GLP-1 & Incretin Science

Adding a Warning Needs Suspicion; Removing One Needs Evidence

UKPWritten and published by UK Peptides4 min readUpdated 3 cited sources

The short answer

Safety warnings are added when a signal is plausible and removed only when evidence of absence is strong enough. That asymmetry is deliberate, and the GLP-1 suicidality question is an unusually complete worked example of it running to a conclusion.

Key facts

Signal raised
Postmarketing reports of suicidal ideation and behaviour
FDA investigation opened
July 2023
FDA preliminary communication
January 2024
EMA PRAC conclusion
12 April 2024
PRAC decision on labelling
No update to the product information warranted
FDA request to remove warning
13 January 2026
Evidence base for removal
91 placebo-controlled trials, 107,910 participants
Elapsed time
Roughly two and a half years

Signals are opened easily and closed slowly, on purpose

The threshold for taking a safety question seriously is deliberately low. Spontaneous reports arrive unverified and uncontrolled, and a regulator that waited for proof before investigating would be too slow to be useful. The threshold for declaring a question settled is deliberately high, because the cost of wrongly removing a warning is borne by patients while the cost of wrongly keeping one is mostly borne by the drug. The result is an asymmetry that looks like inconsistency and is not: the same body can say a warning is justified on suspicion and, later, that removing it requires far more than suspicion to the contrary.

The sequence, with dates

The FDA opened a formal investigation in July 2023 after postmarketing reports of suicidal ideation and behaviour in people taking GLP-1 receptor agonists. Its preliminary Drug Safety Communication came in January 2024 and reported no clear evidence of causation, while noting that only a small number of cases had been observed in individual trials, so the risk estimate was uncertain. In April 2024 the European Medicines Agency's Pharmacovigilance Risk Assessment Committee concluded that the available evidence did not support a causal association for dulaglutide, exenatide, liraglutide, lixisenatide and semaglutide, having reviewed non-clinical studies, clinical trials, post-marketing surveillance and additional studies including one the agency conducted itself. On 13 January 2026 the FDA requested that manufacturers remove the suicidal behaviour and ideation warning from labelling for Saxenda, Wegovy and Zepbound.

Research material referenced

Retatrutide 10mg, third-party HPLC tested

Buy Retatrutide · £59.99

The two regulators did different things, and the difference matters

This is where most summaries collapse the story. The PRAC's April 2024 conclusion was that no update to the product information was warranted. In other words, it declined to add anything, and left existing labelling as it stood. The FDA's January 2026 action was to request removal of a warning that was already there. Declining to add and choosing to subtract are different regulatory acts requiring different amounts of evidence, and the twenty-one months between them is roughly what it took to assemble the second.

What made the removal possible

Not more reports, and not more observational studies. The FDA's meta-analysis pooled 91 placebo-controlled clinical trials from across GLP-1 receptor agonist development programmes: 107,910 participants, 60,338 on a GLP-1 receptor agonist and 47,572 on placebo, examining suicidal behaviour and ideation along with other psychiatric adverse events including anxiety, depression, irritability and psychosis. Randomised comparison addresses the central confounder directly (that people prescribed these drugs differ systematically from those who are not, including in baseline psychiatric risk), and a hundred thousand participants supplies the precision that individual trials, each containing only a handful of events, could not.

What the closure does and does not say

It says that across a very large randomised dataset, no increased risk of these outcomes was detected. It does not say the risk is zero, because no study design establishes that; a large randomised analysis places an upper bound on how big an effect could have gone unnoticed, which is the strongest available form of the claim. It also does not extend to populations the trials excluded, to durations longer than the trials ran, or to compounds not in the analysis. And it says nothing about whether any individual person should stop or continue a prescription, which is a clinical question.

Why this case is worth keeping as a template

Most safety signals do not resolve. They persist as unquantified concerns, attached to labels, cited in coverage, and never formally closed, because the study that would close them is expensive and nobody's incentive is to run it. This one closed, in public, with dated regulatory documents on both sides of the Atlantic and a stated evidence base. When the next signal in this class arises, as it will, the useful questions are the ones this case answers: who opened it and when, what evidence would close it, has anyone assembled that evidence, and has any regulator acted on it.

Quick reference

DateBodyAction
July 2023FDAInvestigation opened after postmarketing reports
January 2024FDAPreliminary communication: no clear evidence, estimate uncertain
12 April 2024EMA PRACNo causal association; no labelling update warranted
13 January 2026FDARequests removal of the warning from labelling

Frequently asked questions

Does removing a warning mean the drug is safe?

It means that specific concern was investigated and not substantiated. Every one of these medicines carries other warnings that remain, and removal of one says nothing about the rest of the safety profile.

Why did the EMA not remove the warning in 2024?

It concluded that the evidence did not support a causal association and that no update to the product information was warranted, which is a decision not to change the labelling in either direction. Removal is a separate act requiring its own justification, and the large trial-level analysis that supported the FDA's 2026 request did not exist in April 2024.

Should someone with a psychiatric condition avoid these drugs?

That is a question for a prescribing clinician who knows the person's history. Commentary accompanying these reviews has consistently recommended monitoring in people with existing psychiatric conditions, which is a different thing from a contraindication.

Does this apply to research peptides?

No. Everything described here concerns licensed medicines, their clinical trial programmes and their regulatory labelling. Material supplied for laboratory research has no labelling, no regulatory assessment and no clinical evidence base of any kind, and is not an alternative to a licensed medicine.

Continue researching

Related guides and reference material.

Primary sources & clinical trials

Published papers, trial registrations, regulator pages and company announcements. Links open the original.

UKP

Written and published by UK Peptides, the trading name of AURYXRESEARCH LTD (company 17388727, registered in England and Wales).

Questions or a correction? Email hello@ukpeptides.org or call +44 7426 987848.

Corrections log

More GLP-1 & Incretin Science articles

Research use only. The information above is provided for scientific and educational reference. Compounds referenced are not approved for human use and are supplied for in vitro research or reference-material purposes only. No efficacy, safety, or therapeutic claims are made.