why fda removed the glp-1 suicide warning: what it reviewed, what is left on the label, and the studies that disagree

from the day they were approved, the weight-loss versions of these drugs carried a warning about suicidal thoughts, and in february 2026 it was gone. if you take one, or were told not to because of your history, the reasonable question is what changed: the drug, the evidence, or the politics.

the answer is in fda's own record: what it reviewed, what the european regulator had already concluded, what is left on the three labels today, and the studies that point the other way. generic names first; brand names are in brackets because that is what is on the box.

why the warning was there in the first place

not because of a finding about glp-1 drugs. fda says the weight-reduction labels carried the warning from approval because “similar information about si/b is also included in the labeling of other types of weight loss medicines and is based on reports of such events observed with a variety of older medicines used or studied for weight loss”^[1]. the versions of the same drug class approved for type 2 diabetes did not carry it^[1]. so the question fda had to answer was not whether to add a warning but whether a warning inherited from other drugs belonged on these ones.

the timeline

• july 2023. after post-marketing reports of suicidal thoughts, fda opens an investigation^[1]; the european medicines agency starts its own review the same month after case reports with liraglutide and semaglutide^[3].
• january 11, 2024. fda's preliminary evaluation “has not found evidence that use of these medicines causes suicidal thoughts or actions,” but because events were few, fda “cannot definitively rule out that a small risk may exist” and starts a meta-analysis and a sentinel study^[2].
• april 12, 2024. the ema's safety committee concludes the evidence “does not support a causal association” and that “no update to the product information is warranted”^[3].
• january 13, 2026. fda asks application holders to remove the warning^[1].
• february 2026. the three labels list the section as removed^[4],[5],[6].

what fda reviewed

the trials. fda pooled 91 placebo-controlled trials across every glp-1 development programme: 107,910 patients, 60,338 on a glp-1 drug and 47,572 on placebo. the results “did not show an increased risk for si/b or for other relevant psychiatric adverse events such as anxiety, depression, irritability, or psychosis”^[1]. this was the step fda said in 2024 it needed, because individual trials had too few events to estimate a risk precisely^[1],[2].

the real-world study. using sentinel, its claims-data network, fda compared new users of a glp-1 drug with new users of an sglt2 inhibitor, another diabetes drug class, among people with type 2 diabetes: 2,243,138 people (1,161,983 and 1,081,155) from 10 data partners, october 2015 to september 2023. after adjusting for baseline differences it found no increased risk of intentional self-harm, and none in the subgroup with both diabetes and obesity^[1].

the published literature. fda also reviewed published observational and pooled studies and concluded that “the totality of these studies does not support a causal relationship”^[1].

two limits fda's own description makes visible. the sentinel cohort was people with diabetes, not the people taking these drugs for weight loss, whose labels carried the warning; the subgroup with obesity narrows that gap without closing it. and neither the meta-analysis nor the sentinel study has been published in full alongside the communication, so outsiders cannot yet check the analysis^[1].

what is left on the labels

we read the current versions on dailymed on october 23, 2026. the semaglutide (wegovy) label, effective june 2026, lists “suicidal behavior and ideation (5.10) … (removed) 02/2026” and contains no other depression warning; its adolescent table reports anxiety in 4% on the drug against 2% on placebo^[4]. the liraglutide (saxenda) label lists the section removed in 02/2026; its adult trial table reports anxiety at 2.0% against 1.6% and insomnia at 2.4% against 1.7%, and its 56-week pediatric trial table reports depression in 4% of 125 children on the drug against 2.4% of 126 on placebo^[5]. the tirzepatide (zepbound) label lists the section removed in 02/2026 and contains no depression warning^[6]. on all three, “mood changes” survive only as a symptom of low blood sugar^[4],[5],[6].

what is not on the label is still in fda's communication: patients should “tell your health care professional if you experience new or worsening depression, suicidal thoughts, or any unusual changes in mood or behavior,” and clinicians should refer anyone who discloses suicidal thoughts to mental health professionals^[1]. if that is you right now, call or text 988.

the studies that found no increase

the strongest designs compare people starting a glp-1 drug with people starting a different drug for the same condition, which removes much of the difference between people who are and are not treated. a swedish and danish register study of 298,553 adults found suicide deaths were rare (77 on glp-1 drugs, 71 on sglt2 inhibitors), with a hazard ratio of 1.25 and a confidence interval of 0.83 to 1.88, so not a significant increase; the combined outcome of suicide and non-fatal self-harm was lower (0.83), and new depression or anxiety diagnoses were unchanged (1.01)^[8]. a uk primary-care study found something instructive: unadjusted, glp-1 users had about double the rate of suicidality against dpp-4 inhibitor users (hazard ratio 2.08), and after accounting for who gets prescribed what, the figure fell to 1.02; against sglt2 inhibitors, 1.60 fell to 0.91^[9]. a us electronic-records study of 240,618 people with overweight or obesity found semaglutide linked to a lower risk of new suicidal ideation than other weight-loss drugs over six months (hazard ratio 0.27)^[7].

the studies that found a signal

they exist, and they are the weaker designs, which is not the same as wrong. a 2024 analysis of the world health organization's database of suspected adverse reactions found semaglutide was reported with suicidal ideation more often than expected against all other drugs (reporting odds ratio 1.45, 95% ci 1.18 to 1.77), and more so when compared with dapagliflozin, metformin or orlistat; the authors said the signal “warrants urgent clarification”^[10]. spontaneous reports cannot give a rate or show cause, and both regulators opened reviews in july 2023, inside the reporting window. a 2024 cohort of people with obesity compared glp-1 users with people taking no such drug and reported a 106% higher risk of suicidal behavior and 195% higher risk of major depression^[11]. comparing users with non-users, rather than with users of another drug, leaves room for the difference to come from who seeks treatment and gets diagnosed; the uk study above shows how large that effect can be^[9],[11].

what to ask your prescriber

these are questions, not advice. they are the ones the record above makes it reasonable to ask.

• i was told earlier that my history of depression mattered for this drug. does the 2026 change alter that advice for me?
• what mood changes would you want to hear about, and how soon?
• if i lose weight quickly or eat much less, could that affect how i feel, separately from the drug itself?
• do any of my other medicines interact with this one in a way that affects mood or blood sugar?
• if i have a hard week, who do i call: you, my therapist, or 988?

adam