Headlines Say Antidepressants in Pregnancy Add No Autism or ADHD Risk. The Welsh Study Behind Them Found Higher Odds.

What is actually known is narrower than either side of this week's headlines. A population study of 167,447 children born in Wales between 2009 and 2016 found that depression during pregnancy is associated with more special educational need in school-age children. It also found that the medication used to treat that depression cannot be cleanly separated from the illness. PLOS Medicine published the work on September 29, 2026.

Over the weekend the study circulated under a tidier line. Time News ran it as "Antidepressants in Pregnancy Not Tied to Extra Autism or ADHD Risk, Study Finds." That sentence is accurate for one comparison in the paper. It is not what the paper reports for all of them.

What the Welsh records show

Researchers at the University of Glasgow linked primary care, maternity and school records for the whole cohort and followed the children until 2022, when their mean age was 5.68 years. Of the 167,447 children, 12,630 (7.6 percent) were exposed to depression that went untreated, 7,005 (4.2 percent) to depression treated with medication, and 5,996 (3.6 percent) to antidepressants with no depression diagnosis on record. Selective serotonin reuptake inhibitors made up 83.4 percent of the prescribing.

Across the cohort, 28.4 percent of children had a recorded special educational need. Autism accounted for 2.0 percent and ADHD for 1.8 percent. After adjustment, the predicted rate of any special educational need was 20.0 per 100 children with no exposure, 23.6 per 100 with untreated maternal depression, and 26.5 per 100 with treated depression.

Two statements from one paper

The authors write that marginal effects showed no significant difference in autism or ADHD risk between untreated and treated depression. That is the sentence the headline compresses. The same results section reports that, against children with no exposure, antidepressant medication carried an adjusted odds ratio of 1.59 for autism and an adjusted hazard ratio of 1.89 for ADHD. Depression alone carried 1.42 and 1.57.

Those higher figures are not a verdict on the drugs. The model included an interaction between depression and medication, and the authors state that the two effects are not independent. They also say they had no measure of how severe each mother's depression was, and that women who are prescribed antidepressants are likely to be the more severely ill. They had no data on alcohol or illicit drug use, and none on postnatal or paternal depression.

The evidence is thin on one point in particular. Tricyclic antidepressants and serotonin and norepinephrine reuptake inhibitors showed an 8.8 per 100 gap between treated and untreated depression (95 percent confidence intervals 3.2 to 14.4 and 2.6 to 15.0). Both classes are usually reserved for harder-to-treat illness, and both are also prescribed for migraine, fibromyalgia and nerve pain. The Welsh data cannot say whether that gap comes from the drug, the diagnosis, or the reason for the prescription.

A pooled analysis points the same direction, with one exception

The Welsh team's own discussion leans on a larger piece of work. The Lancet Psychiatry published a meta-analysis in May 2026 by Joe Kwun-Nam Chan and colleagues at the University of Hong Kong, covering 37 studies, 648,626 exposed pregnancies and 24,967,806 unexposed ones. The Lancet paper itself would not load for this reporter, so the figures here come from Contemporary OB/GYN's report on it, which is dated May 15, 2026.

Before adjustment, prenatal antidepressant use carried a relative risk of 1.35 for ADHD and 1.69 for autism. When the analysis accounted for maternal mental disorders, family influences and genetic predisposition, those associations were no longer statistically significant. Fathers who took antidepressants around conception or during the pregnancy showed a 46 percent higher ADHD risk and a 28 percent higher autism risk in their children. A father's medication cannot act on a fetus the way a mother's can, and the researchers read the pattern as pointing toward shared genes or home environment.

“While all medications carry risks, so too does stopping antidepressants during pregnancy”

The exception is two tricyclics. Even after minimizing confounding by indication, amitriptyline stayed associated with ADHD (relative risk 1.74, confidence interval 1.00 to 3.03), and amitriptyline combined with nortriptyline stayed associated with autism (relative risk 2.02, confidence interval 1.32 to 3.10). The ADHD interval touches 1.00. The Glasgow authors note that the only sibling analysis to report on tricyclics was underpowered, and that sibling comparisons generally erase the associations for other classes but can still carry residual bias.

Why the stakes run in both directions

Untreated depression is itself associated with higher need in this dataset: 23.6 per 100 children against a 20.0 baseline. Chang noted that stopping medication carries a relapse risk, and the Lancet Psychiatry authors concluded that the evidence supports continuing treatment in moderate-to-severe depression. That is their conclusion, drawn from pooled observational data. Neither study randomized anyone, so neither can establish cause in either direction.

Replication is partial. Many earlier studies found links between antidepressants and ADHD, and others found none, and the Welsh authors say sibling studies mostly find the links vanish. Nothing here is a reason to read one cohort as the last word. This article is reporting on research, not offering medical advice, and decisions about medication belong with a patient and a prescriber.