the public recalling a shadowy link between epidurals and autism, and not much else.

Yet it did not. As far as anyone could tell me, the sweeping fear over epidurals hasn’t materialized in the general population. No one could cite more than a handful of instances from the past three years of a pregnant person asking about the link between autism and epidurals; most said they had never been asked about it. Searches for the topic turned up little on social media. It’s possible this is because the press was otherwise occupied in October 2020: It was furiously covering the global pandemic, and the United States was a month away from a contentious presidential election. Major national outlets such as The New York Times, The Washington Post and The Los Angeles Times, as well as popular science news sites such as Scientific American and New Scientist, did not cover Xiang’s study. (Spectrum ran a story the day after the study came out.)

That’s one theory. But Brossard has another. And that is that there was no organized movement eagerly waiting to pounce on new data about epidurals, no group that could use it to further a specific viewpoint. There is no anti-epidural lobbying group, for instance, and there is no epidural-alternative product to sell. If there had been people seeking to amplify Xiang’s study for personal gain, maybe it would have caught fire in the public consciousness.

Instead, though, the issue fizzled. That just left the scientists, gathering their evidence, publishing the story it told.