In mid-2023, case reports of suicidal thoughts in GLP-1 users triggered regulatory reviews on two continents and a wave of alarming headlines. What followed was one of the more instructive safety investigations of the decade — because the large-scale evidence, when it arrived, pointed somewhere the headlines never went. This file walks the whole arc: the scare, the reviews, the cohort data, the legitimate residual cautions, and the practical questions for anyone starting therapy with a psychiatric history. One note before the data: this article discusses suicidality as a research topic; if any of it is personally close to home, it deserves a human conversation — your clinician, or in the U.S. the 988 Suicide and Crisis Lifeline — more than it deserves another paragraph of epidemiology.
Where the concern came from
Two tributaries. The older one: weight-loss drugs as a class carry psychiatric scar tissue — rimonabant was withdrawn in Europe in 2008 over depression and suicidality, and that history hardwired mood surveillance into how regulators approach the category. The newer one: in 2023, Iceland’s regulator flagged case reports of suicidal ideation and self-harm in semaglutide and liraglutide users, prompting the European Medicines Agency to open a formal review, with the FDA following via its own adverse-event and surveillance analyses. Case reports are how signals start; they are also, notoriously, how false alarms start — people taking any medication also live lives in which psychiatric events occur.
What the regulators concluded
In January 2024 the FDA published a preliminary evaluation stating that its review of reports and trial data did not find evidence of a causal link between GLP-1 receptor agonists and suicidal thoughts or actions, while noting it could not definitively rule out a small risk and would continue looking, including through analyses of large insurance-claims databases in its Sentinel system. In April 2024, the EMA’s safety committee closed its review similarly: available evidence did not support a causal association, no label change required. Two independent regulators, same destination. Neither declared the question eternally closed — regulators never do — but “no evidence of causation on current data” is the operative finding, and it has held through subsequent updates as of our last verification.
What the big observational studies found
The cohort evidence went a step further than “no signal.” A widely cited analysis of electronic health records (Wang and colleagues, Nature Medicine, 2024) compared people prescribed semaglutide for obesity against those prescribed other anti-obesity medications and found lower rates of first-time and recurrent suicidal ideation in the semaglutide group — a protective-direction association, replicated in the diabetes comparison. Other pharmacovigilance and cohort analyses have landed variously at null or protective, with the occasional disproportionality analysis on the other side illustrating mostly the limits of spontaneous-report databases. The sober reading: randomized trials (which screened out severe psychiatric illness) show no excess; large real-world cohorts trend null-to-favorable; and a hypothesis even exists for benefit, since the same reward-circuit effects being studied for alcohol craving plausibly touch mood-adjacent machinery. Association is not proof in the favorable direction either — but the asymmetry with the 2023 headlines is stark.
Why the labels still say “monitor”
Wegovy’s and other weight-management labels advise monitoring for depression and suicidal thoughts, and that guidance predates and survives the 2023–24 reviews. Three reasons it stays. Category history: the rimonabant precedent makes mood monitoring the class default for weight-management indications regardless of molecule. Trial exclusions: people with recent severe depression or suicidality were kept out of the pivotal studies, so the drugs are least-tested in exactly the highest-risk group. And plain prudence: rapid physiological and life change — which successful weight loss is — interacts with mental health in both directions, drug or no drug. “Monitor” is not a coded confession of causation; it is the reasonable posture toward an incompletely characterized population.
The lived-experience middle ground the studies undercount
Between “no suicidality signal” and “all clear” sits a band of real, commonly reported experience: mood dips during rough titration weeks (feeling unwell does that), the psychological weirdness of losing “food noise” — for some people, eating was load-bearing coping, and its sudden absence leaves a gap (the food-noise file covers the mechanism); occasional reports of emotional blunting; and, on the positive side, meaningful mood improvement tracking health gains. None of this is well captured by suicidality endpoints, most of it is manageable when named in advance, and all of it argues for treating mental health as part of the treatment conversation rather than an FAQ footnote.
Starting therapy with a psychiatric history: the practical version
A history of depression or anxiety is not a listed contraindication, and psychiatric medications are not, as a class, incompatible — but three practicalities matter. Disclose the history and current medications at intake, and treat an intake that never asks as the red flag it is — psychiatric screening quality is one of our platform criteria. Loop in whoever manages your mental health, both because coordination beats surprise and because appetite and absorption changes can matter for some regimens — prescriber territory, not forum territory. And pre-agree a monitoring plan that names concrete triggers: mood changes persisting beyond a rough titration week, new hopelessness, any self-harm thought — each of which means contacting the clinician promptly, and the last of which means immediate help (988 in the U.S.), full stop. Recent severe depression, a past suicide attempt, or active instability moves the whole decision into specialist-involved territory — the population the trials excluded deserves more than a checkout flow’s judgment.
How to read the next scary headline
This topic will generate future news cycles; here is the durable rubric. Check the study type: spontaneous-report disproportionality analyses generate hypotheses, not conclusions; large active-comparator cohorts and randomized data outrank them. Check the comparator: “versus the general population” confounds; “versus similar patients on other treatment” informs. Check whether regulators moved: label changes and formal review outcomes are the scoreboard, and as of our last verification the scoreboard reads no established causal link. And check the base-rate framing — millions of users guarantee thousands of coincident psychiatric events annually with or without causation, which is precisely why the question requires the big studies rather than the case reports.
The bottom line
The 2023 scare was investigated the way drug-safety questions should be, and the answer, so far, is reassuring: no causal link established by FDA or EMA, cohort data trending null-to-favorable, monitoring language on labels reflecting category history and trial exclusions rather than hidden evidence. The genuinely open frontier is the excluded population — people with severe recent psychiatric illness — where individualized, specialist-involved decisions are the honest standard. And the sentence worth keeping from this entire file: if any of this is personal rather than academic, the right resource is a clinician or crisis line today, not a better literature review.
What about tirzepatide specifically?
The 2023 signal and the headline reviews centered on semaglutide and liraglutide — partly a market-share artifact, since tirzepatide was newer with less accumulated exposure. Tirzepatide’s weight-management label carries the same category-standard monitoring language, its trials (which used the same psychiatric exclusions) showed no suicidality excess, and no distinct tirzepatide signal has been established in the surveillance data as of our last verification. The honest asterisk is simply thinner data: fewer patient-years than semaglutide, a gap that closes on its own with time. Nothing currently suggests the molecules differ on this axis; nothing yet proves they don’t.
Questions worth asking any program
Does your intake ask about psychiatric history, current medications, and past self-harm — and does a clinician actually read the answers? Who do I contact, and how fast is the response, if my mood changes mid-treatment? Will you coordinate with my therapist or psychiatrist if I sign a release? A program that treats those as unusual questions has told you its model. The monitoring the label asks for is only real if someone is on the other end of it.
Where the research goes next
Three threads worth watching: the FDA’s ongoing Sentinel-system analyses, which can detect small effects case reports never could; emerging studies that deliberately include people with psychiatric conditions, closing the trial-exclusion gap; and the mechanistic reward-pathway work — spanning alcohol, nicotine, and compulsive behaviors — that may eventually explain why the cohort data keeps trending favorable. Each will generate headlines; the reading rubric above will still apply.
References
FDA preliminary evaluation of suicidality reports with GLP-1 RAs, January 2024 — fda.gov. EMA PRAC review conclusion, April 2024 — ema.europa.eu. Wang W, et al. Semaglutide and risk of suicidal ideation, Nature Medicine, 2024. Wegovy prescribing information (psychiatric monitoring language) — novo-pi.com. Locate studies via PubMed; verify current review status before quoting. Educational content, not medical advice. If you are struggling, contact your clinician or, in the U.S., call or text 988.