In July 2023 the Icelandic Medicines Agency passed on reports of suicidal ideation and self-injury in people taking liraglutide and semaglutide. Regulators in Europe and the United States opened reviews, and the question reached every buyer of these drugs long before the studies answering it were published.[1][2] What follows is what those studies found, stated at the scale the numbers actually support, alongside the better-quantified effects of this class.
Where the alarm came from
The original signal was made of spontaneous adverse-event reports. Anyone may file one, nothing in the process establishes that the drug caused the event, and reporting rates rise sharply once a drug is in the news. A signal of this kind is a reason to investigate, and it is not an estimate of risk.
A replication using the World Health Organization pharmacovigilance database shows exactly why. Reporting odds ratios for suicidal ideation were raised for semaglutide at 5.82, liraglutide at 4.03 and tirzepatide at 2.25. For suicide attempts and completed suicide, the same analysis found ratios below one for the same drugs, at 0.11 and 0.01 for semaglutide.[3]
Both directions cannot be causal. The authors state plainly that causation, increased or decreased, cannot be established from disproportionality data. That pattern is the strongest available warning against quoting a reporting ratio as if it were a rate.
What the population studies found
The most direct evidence on death comes from a new-user cohort built on nationwide registers in Sweden and Denmark between 2013 and 2021. Investigators compared 124,517 adults starting a GLP-1 receptor agonist with 174,036 starting an SGLT2 inhibitor, a comparator chosen because it treats similar patients.[4]
Over a mean 2.5 years there were 77 suicide deaths among GLP-1 users and 71 among the comparison group. Weighted incidences were 0.23 against 0.18 events per 1,000 person-years, a hazard ratio of 1.25 (95% CI, 0.83 to 1.88). The absolute difference was 0.05 events per 1,000 person-years, with a confidence interval running from a slight decrease to an increase of 0.16.[4]
The same study found a hazard ratio of 0.83 (95% CI, 0.70 to 0.97) for the composite of suicide death and nonfatal self-harm, and 1.01 (95% CI, 0.97 to 1.06) for incident depression and anxiety disorders.[4]A retrospective analysis of electronic health records covered 240,618 United States patients with overweight or obesity. Incident suicidal ideation was less likely on semaglutide than on other weight-loss drugs, at a hazard ratio of 0.27 (95% CI, 0.20 to 0.36), and the finding was replicated across 1,589,855 patients with type 2 diabetes.[2]That comparison ran over six months and set one drug against another rather than against no treatment at all.
What the randomized trials found
Pooled post hoc analysis of four phase 3 semaglutide obesity trials covered 3,377 participants in three of them and 304 in the fourth. Depressive symptoms were measured with the PHQ-9 and suicidal ideation with the Columbia-Suicide Severity Rating Scale.[5]
Mean PHQ-9 scores at baseline were 2.0 and 1.8, which is the no-symptom end of the scale. At week 68 they were 2.0 on semaglutide and 2.4 on placebo, an estimated treatment difference of 0.56 points in semaglutide's favor. Participants on the drug were less likely to shift into a more severe depression category, at an odds ratio of 0.63 (95% CI, 0.50 to 0.79).[5]
One percent or fewer of participants reported suicidal ideation or behavior during treatment, with no difference between semaglutide and placebo.[5] The authors describe the mood difference as statistically significant and not clinically meaningful, which is the correct reading of two-tenths of a point on a 27-point scale.
What remains genuinely uncertain
Those trials enrolled people without known major psychopathology. They therefore say very little about the population the original reports concerned, and nothing about someone with active depression or a history of a suicide attempt.[5] That is the largest gap in the evidence and it has not been closed.
Precision is the second problem. A Spanish cohort of 3,040 people starting a GLP-1 receptor agonist against 11,627 starting an SGLT2 inhibitor found no increase in suicidal ideation and self-injury, at a hazard ratio of 1.04. Its confidence interval ran from 0.35 to 3.14, and its authors wrote that an effect as high as threefold remained compatible with their data.[1]
Rare outcomes produce wide intervals, and wide intervals are not the same as evidence of safety. Read together, the fair summary is that the large studies do not support an increased risk of suicide death, self-harm or new depression, that each carries real imprecision, and that people with significant psychiatric history were largely studied by exclusion. The standards these readings are held to are set out in the methodology.
What this means for a reader
Nothing here tells anyone to begin or discontinue a medicine, and no website is in a position to. Mood, sleep and appetite change during rapid weight loss whatever causes it, and a prescriber who knows a reader's psychiatric history is the only person who can weigh that.
Worth raising at an appointment: any history of depression, self-harm or a suicide attempt, any psychiatric medicine already being taken, and any new low mood or hopelessness after starting treatment. Conditions that change the calculation entirely are listed in the contraindications article, and the mood effects of coming off treatment are not something the stopping evidence has measured.
If thoughts of suicide are present now, contact a clinician or a crisis line rather than a search engine. In the United States the 988 Suicide and Crisis Lifeline answers calls and texts at any hour. This site earns money from some of the sellers it writes about, which is disclosed in full on the disclosures page, and no safety page here carries a buy button.