Illustration of hybrid objects: part light bulb, part lab vial, some in blue and some in red to signify null and replicated results.
Illustration by Laurène Boglio

The true meaning of a null result

This edition of Null and Noteworthy highlights results that reveal the difficulty in drawing definitive conclusions from data, including new findings about epidurals that contradict several others and an apparent null result on sex differences that may derive from “circular logic.”

In May, I had the pleasure of attending a symposium at the Massachusetts Institute of Technology’s Picower Institute that featured wide-ranging talks on the early-life factors that may impact lifelong health. Among the speakers was Ran Rotem, a research associate at Harvard University, who discussed his approach to analyzing data from medical records. Though the findings he presented were published in 2020, they provided a new way for me to think about the value of null results.

In data from more than 400,000 births, Rotem and his colleagues identified an association between a woman’s underactive thyroid and autism in her child. But medications to correct hypothyroidism had no effect on the child’s chances of having autism, suggesting that the thyroid problem itself is not driving the association.

The presentation reminded me of the power of null findings, which are often thought of as simply a lack of a result rather than informative outcomes that can reshape thinking. It’s one of the reasons I started this newsletter. As always, please send your perception-shifting null results to [email protected].

Epidurals, resurrected:

Can epidurals contribute to autism? A 2020 JAMA Pediatrics study suggested they might, but a string of studies (many of them covered in this newsletter) looking at births in several countries said otherwise. Siblings are equally likely to be autistic whether their mother had an epidural or not, according to a massive 2022 study in 4.5 million people born Finland, Norway or Sweden, the largest study on the topic to date.

But even all this null evidence cannot close the door on this question entirely. In a July study in JAMA Network Open, the investigators on the 2020 work compared autism diagnoses among the children of people who had an epidural, oxytocin — a hormone that aids labor — or both. Children whose mothers had an epidural, either alone or with oxytocin, were more likely than those whose mothers had only oxytocin or neither intervention to be diagnosed with autism, the investigators found, even after controlling for factors such as the mother’s mental health during pregnancy.

The new work doesn’t negate the previous null findings, says Amir Sariaslan, a psychiatric epidemiologist at the University of Oxford in the United Kingdom, who led the 2022 work. The researchers did not compare children born to the same mother, which would have helped to eliminate confounding factors. And Sariaslan’s team found no association between epidurals and autism in their data. “If it is causally related, it should be seen in the population-wide estimates all across the world,” Sariaslan says. “That on its own implies that maybe something is wrong here.”

Sex differences:

Just as confounding factors can falsely give the appearance of a significant result, gaps in data can turn up an erroneous null. Case in point: A meta-analysis of 35 studies involving 8,794 boys and 3,202 girls aged 6 months to nearly 7 years detected no statistically significant differences between the sexes in core autism traits, social skills, cognition or co-occurring conditions. But because the studies only include children diagnosed with autism, the apparent null finding could represent what the investigators call “circular logic”: Girls diagnosed at a young age have traits that align with diagnostic tests based on boys’ typical presentations. “It remains possible that a subgroup of girls presenting with a distinct [autism] phenotype does exist but is missing from the pool of data,” the investigators write.

The findings were published in Research in Autism Spectrum Disorders in July.

Et al.:

  • Children exposed to anti-seizure medications in utero fare no differently on a measure of verbal skills at age 3 than children not exposed to such drugs. The investigators did not include valproate, which has been linked to autism and other neurodevelopmental issues. The Lancet Neurology
  • Taking folic acid supplements during pregnancy does not increase the likelihood of having a child with autism, according to a new review of seven studies. JAMA
  • The balance of nutrients consumed during pregnancy has little effect on a child’s scores on the Social Responsiveness Scale, a measure of autism traits. Current Developments in Nutrition