One drug in the class showed something, and the design is better than most. A nationwide cohort of 14,694 people with bipolar disorder and diabetes or obesity compared each person against their own periods on and off these drugs [1]. Semaglutide came with a 21% lower risk of psychiatric hospitalization, adjusted HR 0.79, 95% CI 0.69 to 0.91. Bipolar relapse was 17% lower at 0.83, 95% CI 0.69 to 0.99, while liraglutide and dulaglutide showed no association at all. Sick leave for psychiatric reasons, the outcome closest to daily life, did not move for any of them. No randomized trial has tested one of these drugs as a psychiatric treatment, and the authors ask for exactly that.
People with bipolar disorder carry a heavy burden of obesity and diabetes, much of it produced by the medications that control the illness. A nationwide cohort spanning 2009 to 2024 asked whether treating the metabolic problem changes the psychiatric course[1]. The broader mortality picture in serious mental illness is in a four-year mortality gap.
The design is the reason to take it seriously. Rather than comparing treated people against untreated people, it compares each person against themselves across periods when they were and were not taking one of these drugs. Everything stable about someone — how severe their illness is, their genetics, their circumstances — is identical in both periods and cancels out. In a condition where severity drives both what gets prescribed and who gets admitted, that removes the dominant objection — the same instrument used to good effect in the alcohol discontinuation study.
Semaglutide was associated with a 21% lower risk of psychiatric hospitalization (adjusted HR 0.79, 95% CI 0.69–0.91) and a 17% lower risk of bipolar relapse (aHR 0.83, 95% CI 0.69–0.99). Liraglutide and dulaglutide showed no such association.
One outcome did not move at all, and it is the one closest to daily life. Sick leave for psychiatric reasons showed no association with any of the drugs studied. Hospitalization is a severe endpoint that also depends on admission thresholds and bed availability; sick leave tracks whether someone can work. A benefit that appears in one and not the other is narrower than the headline suggests.
The two neighboring psychiatric results
Two other findings sit next to this one and neither is about bipolar disorder. A target trial emulation matched 764,115 pairs starting a GLP-1 drug or an SGLT2 inhibitor, including 195,184 pairs with serious mental illness, and four-year mortality in that cohort was 4.91% against 6.45% [2]. That is a hazard ratio of 0.76 and an absolute difference of 1.54 percentage points, measured against another effective drug rather than against placebo. Separately, a 16-week double-blind trial randomized 72 adults with major depressive disorder to oral semaglutide 14 mg or placebo and found more willingness to exert effort as reward value rose, at P = .02 [3]. Its outcome was a laboratory task rather than a depression rating scale, which is set out in the effort and reward trial.
The authors ask for a randomized trial, which is the right conclusion and an achievable one here — unlike many populations this site covers, people with bipolar disorder and obesity are numerous and already in regular contact with services. Until then this sits alongside the other psychiatric findings on this site, the effort and reward trial and the clozapine cohort, as a signal worth testing rather than a reason to prescribe.