The Lancet Psychiatry on March 18 published a national Swedish cohort of 95,490 adults followed for up to four years on semaglutide. [1] The paper's adjusted hazard ratios are now in front of regulators on both sides of the Atlantic: 0.58 for any new mental illness, 0.56 for depression, 0.62 for anxiety, 0.53 for substance-use disorders, and 0.56 for self-harm. The pattern held when the authors restricted the analysis to patients with no prior psychiatric history and when they adjusted for baseline weight, glycemic control, and antidepressant use.
The paper's Friday brief on the cohort's arrival at the FDA argued the regulatory window had narrowed — the prescription denominator crossed two million as the safety signal sharpened. What the brief left for today is the second study.
A ScienceDirect paper from the same window compared semaglutide against SGLT2 inhibitors and DPP4 inhibitors in a smaller U.S. cohort and reported the opposite signal: GLP-1 patients had higher, not lower, rates of incident depression and anxiety after one year of follow-up. [2] The methodological cut is the difference. The Swedish paper compared semaglutide users against the Swedish general population. The U.S. paper compared semaglutide users against patients on other diabetes medications — a tighter comparator that controls for the baseline metabolic disease but loses the population-level reassurance the Swedish design provides.
That is the regulatory question. The FDA has both designs on its desk inside the same review cycle. Each is large, peer-reviewed, and methodologically defensible. Neither answers the other. A label expansion for semaglutide into psychiatric indications turns on which design the agency treats as primary. Karolinska Institutet's lead author, Pouya Tahsili-Fahadan, told the Karolinska press office the Swedish data "suggest GLP-1 receptor agonists may have neuroprotective effects beyond glycemic control," and that follow-up sub-population analyses are in progress. [3]
The mechanism debate has not waited. GLP-1 receptors are densely expressed in the hippocampus, the prefrontal cortex, and the brainstem reward circuitry. The neuroprotective hypothesis is that semaglutide reduces neuroinflammation and improves insulin signaling in regions implicated in mood disorders. The competing hypothesis — the one the comparator-class study supports — is that any apparent mental-health benefit on a population scale washes out when you control for the metabolic benefits the comparator drugs deliver too. Both hypotheses have biology behind them. Neither has resolved.
What is resolved is the prescribing reality. U.S. semaglutide volume crossed two million weekly scripts in mid-April. [1] The pill formulation that Novo Nordisk launched in late January cleared a million prescriptions inside its first sixteen weeks — the fastest U.S. pharmaceutical launch outside vaccines, by Novo CEO Lars Doustdar's metric. The label question is no longer an abstract regulatory exercise. Whatever the FDA does with the mental-health data lands on a denominator the agency itself has not seen since the SSRI rollouts of the late 1990s.
Two further timing pressures matter. Novo Nordisk has signaled it will file a label-expansion supplement for cardiovascular indications by the third quarter, with a depression sub-claim under internal review. The Karolinska group has a follow-on paper in peer review on whether the protective signal holds in patients with a documented prior psychiatric admission — the population the FDA most needs to understand. And the comparator-class authors are reportedly preparing a response to the methodological critiques that have arrived in the weeks since publication. None of those drops has a date yet.
For the clinician reading both papers tonight, the practical answer is unsatisfying. The Swedish cohort suggests semaglutide does not worsen mental health and may improve it. The U.S. comparator study suggests, against the right reference class, the picture is muddier. Patients on the drug are not asking which design the FDA prefers. They are asking whether a refill in May is the same drug they started in January. The answer, for now, is yes — and that is the only part of the picture the regulators agree on.
-- NORA WHITFIELD, Chicago