One in ten babies born with diagnosed infections was later diagnosed autistic

Of 3.7 million people born in Sweden from 1987 to 2021, 975 had a diagnosed infection caught in the womb. About 1 in 10 of them was later diagnosed with autism, against about 1 in 32 of everyone else, and the gap held when they were compared with their own siblings. The study counted only infections serious enough to be diagnosed, which account for a tiny share of autism.

8 October 2026 · How we rate evidence

In a typical year, between 10 and 20 babies in Sweden are diagnosed with a cytomegalovirus infection caught in the womb, and fewer than 10 each with toxoplasma or herpes. Across 35 years of births, 3,666,002 people in all, researchers at Karolinska Institutet found 975 children diagnosed with one of these infections, about 1 in 3,760 births. Then they followed them for up to 36 years, into adulthood for the older ones.

The claim going around. The paper came out in JAMA Pediatrics on 21 September. Karolinska's press release said infected children were "approximately three times more likely to be diagnosed with autism," and that "approximately one in five children born with a TORCH infection may later develop autism." ScienceDaily reprinted the release under the headline "Rare pregnancy infections may triple the risk of autism," and indica News and Yahoo carried the same figures a week later.

What they did. The team, led by doctoral student Hugo Sjöqvist, linked Sweden's birth, patient, insurance, school and death registers. They looked for TORCH infections, the handful of germs that can cross the placenta and infect a fetus directly: toxoplasma (from undercooked meat or cat litter), syphilis and a few others, rubella, cytomegalovirus (CMV, common in young children's saliva and urine) and herpes simplex. CMV made up about half the cases.

They tracked seven outcomes, none marked as the main one: autism, intellectual disability, ADHD, OCD, tic disorders, psychosis and school grades at 16. They adjusted for family background and problems around birth. Then they compared infected children with their own uninfected brothers and sisters, which removes anything siblings share, such as parents, home and many of their genes. Few datasets could do this: an infection this rare takes a whole country's registers, siblings included.

What they found. By the end of follow-up, 102 of the 975 infected children had an autism diagnosis, about 105 per 1,000, against about 32 per 1,000 among everyone else. For intellectual disability the gap was wider: about 152 per 1,000 against 13.

The ratios are relative, after adjustment. Against siblings they were similar or larger, with wider ranges: 2 to 5 times for autism. For intellectual disability the ratio climbed with severity, from about 3 for mild to about 23 for severe to profound. No formal test backs that rise, though the ranges don't overlap; that puts it between Early and Promising. Of the extra autism, more came with intellectual disability than without it. The ADHD link vanished between siblings, and there was none for OCD. A design that can come up empty is more believable when it doesn't. Infected children with neither diagnosis finished school about 1.5 points lower on a scale averaging 100. That is a small shift, and it wasn't clear of chance between siblings.

The release's "up to 30 times" for severe intellectual disability matches the sibling estimate (about 33), though its range runs from 4 to over 200; the main estimate is 23. "One in five" comes from the supplement, which calculates how many infected children it takes for one to be diagnosed with autism. It doesn't show the method, and we couldn't reproduce the figure from the paper's own counts. Those give about one in ten diagnosed, or one extra case per 14 infected children above the background rate, with the average person aged about 19.

ScienceDaily's "may triple the risk" gets the size right for children born infected. But the study counted babies born with a diagnosed infection, not infections during pregnancy, most of which never reach the fetus, as the release itself notes.

What it doesn't show. First, it counts only diagnosed infections. Sweden doesn't screen newborns for these germs, so the 975 are mostly babies whose infection caused problems, and nearly a third were born early. The authors say plainly that silent infections couldn't be assessed. A French clinic that followed 642 children with CMV reported autism in 11 by age four, all of them babies who had symptoms at birth, according to the study's abstract.

Second, a child known to have a congenital infection is watched closely, with hearing tests, developmental checks and specialist visits. More looking can mean more diagnoses, especially of milder autism, and the sibling comparison doesn't fix that. The authors tested a related problem, infections found only after later symptoms, by counting only those diagnosed in the first year. Autism and intellectual disability held, but the sibling estimate for autism without intellectual disability was no longer clear of chance. The ADHD link fading between siblings cuts the other way, since closer follow-up should turn up more ADHD too.

Third, autism was one of seven outcomes, with no registered plan naming a main question.

What it means beyond the study. For people, the question is what this means for someone who is pregnant. The infections are rare: about 1 in 3,760 births here. If the links are causal, the authors estimate that eliminating these infections would remove 0.034% of autism diagnoses and about 1.2% of severe to profound intellectual disability. We couldn't reproduce the autism figure: standard formulas with the paper's own ratio give about 0.06%, some 70 of the cohort's 116,000 cases. Either way, a tiny share. For the families affected, though, the risk is large, and several of these infections can be prevented. The release lists rubella vaccination, care with raw meat and cat litter, hand hygiene around young children's saliva and urine, and syphilis screening and treatment in pregnancy. In Sweden, the authors write, rubella "virtually disappeared" after vaccination.

Why it's still interesting. Studies of common infections during pregnancy have found modest links with autism that largely faded when siblings were compared, which points to family factors rather than the infection. This study finds the opposite for the few germs that reach the fetus directly. The link survives the sibling comparison and, for intellectual disability, grows with severity. That helps sort a confusing literature into two piles, and it puts numbers on a rare, partly preventable link to intellectual disability and, more tentatively, to autism.

Open questions.

Is closer follow-up finding more autism? Compare infected children with children followed just as closely for other reasons, such as very early birth or inherited hearing loss. Or assess infected children and their siblings with the same standard autism tools. If the milder-autism link survives, it points to the infection's effect on the developing brain. If it fades, the autism link rests mainly on children who also have intellectual disability.

Do silent infections carry any risk? Places that test every newborn for CMV could follow babies who test positive but seem well alongside babies who test negative, with identical checks. The answer decides whether universal newborn screening, and the follow-up it brings, is worth the cost.

Does treatment change the outcome? The study couldn't assess whether treatment made a difference. Following children from treatment trials to school age would show whether early treatment makes autism or intellectual disability less likely, less severe, or neither. That matters for what doctors offer a newborn diagnosed today.

A note on timing. The paper and the Karolinska release came out on 21 September 2026. Coverage followed within days and was still arriving a week later.