Before Blaming the Prescription

A Welsh study, reported by Eric W. Dolan in PsyPost on October 11, interrupts that shortcut. Researchers found no statistically significant differences in autism or ADHD risk when comparing children whose mothers had recorded depression with antidepressant prescriptions against those whose mothers had recorded depression without them. That is a narrower finding than “antidepressants carry no developmental risks.” It is also more useful than treating every association as an accusation.

Compared with whom?

Saraid McIlvride and colleagues linked health and education records for more than 167,000 children born in Wales between 2009 and 2016, following them through 2022. Their paper, published in PLOS Medicine on September 29, examined special educational needs and ADHD.

The categories depended on records: a maternal depression diagnosis in the two years before delivery, and at least one antidepressant prescription during pregnancy or the month before conception. “Treated” and “untreated” are therefore useful shorthand, not complete accounts of anyone's care. A prescription does not establish that medication was taken, and a missing diagnosis does not establish that depression was absent.

The researchers' adjusted models estimated special educational needs in 20 per 100 children with neither recorded exposure, 23.6 per 100 with maternal depression but no antidepressant prescription, and 26.5 per 100 with both. The difference between the latter two groups was 2.9 per 100 children, with a 95% confidence interval of 1.7 to 4.1.

That difference should not disappear behind a reassuring headline. Nor should it become proof of medication harm. The researchers lacked a measure of depression severity: mothers prescribed medication may have had more severe illness than those without prescriptions. The study could not fully separate those possibilities.

For autism and ADHD specifically, the treated-versus-untreated depression differences were not statistically significant. Small significant differences remained for learning difficulties and behavioral-emotional-social difficulties. The broad measure of special educational needs must not be substituted for any one diagnosis, and a non-significant result does not prove two groups identical.

The blame does not transfer

The methodological problem has a name: confounding by indication. People receive a treatment for a reason; that reason may itself be associated with the outcome being studied. Comparing those who receive medication with those who do not can mix the treatment with the circumstances that brought it into the room.

Here is the cultural temptation: if the prescription cannot be convicted, put the mother's depression on trial instead. Same courtroom, different defendant.

But this study establishes associations, not that a mother's depression caused her child's autism, ADHD or educational needs. Its findings cannot assign responsibility to an individual parent. Unmeasured differences between families remain possible explanations. Precision means resisting both the medication scare and the soothing overcorrection that declares the entire question settled.

Coherenceism's demand to include the people affected changes what we ask of a risk story. A pregnant person is not merely a possible exposure in a child's record; her health matters in its own right. A child who needs educational support is not evidence of a failed pregnancy. Both deserve more than an argument over who can be blamed.

This cohort cannot tell an individual whether starting, continuing or stopping a particular antidepressant would change their child's developmental outcome. Those decisions require individualized clinical discussion, not a headline acting as a prescription.

What the evidence can invite is a better conversation: which comparison produced this risk estimate, what remains unmeasured, and what support does this family need? Uncertainty is a reason to investigate carefully. It is not a vacancy that blame needs to fill.