GLP-1 Drugs and Suicidal Ideation: Separating Evidence from Headlines

In July 2023, three patients in Iceland -- all taking liraglutide or semaglutide -- reported suicidal thoughts. The Icelandic Medicines Agency forwarded these reports to the European Medicines Agency. Within weeks, the EMA opened a formal safety review. The FDA followed. Headlines exploded.

Two years later, both agencies have completed their investigations. The FDA has gone further, requesting that manufacturers remove suicidality warnings from GLP-1 drug labels. But the headlines left a mark. Patients are worried. Some have stopped their medications. Others never started.

This article walks through every layer of evidence -- from the initial signal through the regulatory conclusions -- because a topic this serious deserves precision, not summaries.

How the Investigation Started: Three Patients in Iceland

The signal that triggered global regulatory attention was remarkably small: three individual reports of suicidal ideation and self-injury in people taking GLP-1 receptor agonists in Iceland. The Icelandic Medicines Agency relayed these to the EMA's Pharmacovigilance Risk Assessment Committee (PRAC) on July 3, 2023.

To be clear, three reports from a single country do not establish a drug safety problem. But regulators are supposed to investigate signals, however small, when they involve serious outcomes. That is exactly what happened.

The EMA review covered all GLP-1 receptor agonists: semaglutide, liraglutide, exenatide, dulaglutide, and lixisenatide.

What followed was an intense 18-month period of investigation that involved three major regulatory agencies, multiple independent research teams, and the analysis of data from millions of patients. The media coverage ranged from responsibly cautious to sensationally alarming. And through it all, tens of millions of people were taking these drugs and wondering whether they should stop.

The EMA Review: No Causal Link Found

The PRAC completed its review in April 2024. The committee's conclusion: the available evidence does not support a causal association between GLP-1 receptor agonists and suicidal or self-injurious thoughts and actions.

The review analyzed:

Results from a large study using electronic health records, comparing the incidence of suicidal thoughts in patients with overweight and type 2 diabetes treated with semaglutide versus non-GLP-1 medications
Non-clinical (animal) studies
Clinical trial data
Post-marketing surveillance reports

The PRAC determined that no update to product information was warranted. This was a clear conclusion: after reviewing every available type of evidence, the EMA's safety committee found no reason to believe these drugs cause suicidal thoughts.

The final study report, dated February 2024, is publicly available. It is worth reading for anyone who wants to see exactly what evidence the committee reviewed and how they weighed it.

The FDA Investigation: From Concern to Label Removal

The FDA took a parallel but more extended path.

January 2024: Preliminary Review

In its initial Drug Safety Communication, the FDA reported that a preliminary review of clinical trial and postmarketing data "did not find an association between the use of GLP-1 RAs and the occurrence of suicidal ideation or behavior," though it noted "considerable uncertainty in the risk estimate due to the small number of cases in individual trials."

The Sentinel Study

To get a definitive answer, the FDA conducted a comprehensive meta-analysis across GLP-1 RA drug development programs and ran a massive study through its Sentinel System -- a network of electronic health records covering millions of patients.

The Sentinel study included 2,243,138 users: 1,161,983 who started a GLP-1 RA and 1,081,155 who started an SGLT2 inhibitor (a different class of diabetes/obesity medication used as a comparator). The data spanned 10 data partners during 2015-2023.

After controlling for baseline confounders, the FDA did not find an increased risk of intentional self-harm in GLP-1 RA users compared to SGLT2i users, including in the subgroup of patients with both type 2 diabetes and obesity.

March 2025: Warning Removal Requested

Based on the totality of evidence, the FDA concluded that its evaluation "did not identify an increased risk of suicidal ideation or behavior with the use of GLP-1 RA medications" and formally requested that manufacturers remove suicidal behavior and ideation warnings from GLP-1 RA labels.

This is notable. Regulatory agencies rarely ask for warning removal. It reflects a high level of confidence in the safety data.

The UK Review

The UK Medicines and Healthcare products Regulatory Agency (MHRA) launched its own review in July 2023. According to the MHRA's Yellow Card reporting scheme, there had been 28 reports of suicidal and self-injurious behavior in patients taking semaglutide and 23 reports for liraglutide. As of this writing, the MHRA has not issued a final determination, though its position has been consistent with the EMA's assessment.

What the Clinical Trial Data Shows

The most rigorous evidence on suicidality comes from the STEP trial post hoc analysis by Wadden et al. (2024, JAMA Internal Medicine), which assessed suicidal ideation and behavior using the Columbia-Suicide Severity Rating Scale across 3,377 participants in STEP 1, 2, 3, and 5.

The findings were straightforward:

1% or fewer participants reported suicidal ideation or behavior in either the semaglutide or placebo group
No statistically significant difference between groups
Results were consistent across all four trials

The limitation, again, is that these trials excluded people with histories of suicide attempts, active suicidal ideation, or recent major depressive disorder diagnoses -- precisely the population at highest baseline risk.

It is also worth noting that the STEP trials were not specifically powered to detect rare events like suicidal ideation. With only about 1% of participants reporting any suicidal ideation in either group, these trials would need to be orders of magnitude larger to detect a small increase in risk with statistical confidence. The absence of a signal in the STEP data is reassuring but not definitive proof that no risk exists -- it means any risk, if present, is too small to detect in a study of this size.

The FAERS Database: The American Pharmacovigilance Story

Before diving into the broader pharmacovigilance data, it helps to understand what the FDA's own adverse event reporting system showed.

The FDA Adverse Event Reporting System (FAERS) receives voluntary reports from healthcare professionals, consumers, and manufacturers. An analysis of FAERS data found that liraglutide was associated with increased reporting of suicidal ideation relative to metformin (ROR: 3.26) and insulin (ROR: 3.62), but -- importantly -- with lower reporting of actual suicide attempts and completed suicides relative to both comparison drugs.

Other GLP-1 RAs in the FAERS database, including dulaglutide, exenatide, and lixisenatide, did not show disproportionate reporting of suicidal events.

This FAERS pattern is interesting because it suggests a disconnect: more reports of suicidal thinking but not more reports of suicidal action. This could indicate reporting bias (people are more likely to report thoughts than actions to a voluntary database), or it could suggest that whatever drives the pharmacovigilance signal is distinct from actual suicide risk.

Pharmacovigilance Data: The Source of the Confusion

This is where things get complicated. Pharmacovigilance databases -- FAERS in the US, EudraVigilance in Europe, VigiBase globally -- consistently show elevated reporting of suicidal ideation with some GLP-1 drugs. Understanding why these numbers exist, and what they mean, is essential.

EudraVigilance Data

An analysis of the European Pharmacovigilance Database identified 230 reports of suicidal events associated with GLP-1 RAs. Suicidal ideation accounted for 65.3% of these reports, and suicide attempts for 19.5%.

Using disproportionality analysis (a statistical method that compares how often a drug-event pair is reported versus how often you'd expect it by chance), the study found:

Semaglutide had a higher reporting probability of suicidal events than dulaglutide (ROR 2.05) and exenatide (ROR 1.81)
Liraglutide had even higher reporting probability vs. dulaglutide (ROR 3.98) and exenatide (ROR 3.52)
Semaglutide had lower reporting probability than liraglutide (ROR 0.51)

WHO VigiBase Data

A study using the WHO global database (data through January 2024) found significantly elevated reporting odds ratios for suicidal ideation with semaglutide (ROR 5.82), liraglutide (4.03), and tirzepatide (2.25).

Those numbers look alarming. But here is what they do not tell you.

Why Pharmacovigilance Data Can Mislead

Spontaneous reporting databases have well-known limitations:

1. Notoriety bias. After the 2023 headlines, clinicians and patients became more likely to report psychiatric events with GLP-1 drugs. This alone can inflate reporting odds ratios substantially.

2. No denominator. We know there were 230 suicidal event reports for GLP-1 drugs in EudraVigilance, but out of how many total users? Semaglutide alone was prescribed to tens of millions of people during the study period. Without knowing the base rate, raw numbers are hard to interpret.

3. Confounding by indication. People taking GLP-1 drugs for obesity often have higher baseline rates of depression and suicidal ideation than the general population. Obesity itself is associated with a roughly 55% increased risk of developing depression.

4. Weber effect. Newly launched or newly popular drugs receive disproportionate adverse event reports because more attention is focused on them.

5. The antidepressant signal. A 2025 meta-analysis found that the reporting odds for semaglutide-related suicidal ideation were significantly elevated in patients co-prescribed antidepressants (ROR 4.45) or benzodiazepines (ROR 4.07). This suggests that pre-existing psychiatric conditions -- not the GLP-1 drug itself -- may drive much of the signal.

Pharmacovigilance data is good at generating hypotheses. It is not good at proving causation. That is why regulatory agencies supplement it with cohort studies and clinical trial analyses.

Cohort Studies: The Middle Ground

Spanish Population-Based Study (2024)

A study published in Diabetologia combined population-wide databases covering five million inhabitants in Spain, including all adults with obesity who started GLP-1 RA or SGLT2 inhibitor treatment for type 2 diabetes from 2015 to 2021.

In per-protocol analyses, there was no evidence that GLP-1 RAs increased suicidal ideation or self-injury (HR 1.04; 95% CI: 0.35-3.14). Intention-to-treat analyses showed a similar non-significant result (HR 1.36; 95% CI: 0.51-3.61).

The researchers noted the rarity of suicidal events made the confidence intervals wide, and recommended caution in interpretation.

The Wang et al. Study (Nature Medicine, 2024)

A large US cohort study using propensity score matching examined 240,618 patients with overweight/obesity and 1,589,855 patients with type 2 diabetes. Semaglutide was compared to non-GLP-1 RA anti-obesity medications.

The findings went against the pharmacovigilance signals:

In patients with obesity, semaglutide was associated with lower risk of incident suicidal ideation (HR 0.27) and recurrent suicidal ideation (HR 0.44)
Similar protective patterns appeared in patients with type 2 diabetes

French Case-Time-Control Study (2024)

A nationwide French study published in eClinicalMedicine used a case-time-control design, which compares medication exposure during a hazard period immediately before the event to a control period earlier in the same patient's history. This design is particularly good at controlling for stable patient characteristics (like baseline psychiatric risk) because each patient is their own control.

The results showed that semaglutide was associated with lower risks of both incident and recurrent suicidal ideation in overweight/obese patients when compared to non-GLP-1 RA medications. Similar findings appeared in patients with diabetes.

What the Cohort Studies Tell Us Collectively

Three independent cohort studies -- from Spain, the US, and France -- using different methodologies and different populations all pointed in the same direction: no increased risk, and possibly a decreased risk of suicidal ideation with GLP-1 RA use.

This consistency across geographies and methods is important. A single study can be wrong for many reasons. Three studies reaching the same conclusion through different paths provides much stronger evidence.

Systematic Reviews: Putting It All Together

Two major systematic reviews in 2025 attempted to synthesize the entire evidence base.

Diabetes/Metabolism Research and Reviews (2025)

A systematic review and meta-analysis found:

Four pharmacovigilance studies showed no disproportionate increase in suicidality
However, a significant signal emerged for semaglutide-related suicidal ideation (ROR 1.45; 95% CI: 1.18-1.77) in pharmacovigilance data
Patients on antidepressants (ROR 4.45) or benzodiazepines (ROR 4.07) had substantially elevated reporting

Journal of Psychiatric Research (2025)

A separate systematic review identified 22 studies meeting inclusion criteria: 10 pharmacovigilance studies and 12 cohort studies. Their assessment:

Pharmacovigilance studies show disproportionate reporting of suicidal ideation with semaglutide and liraglutide
Cohort studies consistently do not show increased suicidality
The overall conclusion: the association between GLP-1 RAs and suicidal outcomes in pharmacovigilance databases has been "mixed," while cohort data and clinical trials have not confirmed an increased risk

Why Headlines Got Ahead of the Evidence

It is worth understanding why the media coverage of this issue was so outsized relative to the actual evidence.

Suicidal ideation is inherently newsworthy. It triggers fear, urgency, and clicks. When a regulatory agency opens a safety review, the nuance of "we are investigating a signal" gets flattened into "Ozempic linked to suicidal thoughts" in headlines. The distinction between "linked" (in a headline) and "causally associated" (in science) is enormous, but most readers do not make it.

The timing also mattered. The investigation opened in mid-2023, at the peak of GLP-1 media frenzy. Semaglutide was already the most-discussed drug in a generation. Any safety concern, however preliminary, was guaranteed front-page treatment.

The result was a feedback loop: headlines drove more reporting of suicidal ideation to pharmacovigilance databases, which generated more data points, which generated more headlines. This is a well-documented pattern in drug safety -- it is one reason pharmacovigilance data should never be interpreted in isolation.

The FDA's March 2025 decision to remove suicidality warnings received a fraction of the coverage that the original investigation received. This asymmetry is predictable but unfortunate, because it means many patients still carry anxiety from headlines that the evidence ultimately did not support.

How to Think About This Evidence

The evidence layers into a clear hierarchy:

Evidence Type	What It Shows	Strength
Clinical trials (STEP)	No increased suicidality; 1% or fewer in both groups	Strongest for causation
Large cohort studies	No increased risk; some suggest protective effect	Strong for real-world risk
Systematic reviews	Pharmacovigilance signals exist but not confirmed by cohort data	Moderate (depends on source)
Pharmacovigilance data	Elevated reporting for semaglutide and liraglutide	Weakest for causation (hypothesis-generating)

The strongest evidence types consistently show no risk. The weakest evidence type shows a signal that the stronger methods have not confirmed.

Both the FDA and EMA arrived at the same conclusion: no causal link established.

Who Should Still Be Cautious

"No causal link" does not mean "no individual risk." Several groups warrant closer monitoring:

People with pre-existing psychiatric conditions. The 2025 meta-analysis finding that reporting odds are 4-5 times higher in people on antidepressants or benzodiazepines suggests this population may be more vulnerable -- or at least deserves more careful observation.

People experiencing rapid lifestyle changes. Dramatic weight loss can trigger identity disruption, loss of food-based coping mechanisms, and shifts in social dynamics. These psychological stressors could contribute to mood changes independent of any pharmacological effect.

People with a history of eating disorders. The relationship between appetite suppression, body image, and suicidality in this population is complex and understudied in the GLP-1 context.

People who use food as their primary coping mechanism. When GLP-1 drugs remove the drive to eat emotionally, the psychological distress that was being managed through food can surface with full force. For someone already vulnerable, this could contribute to a mental health crisis -- not because the drug caused suicidal ideation pharmacologically, but because it removed the coping tool that was preventing it.

Anyone discontinuing GLP-1 therapy. The period after stopping these drugs is psychologically complex. Weight regain is common and can trigger significant distress, especially if the patient built their self-concept around their new body. Mental health monitoring during and after GLP-1 discontinuation is a gap in current clinical practice that deserves attention.

The Responsible Takeaway

Here is what the evidence actually supports:

GLP-1 receptor agonists, including semaglutide, do not appear to cause suicidal ideation or behavior. Two major regulatory agencies reviewed the full evidence base and reached this conclusion. The FDA went so far as to remove suicidality warnings from labels.
Pharmacovigilance signals are real but almost certainly driven by confounding, reporting bias, and population differences rather than a true pharmacological effect.
Suicidal ideation is rare but serious at any rate. Whether or not GLP-1 drugs contribute to risk, any patient experiencing changes in mood, hopelessness, or suicidal thoughts should contact their healthcare provider immediately -- regardless of what they are taking.
Patients with psychiatric histories were underrepresented in the strongest studies. The evidence is most reassuring for psychiatrically healthy individuals. More data is needed for people with depression, bipolar disorder, or histories of suicidal behavior.
Monitoring is still appropriate. Even without a confirmed causal link, tracking mood and mental health during any significant medical treatment is good clinical practice. The psychological effects of rapid weight loss alone justify paying attention.

If you are on a GLP-1 medication and experiencing thoughts of self-harm, call 988 (the Suicide and Crisis Lifeline) or go to your nearest emergency department. These symptoms need evaluation regardless of their cause.

This article is part of the GLP-1 and Mental Health series. For related topics, see our coverage of semaglutide and depression, emotional eating and food noise, body image after weight loss, and GLP-1 agonists and alcohol reduction.