GLP-1 FAERS safety signals explained

A pharmacovigilance analysis published in Diabetes, Obesity and Metabolism on 3 September 2026 examined 243,114 primary-suspect adverse event reports for GLP-1 receptor agonists and tirzepatide filed with the US FDA Adverse Event Reporting System (FAERS) from 2021 through early 2026. The authors, Ren L, Zhang Y, Fang F, and Ma Y from Shanghai East Hospital, Tongji University School of Medicine, set out to describe safety signal patterns and medication-use risks across the approved GLP-1 class^[1].

This is a large FAERS analysis, not a clinical trial. It identifies patterns in spontaneous reports, so the counts reflect reporting behaviour as much as underlying pharmacology. What it adds is a structured look at how the GLP-1 class is faring in the real world, across a period when prescribing volumes grew sharply following approvals for semaglutide and tirzepatide in obesity^[2].

The report counts by drug

Of the 243,114 primary-suspect reports in the analysis window, tirzepatide accounted for 133,100 records. Semaglutide followed with 55,619 reports. Dulaglutide contributed 38,406. Liraglutide, exenatide, and lixisenatide made up the remainder^[1].

The tirzepatide figure is the largest because tirzepatide is the most recently authorised agent in this group and because it entered the obesity market with a strong commercial push. FAERS report counts are driven by prescribing volume, by how long a drug has been on the market, and by the degree to which adverse events prompt contact with a prescriber or pharmacist. A higher count does not mean a higher per-patient risk.

What types of events appeared most often

The analysis found four dominant adverse-event categories across the GLP-1 class: gastrointestinal effects, injection-site complications, dosing errors, and gastric-emptying impairment^[1]. This is consistent with the class pharmacology and with what the approved prescribing information already lists.

For tirzepatide specifically, the Mounjaro prescribing information (DailyMed) lists nausea, diarrhoea, decreased appetite, vomiting, and constipation as the most common adverse reactions. Severe gastrointestinal events occurred in 1.2 to 1.3 percent of tirzepatide-treated patients in the SURPASS programme, against 0.9 percent on placebo. Acute pancreatitis was observed at a rate of 0.23 per 100 exposure years across the tirzepatide-treated population in the programme^[3].

Dosing errors in the FAERS data are a separate signal from pharmacological adverse effects. Pen-sharing, incorrect dose dialling, and confusion between brands or strengths all generate error reports. The DailyMed labelling for Mounjaro contains an explicit warning against sharing the KwikPen between patients because of bloodborne pathogen transmission risk, which places at least part of the medication-error category outside the pharmacology entirely.

How to read a FAERS signal

FAERS is a spontaneous reporting system. It collects adverse event reports from manufacturers (required by regulation), healthcare professionals, and consumers. Reports are submitted after a suspected event, not in response to systematic follow-up. The FDA uses FAERS data to identify potential safety signals for further review, not to establish causation or calculate incidence rates^[4].

Two features of spontaneous reporting particularly matter for interpreting a paper like this one. First, under-reporting: the reported count is a fraction of events that actually occur, with the fraction varying by event severity and by how widely a drug is covered in media or social channels. Second, confounding by indication: patients taking GLP-1 agents for obesity or type-2 diabetes already carry elevated background rates for some of the events being reported, including renal injury and cardiovascular events.

The authors of the Ren et al. analysis were explicit that their findings represent signal identification and medication-safety patterns, not definitive causal proof or population-level risk comparisons. That framing is the correct one for FAERS work.

What this is not

This analysis does not show that tirzepatide is less safe than semaglutide, or vice versa. The report counts are driven by market dynamics as much as by pharmacology. Tirzepatide's 133,100 reports versus semaglutide's 55,619 reflect tirzepatide's faster uptake in the US obesity market, not a proportionally higher per-patient risk. Head-to-head safety comparisons require active surveillance or randomised trial data with comparable follow-up periods.

The study also does not change the regulatory status of any drug it examined. All six agents were reviewed through formal approval programmes with pre-specified safety databases. The EMA maintains ongoing pharmacovigilance requirements for Mounjaro under its EU marketing authorisation, which has been in place since September 2022. The FDA maintains the same for US approvals. FAERS feeds into those surveillance systems but does not replace them.

Where this lands for readers following GLP-1 safety

The Ren et al. paper is useful for two things. First, it confirms that the known adverse-event profile from clinical trials, gastrointestinal effects and gastric-emptying disruption principally, is showing up in real-world reporting at scale. The signal is coherent with the pharmacology. Second, it flags medication-use errors as a numerically significant category, which points to prescriber education and pharmacy counselling as practical intervention points rather than pharmacological ones. For the per-country regulatory picture, the tirzepatide regulation pages track each jurisdiction's approved labelling and any post-approval safety communications^[2].

What the 243,114 report count does not resolve is whether any of the listed signal categories carries a risk that post-marketing surveillance has not already characterised. That question requires longer follow-up and, for some signals, active surveillance studies rather than FAERS mining. The authors call for integrated safety surveillance methods, and that is the appropriate conclusion from work of this type.