Whether GLP-1 drugs raise the risk of depression depends almost entirely on which other diabetes drug you line them up against. A nationwide study out of South Korea ran the same GLP-1 patients against two different comparison groups and got two different answers, and the gap between them is the actual finding.
The analysis, published in Diabetes, Obesity and Metabolism ↗, used a Korean national claims database covering 2012 to 2022. It followed adults with type 2 diabetes who newly started a GLP-1 receptor agonist (a GLP-1 drug, the class that includes semaglutide ↗, sold as Ozempic and Wegovy). It then compared them against two other newer diabetes drug classes: DPP-4 inhibitors like sitagliptin and SGLT2 inhibitors like dapagliflozin. The design is called an active-comparator new-user study. Everyone in it had just started some diabetes drug, so the groups are more alike than comparing drug-takers to the general population would be. GLP-1 starters were matched to 874,902 DPP-4 users in one arm and to 623,750 SGLT2 users in the other. The outcomes tracked were depression, anxiety, sleep disorders, and suicide.
Same drug, two verdicts
Against DPP-4 inhibitors, GLP-1 drugs looked clean. The risk of depression was identical (hazard ratio 1.00, 95 percent confidence interval 0.83 to 1.20), anxiety was if anything slightly lower (0.92, 0.77 to 1.08), and sleep disorders showed no clear difference (1.14, 0.97 to 1.33). Every one of those ranges crosses the no-difference line, which is the statistical way of saying the study could not tell the two drug groups apart on mental health.
Against SGLT2 inhibitors, the same GLP-1 patients looked slightly worse. Depression risk came out 20 percent higher (1.20, 1.09 to 1.31) and anxiety about 9 percent higher (1.09, 1.00 to 1.19), both just clearing the line into a real difference. Suicide was too rare to analyze, and no suicides occurred among the GLP-1 users at all.
So the same molecules were psychiatrically neutral against one comparator and mildly risky against the other. The authors' own conclusion is that psychiatric safety signals may vary by active comparator, which is a careful way of saying the comparison drug is doing half the work in every headline you have read about GLP-1 and mood.
Why the comparator is the story
A head-to-head study names a winner, not a cause. An elevated risk versus SGLT2 inhibitors can mean GLP-1 drugs nudge depression up, or it can mean SGLT2 inhibitors nudge it down, and this design cannot separate the two. SGLT2 inhibitors are not an inert stand-in. They have their own cardiovascular and metabolic effects, and if they carry any mood benefit, a neutral GLP-1 drug would look worse simply by standing next to them. Against DPP-4 inhibitors, a genuinely low-key comparator, the GLP-1 signal disappears.
This is why the GLP-1 mental-health literature keeps contradicting itself. A large post-marketing analysis tied semaglutide and liraglutide to more depression and anxiety, while a US Medicare emulation found no depression signal against SGLT2 inhibitors. peptidemodel has tracked both directions of this: semaglutide was earlier tied to fewer psychiatric hospitalizations in bipolar disorder ↗ than its drug siblings. None of these studies is wrong. They are answering different questions dressed as the same one.
For a patient or a prescriber, the honest read is modest. Starting a GLP-1 drug does not appear to move depression or anxiety much on its own, the small differences that do show up are sensitive to what you measure them against, and no claims database, however large, settles a causal question that only a randomized trial with mental-health endpoints can close.