Researchers measured selenium in 720 pregnant women in Nepal, then asked about their children's behaviour four years later. Out of 20 comparisons, one came up positive, and the authors did not oversell it

Selenium is one of those nutrients the body needs in tiny amounts and can't do without. It is built into a family of proteins that protect cells from oxidative damage, and it is involved in how the brain's chemical messengers work. That biology has made it a natural suspect in child development. A US study linked higher selenium in pregnancy to fewer behaviour problems in middle childhood. A Polish one found a link between selenium in cord blood and emotional problems. A Danish study tied a selenium-dependent enzyme to fewer ADHD symptoms in boys only, and another American one found more diagnosed ADHD with higher selenium, in girls only. The picture, in other words, was already untidy.

The new study, published in PLOS One on 9 October 2026, adds a piece from a very different setting. Researchers at the University of Bergen, hospitals in Norway and the Child Health Research Project in Kathmandu used a cohort from Bhaktapur, a densely populated district next to Nepal's capital. The women had originally been recruited, between 2017 and 2020, for a trial of vitamin B12 in pregnancy. All were 20 to 40 years old, healthy, and no more than 15 weeks pregnant when a blood sample was taken. The plasma was frozen, shipped to Tromsø in northern Norway on dry ice, and measured for selenium. Four years later, trained field workers sat down with each child's caregiver and went through the Strengths and Difficulties Questionnaire, 25 statements about a child's emotions, conduct, restlessness, friendships and kindness, each answered 'not true', 'somewhat true' or 'certainly true'. In all, 720 mother and child pairs had both measurements, 95% of the live births in the original trial.

Selenium was on the low side. The average was 74.7 micrograms per litre, and 35.8% of the women fell under 71.1, the cut-off the team borrowed from a study of pregnant Norwegians. For comparison, the US National Institutes of Health says 80 or more is usually considered enough in healthy people. If low selenium in pregnancy leaves a mark on children's behaviour, this is the kind of population where it ought to show.

It mostly didn't. Treating selenium as a simple number, there was no link with any of the five scales, before or after allowing for the mother's age, the family's living standard and the child's sex. Comparing women below the cut-off with those above it: nothing. Only when the women were split into thirds did one result cross the usual line. Children whose mothers were in the lowest third scored 0.33 points higher for hyperactivity and inattention than children whose mothers were in the middle third (95% confidence interval 0.02 to 0.64, p = 0.035), on a scale that here ran from 0 to 8.

Before reading much into that, look at what sits around it. By our count the paper's two main tables hold 20 adjusted comparisons, and one in 20 is what chance alone hands you at that threshold. Children of mothers in the highest third also scored a little higher than the middle third (0.15 points, not significant), which is not how a nutrient shortage usually behaves. And in the below-or-above comparison, the children of the women with normal selenium scored slightly higher for hyperactivity than those of the low group, the opposite direction, though again well within chance. The authors do not hide any of this. They call it 'limited evidence', note that 'no clear exposure-response pattern was observed', and write that 'this isolated finding requires confirmation'. In a field where a single significant number often becomes the title, that restraint is worth noticing.

The paper's second job was to test the questionnaire itself, which had an authorised Nepali translation whose behaviour in Nepal had never been checked. It needed some surgery. The standard five-scale structure would not fit until three of the 25 items were removed, including the one about attention span, which the authors think may have suffered in translation or may simply be hard for parents to judge in a four-year-old. Even then, the hyperactivity scale's internal consistency was 0.49 where 0.7 is the usual pass mark, and the peer problems scale managed 0.05, close to noise. So the one positive result sits on a scale that was not measuring very steadily. The same weakness cuts the other way as well: a blunt instrument can miss a real effect, and a single plasma sample reflects recent intake more than long-term stores.

Our reading: this is a careful study with an honest near-null, and it should cool the idea that selenium in pregnancy shapes a child's behaviour in any way large enough to see easily. It does not prove there is no effect, and it was never a test of supplements; nobody was given selenium. For anyone pregnant, the practical point is unchanged. The US recommended intake in pregnancy is 60 micrograms a day, and selenium is one of the nutrients where more is not safer: the upper limit is 400 micrograms a day, and an ounce of Brazil nuts can pass it. The analysis plan was registered before the analysis was run, the work was funded by the Research Council of Norway, and the authors declare no competing interests.