Peptide Post-Marketing Surveillance Explained
Peptide Safety
29,000+ reports
Nearly 29,000 neurological adverse event reports associated with GLP-1 receptor agonists were identified in the FAERS database, yielding 19 distinct safety signals.
Chen et al., Scientific Reports, 2025
Chen et al., Scientific Reports, 2025
View as imageClinical trials test peptide drugs in thousands of patients over months to years. Post-marketing surveillance monitors them in millions of patients over decades. This difference in scale is why some safety signals only emerge after a drug reaches the market. The FDA Adverse Event Reporting System (FAERS) is the primary tool for detecting these signals, and it has been especially active for GLP-1 receptor agonists, where new pharmacovigilance analyses are published almost weekly. Chen et al. (2025) identified 19 distinct neurological safety signals from nearly 29,000 reports associated with GLP-1 agonists in FAERS.[1] This article explains how post-marketing surveillance works, what it has found for peptide drugs, and what these signals mean. For the broader picture of peptide safety gaps, see our pillar article on compounded peptide safety monitoring.
Key Takeaways
- FAERS is the FDA's primary post-marketing surveillance database, collecting voluntary reports of adverse events from healthcare providers, patients, and manufacturers
- Disproportionality analysis compares observed adverse event rates for a drug against expected background rates to identify statistical safety signals
- Chen et al. (2025) found 19 neurological safety signals for GLP-1 agonists from nearly 29,000 FAERS reports, including dizziness, tremor, and dysgeusia
- Compounded GLP-1 agonists showed higher rates of abdominal pain, nausea, gallbladder inflammation, and suicidality reports compared to branded formulations (McCall et al., 2026)
- Cheng et al. (2026) used both FAERS and VigiBase (WHO) to assess ophthalmic risks, finding semaglutide had the strongest signal for nonarteritic anterior ischemic optic neuropathy
- Safety signals from FAERS are hypothesis-generating, not proof of causation; they trigger focused investigations, not automatic drug withdrawals
How FAERS Works
The FDA Adverse Event Reporting System (FAERS) is a database of spontaneous adverse event reports submitted to the FDA. Healthcare professionals, patients, and drug manufacturers can report any suspected adverse event associated with any marketed drug. Manufacturers are required by law to submit reports they receive; reports from clinicians and patients are voluntary.
FAERS contains over 20 million reports and receives more than 2 million new reports annually. Each report includes the drug name, adverse event description (coded using the Medical Dictionary for Regulatory Activities, or MedDRA), patient demographics, outcome severity, and reporter type. The database is publicly searchable and has become the primary data source for pharmacovigilance studies of peptide drugs.
The critical limitation: FAERS captures spontaneous reports, not systematic data collection. This means reporting rates depend on awareness, motivation, and access. A drug that receives media attention (like semaglutide) will accumulate more reports than an equivalent drug that does not, regardless of actual adverse event rates. This "notoriety bias" or "stimulated reporting" must be accounted for in any FAERS analysis.
Disproportionality Analysis: The Core Method
Pharmacovigilance researchers do not simply count adverse event reports. They use disproportionality analysis to compare the proportion of a specific adverse event reported for a drug against the proportion expected based on all other drugs in the database.
The most common metric is the Reporting Odds Ratio (ROR). If semaglutide has 500 reports of pancreatitis out of 10,000 total reports, and the background rate of pancreatitis reports across all other drugs is 50 out of 10,000, the ROR would be approximately 10, indicating a disproportionate signal. Additional metrics include the Proportional Reporting Ratio (PRR), Information Component (IC), and Empirical Bayesian Geometric Mean (EBGM), each with different statistical properties.
A positive signal does not prove causation. It means the adverse event is reported more frequently than expected and warrants further investigation through controlled studies, mechanistic research, or regulatory action.
Neurological Safety Signals for GLP-1 Agonists
Chen et al. (2025) conducted one of the most comprehensive neurological pharmacovigilance analyses of GLP-1 agonists to date. From the FAERS database, they identified nearly 29,000 neurological adverse event reports associated with GLP-1 receptor agonists and found 19 distinct safety signals.[1]
The signals included dizziness, tremor, dysgeusia (altered taste), lethargy, taste disorder, presyncope, parosmia (distorted smell), and allodynia (pain from normally non-painful stimuli). Several of these, particularly taste and smell disturbances, had not been prominent in clinical trial data and emerged primarily through post-marketing surveillance.
Lev et al. (2026) identified an unexpected signal: Wernicke encephalopathy (a serious neurological condition caused by thiamine deficiency) associated with GLP-1 agonists. The proposed mechanism is that severe nausea and vomiting, combined with reduced food intake, can deplete thiamine stores in vulnerable patients. This signal would be nearly impossible to detect in clinical trials, where nutritional monitoring is routine, but emerges in real-world use where patients may not receive adequate nutritional support.[2]
Ophthalmic Safety Signals
Cheng et al. (2026) conducted a multi-database analysis using both FAERS and VigiBase (the WHO's global adverse event database) to assess ophthalmic risks of GLP-1 agonists. They used advanced signal detection methods to minimize false positives.[3]
The strongest ophthalmic signal was for nonarteritic anterior ischemic optic neuropathy (NAION), a sudden loss of vision caused by reduced blood flow to the optic nerve. Semaglutide had 2,878 NAION-related cases in VigiBase and 2,047 in FAERS. Whether GLP-1 agonists directly cause NAION or whether the signal reflects confounding factors (patients with obesity and diabetes are already at higher NAION risk) remains unresolved. The FDA added language about NAION to semaglutide's prescribing information in 2024 based on these post-marketing signals.
Gastrointestinal Adverse Events
Chiang et al. (2025) published a systematic review and meta-analysis in Gastroenterology examining GLP-1 agonist gastrointestinal adverse events, combining both clinical trial data and real-world evidence. They confirmed that nausea, vomiting, and diarrhea are dose-dependent class effects, and identified delayed gastric emptying (gastroparesis) as an area requiring longer-term surveillance.[4]
For a detailed discussion of GLP-1 gastrointestinal effects, see our article on GLP-1 side effects.
Tirzepatide-Specific Surveillance
As a newer drug with a dual mechanism (GLP-1 + GIP), tirzepatide has been subject to intensive post-marketing monitoring.
Chen et al. (2025) published one of the first systematic FAERS analyses of tirzepatide, identifying safety signals across gastrointestinal, hepatobiliary, and metabolic categories. The signal profile was broadly similar to GLP-1 agonists but with some differences attributable to GIP receptor activation.[5]
Gu et al. (2026) updated the tirzepatide pharmacovigilance analysis with FAERS data through Q1 2025, using an adapted time-to-onset method to distinguish early-onset from late-onset adverse events. Early-onset events (within the first weeks) were predominantly gastrointestinal (nausea, vomiting), while late-onset events included more metabolic and musculoskeletal signals.[6]
Almansour et al. (2025) analyzed real-world tirzepatide safety concerns across 2022-2025, providing the most comprehensive early post-marketing safety profile available for the drug.[7]
Compounded vs. Branded Peptide Safety
One of the most consequential pharmacovigilance findings for the peptide space involves compounded GLP-1 agonists, versions of semaglutide and tirzepatide produced by compounding pharmacies rather than the original manufacturers.
McCall et al. (2026) compared FAERS reports for compounded GLP-1 agonists versus FDA-approved formulations. Compounded products showed a higher likelihood of reports for abdominal pain, nausea, diarrhea, gallbladder inflammation, and suicidality. The authors noted that these differences could reflect lower product quality, dosing inconsistencies, or differences in the patient populations using compounded products (who may have less medical oversight).[8]
This finding contributed to the FDA's decision to send more than 50 warning letters to GLP-1 compounders and manufacturers in September 2025. For more on the compounding safety gap, see our pillar article on compounded peptide safety. For the longer-term view on what remains unknown, see our article on long-term peptide safety data gaps.
Suicidality and GLP-1 Agonists: A Case Study in Signal Detection
The question of whether GLP-1 agonists increase suicidal ideation illustrates both the power and limitations of post-marketing surveillance.
Seijas-Amigo et al. (2026) conducted a comparative pharmacovigilance analysis of suicidality-related adverse events among GLP-1 and non-GLP-1 anti-obesity drugs in FAERS. They found that suicidality signals were present for GLP-1 agonists but also for other weight loss medications, making it difficult to determine whether the signal reflects a drug-specific effect or a consequence of rapid weight loss and body image changes common to all effective obesity treatments.[9]
The European Medicines Agency (EMA) and FDA both investigated GLP-1 agonist suicidality signals in 2023-2024. The EMA concluded that available evidence did not support a causal association, while the FDA continued monitoring. This outcome is typical for FAERS signals: the database generates hypotheses that require different study designs (cohort studies, case-control studies, mechanistic research) to confirm or refute.
Thyroid Cancer and the Limits of Rodent Signals
Abi et al. (2025) examined the connection between weight-loss medications and thyroid cancer using FAERS data. GLP-1 agonists carry a black box warning for thyroid C-cell tumors based on rodent studies, but human post-marketing data has not confirmed a clear increase in thyroid cancer risk.[10]
This illustrates a different kind of safety signal: one that originated in preclinical studies and persists as a theoretical concern that post-marketing surveillance has not validated. The discrepancy may reflect species-specific biology (rodent thyroid C-cells express GLP-1 receptors more abundantly than human C-cells) or insufficient follow-up time in human data. For a deeper look, see our article on GLP-1 agonists and thyroid cancer risk.
International Surveillance: VigiBase and EudraVigilance
FAERS is not the only post-marketing safety database. VigiBase, maintained by the WHO's Uppsala Monitoring Centre, collects adverse event reports from over 170 countries. EudraVigilance, operated by the European Medicines Agency (EMA), covers the European Union. Japan's PMDA maintains its own system, as does Health Canada.
Multi-database analyses, like the Cheng et al. (2026) study that used both FAERS and VigiBase to assess ophthalmic risks, are increasingly common and more reliable than single-database studies. If a safety signal appears consistently across multiple independent databases with different reporting populations and healthcare systems, it carries more weight than a signal from one database alone.
The practical difference: a FAERS-only analysis might detect a signal that reflects uniquely American prescribing patterns, insurance coverage limitations, or media-driven reporting surges. A signal that replicates across FAERS, VigiBase, and EudraVigilance is more likely to reflect a genuine drug effect.
For peptide drugs, international surveillance is especially important because prescribing patterns differ dramatically between regions. Semaglutide for weight loss is prescribed far more frequently in the United States than in Europe or Asia, creating different exposure levels and patient demographics across databases. GLP-1 agonist compounding is primarily an American phenomenon, so FAERS captures compounding-related safety data that VigiBase does not.
Limitations of Post-Marketing Surveillance
FAERS-based pharmacovigilance has inherent limitations that apply to all peptide drug safety analyses. Spontaneous reporting captures only a fraction of actual adverse events (estimated at 1-10% of events). Reporting is biased toward serious events, new drugs, and drugs receiving media attention. FAERS lacks denominator data (total number of patients taking a drug), making true incidence rates impossible to calculate. Confounding by indication is pervasive: patients taking GLP-1 agonists have obesity and diabetes, conditions that independently increase risk for many of the adverse events detected. Disproportionality signals cannot distinguish causation from correlation, detection bias, or confounding. These limitations do not make FAERS data useless, but they require that signals be interpreted as hypothesis-generating rather than hypothesis-confirming.
The Bottom Line
Post-marketing surveillance through the FAERS database has identified dozens of safety signals for peptide drugs like semaglutide and tirzepatide, from neurological effects (dizziness, dysgeusia, Wernicke encephalopathy) to ophthalmic risks (NAION) to gastrointestinal complications. Compounded GLP-1 agonists show higher adverse event reporting rates than branded formulations. These signals are hypothesis-generating, not proof of causation: they trigger focused investigations that may or may not confirm a true safety risk. Understanding how this system works is essential for interpreting the steady stream of peptide drug safety headlines.