Ozempic (Semaglutide) Adverse Events in FAERS/AEMS: A 2026 Analysis
Semaglutide — sold under the brand names Ozempic (for type 2 diabetes), Wegovy (for chronic weight management), and Rybelsus (oral formulation) — has been one of the most-prescribed drugs in the world since 2023. With prescription volume has come a corresponding wave of adverse event reports in FDA's safety database. This article walks through what's actually in the FAERS/AEMS data for semaglutide across the 2021 Q2 – 2026 Q1 window, what signals are real, what got investigated and dismissed, what's emerging, and how to read the numbers without overinterpreting them.
If you're a clinician, researcher, patient, or journalist trying to understand semaglutide's real-world safety profile beyond what's on the label, this is meant to be a useful starting point. We use the same public AEMS data anyone can access — what we add is the disproportionality analysis, time-series context, and methodology notes that turn a list of reports into a signal picture. Every number below comes from our analytics pipeline running against the latest AEMS quarterly files.
Data window for this analysis: 2021 Q2 through 2026 Q1 (20 quarters). This covers the period from just before the Wegovy weight-management approval (June 2021) through the most recent available AEMS quarterly file. Semaglutide reports before 2021 Q2 — primarily for Ozempic during its diabetes-only period after the December 2017 approval — are not included in the case counts below. Including them would not change the overall picture meaningfully (the pre-2021 reporting volume was small relative to the 2022-2026 surge), but it's worth knowing what's in scope.
A note on terminology: FAERS was renamed to AEMS (Adverse Event Monitoring System) on March 11, 2026. The underlying database is the same; the dashboard is new. We use both names depending on the source date.
The headline numbers
Within the 2021 Q2 – 2026 Q1 window, AEMS contains:
- 66,321 cases involving semaglutide
- Out of 8,531,805 total cases across all drugs in the database for the same window
The reporting curve has been steep. From quarterly data spanning 2021 Q2 through 2026 Q1:
- 2021 Q2 had 477 semaglutide cases
- 2024 Q3 (the largest single quarter) had 7,244 cases
- 2025 Q3 had 15,081 cases
The 30x increase from 2021 to peak quarters reflects the explosion of semaglutide prescribing for weight loss following the Wegovy approval (June 2021) and the broader cultural inflection around GLP-1 drugs in 2023.
Three observations about these numbers before digging deeper:
Report counts are not incidence rates. AEMS captures reported events, not all events. A drug with high public attention sees a higher reporting rate than a drug with similar safety issues but less attention. Semaglutide has had unusually high attention.
A "case" is one ICSR. A single case can report multiple reactions. The 66,321 total cases contain hundreds of thousands of individual reaction events. When you see counts for specific reactions below, those are reaction occurrences within those cases, not separate cases.
FAERS/AEMS doesn't establish causation. This is the standard pharmacovigilance disclaimer. A report of nausea following semaglutide doesn't prove semaglutide caused it. What the data does reveal is patterns that warrant further investigation.
Top reported reactions
Across all 66,321 semaglutide cases, the most-reported adverse events by rate per 1,000 cases:
| Reaction | Cases | Serious | Deaths | Rate / 1k cases |
|---|---|---|---|---|
| Nausea | 11,372 | 3,566 | 101 | 171.5 |
| Vomiting | 8,141 | 3,507 | 93 | 122.8 |
| Diarrhoea | 6,381 | 2,408 | 61 | 96.2 |
| Off-label use | 6,375 | 1,896 | 50 | 96.1 |
| Constipation | 5,131 | 1,786 | 29 | 77.4 |
| Decreased appetite | 4,694 | 1,416 | 41 | 70.8 |
| Impaired gastric emptying | 4,557 | 2,701 | 39 | 68.7 |
| Weight decreased | 3,697 | 735 | 50 | 55.7 |
| Abdominal pain | 3,121 | 1,666 | 46 | 47.1 |
| Product use in unapproved indication | 3,032 | 722 | 44 | 45.7 |
| Abdominal pain upper | 2,942 | 1,039 | 23 | 44.4 |
| Headache | 2,592 | 433 | 15 | 39.1 |
| Fatigue | 2,469 | 508 | 13 | 37.2 |
| Dehydration | 2,314 | 1,626 | 41 | 34.9 |
| Intestinal obstruction | 1,715 | 1,173 | 46 | 25.9 |
A few things worth noting in this table:
- Nausea, vomiting, diarrhoea, constipation — the predictable GI cluster, consistent with semaglutide's mechanism (it slows gastric emptying as part of how it produces satiety and glycemic control).
- Impaired gastric emptying at 68.7 per 1,000 cases is large. Of those 4,557 cases, 2,701 were classified as serious (59%) — a substantially higher serious-case rate than nausea or vomiting, reflecting that gastroparesis is genuinely a more clinically concerning outcome.
- Off-label use and product use in unapproved indication together account for about 14% of all semaglutide cases. This is the weight-loss-prescribing story showing up in the safety data.
- Intestinal obstruction: 1,715 cases, 1,173 (68%) serious. Smaller in absolute count than gastric emptying but a much higher serious-rate.
Disproportionality: which reactions are unusually elevated for this drug
Disproportionality analysis compares how often a reaction appears with semaglutide versus how often it appears with all other drugs in the database. Three metrics together identify "signals":
- PRR (Proportional Reporting Ratio): how much more often a reaction shows up with this product vs all others. PRR ≥ 2 is the standard threshold.
- ROR (Reporting Odds Ratio): the odds of reporting this event with this product vs without it. The 95% confidence interval tells you whether the signal is statistically distinguishable from chance.
- χ² (Yates-corrected): χ² ≥ 4 corresponds to roughly p < 0.05.
A reaction flags as a signal when all three hold (PRR ≥ 2, χ² ≥ 4, with at least 3 cases). Running this against the full database for semaglutide produces a long list. The most-cited signals worth understanding:
Gastrointestinal: large case counts, very strong disproportionality
| Reaction | Cases | PRR | χ² |
|---|---|---|---|
| Cyclic vomiting syndrome | 324 | 285.2 | 28,284 |
| Gallbladder injury | 170 | 217.0 | 13,457 |
| Impaired gastric emptying | 4,557 | 108.2 | 262,139 |
| Gastrointestinal hypomotility | 457 | 81.5 | 22,126 |
| Gastric hypomotility | 76 | 67.8 | 3,224 |
| Cholelithiasis migration | 6 | 51.1 | 176 |
| Gastric ischaemia | 8 | 46.4 | 228 |
| Ileus | 1,057 | 33.6 | 26,425 |
| Obstructive pancreatitis | 91 | 25.9 | 1,788 |
| Acute cholecystitis necrotic | 30 | 24.4 | 545 |
The GI signals dominate. Impaired gastric emptying — at PRR 108 across 4,557 cases — is one of the strongest disproportionality signals you can find for any drug in the database. Ileus (PRR 33.6, 1,057 cases) and gallbladder injury (PRR 217, 170 cases) round out the picture. These signals led to concrete regulatory action: per the FDA's Drug Safety-related Labeling Changes database, FDA added ileus to Ozempic's postmarketing-experience section on September 22, 2023, added "not recommended in patients with severe gastroparesis" along with acute kidney injury and severe GI reaction warnings on January 28, 2025, and strengthened pancreatitis warnings on October 14, 2025.
NAION: the eye signal
The most discussed non-GI signal for semaglutide is non-arteritic anterior ischemic optic neuropathy (NAION), a rare condition where blood flow to the optic nerve is interrupted, often causing sudden permanent vision loss in one eye.
In the AEMS data through 2026 Q1:
- Optic ischaemic neuropathy: 745 cases, PRR 114.3, χ² 44,090 — a very strong signal
This is consistent with the observational research that surfaced the signal. A July 2024 study in JAMA Ophthalmology by Hathaway and colleagues found that patients prescribed semaglutide had a 4.28-fold higher risk of NAION in the diabetes cohort and a 7.64-fold higher risk in the overweight/obese cohort, compared with matched patients on non-GLP-1 RA medications. The European Medicines Agency's safety committee (PRAC) concluded a comprehensive review on June 6, 2025 and recommended adding NAION as a "very rare" side effect (up to 1 in 10,000 people) to the product information for Ozempic, Rybelsus, and Wegovy. The EMA's product information for these semaglutide medicines was updated September 30, 2025.
NAION is a useful example of how postmarketing signals develop. The condition was too rare to surface in pre-approval clinical trials. Real-world surveillance through FAERS/AEMS plus observational cohort studies caught it, and label updates followed. This is exactly the pattern postmarketing pharmacovigilance is designed for.
Thyroid: a smaller but watched signal
Semaglutide carries an FDA boxed warning for thyroid C-cell tumors, based on rodent studies. In the human AEMS data:
- Medullary thyroid cancer: 59 cases, PRR 71.7, χ² 2,589
The signal exists, but the case count is small relative to the prescribed population, and the relationship between rodent findings and human risk has been debated for years. The boxed warning remains.
Weight loss and metabolic effects (also expected)
| Reaction | Cases | PRR |
|---|---|---|
| Weight loss poor | 914 | 119.9 |
| Lack of satiety | 37 | 51.3 |
| Starvation ketoacidosis | 38 | 45.8 |
| Food craving | 452 | 29.7 |
"Weight loss poor" is an interesting one. With 914 cases, this signal reflects patients reporting the drug isn't producing the weight loss they expected. It's a hybrid: partly a real pharmacologic non-response phenomenon, partly a reflection of mismatched expectations from off-label use.
Use- and counterfeit-related signals
| Reaction | Cases | PRR |
|---|---|---|
| Drug dose titration not performed | 308 | 67.4 |
| Counterfeit product administered | 277 | 46.8 |
| Product dose confusion | 60 | 25.4 |
The counterfeit signal (PRR 46.8, 277 cases) is worth flagging. Semaglutide shortages from 2022 through 2025 drove a substantial market for compounded and counterfeit semaglutide. FDA has warned about this repeatedly. The disproportionality analysis surfaces it clearly.
A note on interpreting very high PRR values
Some PRR values in the raw output are absurdly large — for instance, MPL gene mutation: 3 cases, PRR 893.5, or Inner ear operation: 5 cases, PRR 319.1. These aren't real drug effects. They're artifacts of small-numbers statistics:
- When a reaction appears in 2 cases in the entire database outside semaglutide and in 3 cases with semaglutide, the ratio explodes. But the absolute counts are too low to support any inference.
A useful rule: filter out signals where the case count (a) is below some threshold — say, 10 — before drawing any conclusions. The signals worth taking seriously are those with both high disproportionality AND meaningful case counts. The GI signals above (thousands of cases) and NAION (745 cases) clear that bar easily. Single-digit case counts don't.
This is one of the most common failure modes when people analyze FAERS data without considering the underlying statistics. A PRR of 893 sounds dramatic; a PRR of 893 across 3 cases is not.
Emerging signals: what's getting reported more in 2025-2026
Disproportionality tells you which reactions are over-represented. Time-series analysis tells you which reactions are increasing. For semaglutide specifically, the emerging-signal picture as of 2026 Q1 is striking.
Methodology: split the analysis range into "baseline" (2021 Q2 through 2025 Q1, 16 quarters) and "recent" (2025 Q2 through 2026 Q1, 4 quarters). Compare event rates between the two using a two-proportion z-test. |z| ≥ 1.96 corresponds to p ≤ 0.05.
Results for the major GI reactions:
| Reaction | Baseline rate | Recent rate | Relative increase | z-score |
|---|---|---|---|---|
| Constipation | 6.31% | 9.85% | +56% | 16.76 |
| Off-label use | 8.42% | 11.38% | +35% | 12.72 |
| Vomiting | 11.14% | 13.96% | +25% | 10.88 |
| Diarrhoea | 8.80% | 10.85% | +23% | 8.78 |
| Nausea | 16.42% | 18.23% | +11% | 6.05 |
| Decreased appetite | 6.91% | 7.33% | +6% | 2.09 |
Every single GI reaction is emerging — reporting rates are increasing in 2025-2026 compared to 2021-2024. The z-scores are massive (6 to 17), meaning these aren't marginal statistical differences.
This is counterintuitive. Usually a new drug's reaction reporting plateaus or declines as the safety profile becomes known and prescribers adjust dosing. For semaglutide, the opposite is happening: GI reactions are getting reported at higher rates per case in the most recent year than in the prior four years.
Three plausible explanations, all probably contributing:
-
The denominator population is changing. As semaglutide prescribing exploded for weight loss, the patient mix shifted toward people without diabetes, often with less prior tolerance for GI side effects, and often using doses titrated differently than diabetic protocols.
-
Off-label and unsupervised use is increasing. The 35% relative increase in "off-label use" reporting (and the rise of compounded/counterfeit semaglutide flagged earlier) reflects more patients getting the drug outside structured medical supervision — which means more GI events without the dose-titration that prevents them.
-
Reporting threshold is dropping. As semaglutide has remained in the news, patients and clinicians are more likely to report milder events that wouldn't have generated reports for a less-watched drug.
The constipation increase (+56%, z=16.76) is the largest. Constipation is a known semaglutide effect, but the reporting rate nearly doubled in the recent four quarters compared to the prior sixteen. That's worth watching.
The suicidal ideation signal that was investigated and dismissed
This deserves a dedicated section because it illustrates how pharmacovigilance is supposed to work.
The signal emerged. In July 2023, FDA began investigating a potential risk of suicidal ideation and behavior associated with GLP-1 medications after receiving postmarketing reports. The reports raised public concern; media coverage was substantial.
FDA's preliminary review. In a January 11, 2024 Drug Safety Communication, the agency reported that its initial review of GLP-1 RA clinical trial data did not find an association with suicidal thoughts or actions. Because the case counts in individual trials were small, the risk estimate was uncertain — so FDA commissioned a comprehensive review.
The comprehensive review analyzed 91 placebo-controlled GLP-1 medication trials involving 107,910 patients (60,338 on GLP-1s, 47,572 on placebo). FDA also ran a retrospective cohort study using the FDA Sentinel System with more than 2.2 million patients (GLP-1 RA vs. SGLT2 inhibitor comparison). Neither analysis detected an increased risk of suicidal ideation, behavior, or related psychiatric adverse events.
The label change. On January 13, 2026, FDA requested that drug application holders remove information regarding suicidal ideation and behavior from the labeling of GLP-1 receptor agonist medications approved for chronic weight management: Saxenda (liraglutide), Wegovy (semaglutide), and Zepbound (tirzepatide).
The AEMS reports still exist. As of mid-2026, the database contains 389 reports of suicidal ideation linked to semaglutide. They are not evidence the drug causes suicidal ideation; the comprehensive trial data — which is the strongest evidence available — shows no causal link.
The takeaway for reading AEMS data on any drug: report counts are starting points for investigation, not conclusions. A reported reaction following a drug exposure can have many causes — chance, confounding, reporting bias toward newsworthy outcomes. The comprehensive evaluation that establishes or refutes a causal link includes controlled trial data, observational studies, and mechanistic plausibility, not just report counts.
How semaglutide compares to other GLP-1 agonists
Semaglutide isn't alone in its class. Tirzepatide (Mounjaro/Zepbound) and liraglutide (Saxenda/Victoza) are the other major GLP-1 receptor agonists in widespread use. Comparing their adverse-event profiles in AEMS reveals genuine drug-specific patterns:
-
Semaglutide has the highest GI rates per case in its class. Nausea 171/1k vs 108/1k (tirzepatide) vs 151/1k (liraglutide). Vomiting 123 vs 54 vs 113. Constipation 77 vs 40 vs 62. The GI signal is present across all three GLP-1s, but stronger per case for semaglutide.
-
Tirzepatide's adverse event profile is dominated by injection and dosing problems rather than systemic drug effects. Incorrect dose administered at 212 per 1,000 cases — by far the highest single reaction across the three drugs. Injection site pain at 100/1k. This reflects the auto-injector design and patient-handling complexity, not a pharmacologic difference.
-
Liraglutide shows the highest gastroparesis rate per case (86/1k vs 69 for semaglutide vs 15 for tirzepatide). This is non-obvious — most coverage focuses on semaglutide for gastroparesis because of its higher prescription volume, but per case, the older drug has a stronger signal.
-
Off-label use is highest for semaglutide (96/1k) by a wide margin over tirzepatide (63/1k) and liraglutide (54/1k). The weight-loss-prescribing wave shows up most clearly with semaglutide.
We have a separate article going into the three-way comparison in detail. For semaglutide specifically, the takeaway is that the GI profile is the standout feature within the class, and it's measurably more severe per case than either tirzepatide or liraglutide.
How we analyzed this
The methodology behind a useful FAERS/AEMS analysis matters as much as the numbers. Here's the approach.
Data source: FDA AEMS public data files, available quarterly from the AEMS dashboard. Each quarter's release is a set of relational files (demographics, drugs, reactions, outcomes, indications, etc.) keyed by case identifier. The full database currently contains over 8.5 million cases across all drugs and all years.
Drug identification: Mapping reports to "semaglutide" requires handling multiple identifiers — brand names (Ozempic, Wegovy, Rybelsus), generic terms, misspellings, capitalizations, and combination product names. We normalize all variants to a canonical drug. Different studies use slightly different mapping rules, which is one reason exact counts vary between analyses.
De-duplication: A single case can be reported multiple times (initial, follow-ups, multiple sources). Counting each as a separate case inflates totals. We deduplicate by case identifier following the convention FDA uses on its public dashboard.
Disproportionality methods: PRR, ROR (with 95% CI from a normal approximation on the log scale), and Yates-corrected χ². Signal threshold: PRR ≥ 2 AND χ² ≥ 4 AND ≥ 3 cases (EMA convention).
Time-series methods: For detecting emerging signals, we split a chosen time window into "baseline" and "recent" periods and apply a two-proportion z-test to the event rate per case. |z| ≥ 1.96 → p ≤ 0.05.
You can run any of this yourself on our analytics page. Type "semaglutide" (or any drug) into the search and the disproportionality, timeline, and comparison views will run against the latest AEMS data.
Limitations of FAERS/AEMS data
Any analysis based on AEMS reports has structural limitations that affect how the data should be read:
Underreporting. Most adverse events are not reported. Reporting rates vary by severity (serious events more likely reported), drug attention (newsworthy drugs more likely reported), and reporter type (clinicians under-report compared to manufacturers, who are mandatory reporters).
Reporting bias. Reactions that get media attention generate more reports. A new "scary" reaction will see a flood of reports — some genuine, some from patients who experienced the symptom regardless of the drug and now connect it to the drug because they read about it. This is the "stimulated reporting" phenomenon. Semaglutide has had unusual media attention; this affects its data more than most.
No denominator. AEMS knows how many reports exist; it doesn't know how many patients used the drug. Without a denominator, you can't compute true incidence. Disproportionality methods work around this by comparing within the database, but they remain vulnerable to selective reporting.
Quality varies. Reports from manufacturers (mandatory, 15-day timeline, structured workflow) tend to be more complete than reports from patients (voluntary, often incomplete).
Report clusters can distort PRR. A single litigation campaign, registry, or coordinated reporting event can deposit hundreds of cases with similar event lists into the database in a short period. This inflates disproportionality for that drug-event pair without representing a real population-level signal. Looking at case-level patterns (do many reports come from the same source? same date range? same submitter?) is part of careful analysis.
Causality is not established. FDA's standard disclaimer: "Reports sent to the FDA don't necessarily mean the drug caused an adverse event."
For semaglutide specifically, the underreporting is likely substantial. With tens of millions of patients globally and 66,321 reports in the 2021 Q2 – 2026 Q1 window, the reporting rate is well below 1% of estimated prescription-receiving patients. But the patterns in the data — which reactions are over-represented, which signals are emerging, which clusters appear — remain informative when read carefully.
Run this analysis yourself. Every chart, table, and signal in this article was produced on our analytics platform at AdverseEvent.ai. Free, no signup. Type "semaglutide" or any drug name into the search; you'll get disproportionality, time-series, and product-comparison views against the latest AEMS data. We built the tool so clinicians, researchers, journalists, and patients can investigate signals without wrangling raw AEMS files. The analytics page is and will remain free.
What to do with this information
If you're a clinician thinking about prescribing semaglutide or counseling a patient:
- The GI signal is real, pharmacologically expected, and dose-dependent. Slow titration helps; recent emerging-signal data suggests too many patients are skipping it.
- Severe gastroparesis is a contraindication per the October 2025 label update. Screen patients accordingly.
- For patients undergoing elective surgery, hold semaglutide per current anesthesiology guidance and confirm extended fasting (semaglutide delays gastric emptying, raising aspiration risk during anesthesia).
- The suicidal ideation signal has been investigated and dismissed by FDA's comprehensive review.
- NAION risk is real but rare; warn patients about sudden vision changes in one eye.
If you're a patient researching the drug:
- Read the current FDA label, not law-firm marketing pages. The label is updated regularly and reflects FDA's current safety assessment.
- Common side effects (nausea, vomiting, diarrhea, constipation) are usually transient and improve with titration. They're also getting reported at higher rates recently — possibly reflecting more aggressive titration in off-label settings.
- Serious side effects (severe gastroparesis, intestinal obstruction, NAION) are rare but real. Know the symptoms; contact your prescriber if they occur.
- If you're considering compounded or non-pharmacy-supplied semaglutide, the counterfeit signal in AEMS is real and clinically meaningful. Stick to FDA-approved supply.
- Don't stop semaglutide abruptly without discussing with your prescriber.
If you're a researcher or journalist:
- AEMS report counts are inputs to investigation, not conclusions. Filter for case count thresholds before interpreting PRR.
- The most recent FAERS/AEMS data files are available quarterly from the FDA dashboard. Many academic papers cite older snapshots.
- For semaglutide specifically, the gastroparesis lawsuit (MDL 3094) has generated substantial legal-marketing content that ranks well in searches but isn't medical literature. Distinguish between the two.
- The emerging-signal pattern in 2025-2026 (GI reactions rising per case, off-label use rising) is worth deeper study.
Further reading:
- FDA AEMS Public Dashboard — the official data source
- Wegovy (semaglutide) prescribing information — current FDA label
- FDA Drug Safety Communications — official safety updates
- How to fill out FDA Form 3500A — for healthcare professionals who need to report adverse events to FDA
If you're seeing a pattern in your practice or your own care that doesn't appear in this article, you can report it to FDA through MedWatch — the same database we analyzed here. Real-world signals emerge from real-world reports.