Use of Glucagon-Like Peptide 1 Receptor Agonists and the Associated Risk of Hospitalisation in Bipolar Disorder, From a Nationwide Cohort, 2009-2024.

This is a methodologically serious registry study: the within-individual design compares each person to themselves across time, which strips out stable confounders like genetics and baseline severity. The fact that the association appeared for semaglutide but not liraglutide or dulaglutide argues against a simple class effect and against pure healthy-user bias, though time-varying confounding can't be excluded. As the authors note, only a randomized trial can settle whether semaglutide itself is doing the work.