What 1,700 adverse event reports do and don't tell you

TL;DR

• The FDA Adverse Event Reporting System (FAERS) collects suspected side effects after a drug is on the market. Anyone can submit a report; manufacturers are required to for their products.
• A report records a suspicion, not a finding. It cannot establish that the product caused the event, it has no count of how many people were exposed, it is not verified, and most events are never reported. A report neither proves nor rules out harm: the question is unestablished until someone studies it properly.
• It is still the main early-warning system for drug safety, and several important safety problems were first seen this way. Its job is to raise questions, not answer them.
• Compounded medicines have a bigger blind spot. State-licensed 503A pharmacies are not federally required to report adverse events, so totals for compounded products are a floor.
• As of 31 May 2026 the FDA had over 1,700 reports linked to compounded semaglutide and tirzepatide. Read correctly, that number justifies attention and monitoring. It does not, on its own, measure a risk.

Educational content, not medical advice. This is a method explainer. It makes no safety claim about any product in either direction, and no doses or protocols appear.

Updated 26 October 2026: added the October compounded tirzepatide recall and what a "Class II" label means, since a recall and an adverse event count are often read as one signal when they are two different things.

Updated October 2026: added KFF Health News reporting (13 October 2026) on individual peptide harm accounts, and why passive surveillance under-counts most when access is expanding.

---

A number like "over 1,700 adverse event reports" arrives with an argument already attached. One side quotes it as proof that compounded GLP-1 medicines are dangerous. The other side waves it away as meaningless noise. Neither reading is right, and the reason is not in the number. It is in the kind of system that produced it.

This page explains that system. It sits alongside two other evidence-literacy guides on this site, and each covers a different kind of evidence with different failure modes: what a study actually measured covers clinical trials, which are designed in advance to answer a question; this page covers post-market surveillance, which collects what happens after a drug is in use and was never designed to answer causal questions on its own.

What FAERS is, and what a report has to establish

FAERS is the database the FDA uses to support its safety monitoring of drugs and therapeutic biologics after approval. Reports come in from three main directions:

• Manufacturers, who are required by regulation to report serious and unexpected adverse events they learn about for their own products.
• Health professionals, who report voluntarily, usually through the FDA's MedWatch programme.
• Patients and caregivers, who can also report directly.

What a report has to establish is the most important thing to know about it, and the answer is: essentially nothing beyond the fact that someone suspected a connection. The FDA does not require that a causal relationship be shown before a report is accepted, and reports often lack the detail needed to evaluate one. A report says that an event happened to a person who had been using a product, and that the reporter thought the two might be linked. That is the entire claim.

The four things a report cannot tell you

1. Causation. The event may have been caused by the product, by another medicine, by the underlying illness, or by chance. A report cannot separate these. Many reported events are also common in the general population, so some will occur in people using any product.

2. A denominator. FAERS counts reports, not people exposed. Without knowing how many people used a product, over what period, a count cannot be turned into a rate. "1,700 reports" out of ten thousand users and out of ten million users would be very different findings, and the database does not say which.

3. Verification. Reports are not checked against medical records before they enter the system. Some are detailed clinical write-ups; some are a single line. Duplicates occur, and the same event can be reported by more than one person.

4. Completeness. Most adverse events are never reported. The fraction that is reported is unknown, varies by drug and severity, and changes over time. Reporting rises after media coverage, after a safety warning, or around litigation, a pattern sometimes called stimulated or notoriety reporting. A jump in reports can reflect attention rather than a change in risk.

The four things it can tell you

It would be a mistake to stop there. Passive surveillance catches what trials cannot: rare events, events in people who would never have qualified for a trial, and events that take longer to appear than a trial runs.

1. Signals. An unexpected event that keeps appearing with one product is a reason to look harder. That is the system's core purpose.

2. Clusters. Several reports with the same unusual feature, such as a particular pattern of harm, a particular presentation of the product, or a particular setting, can point to a specific cause worth investigating.

3. Unexpected events. Events not listed in a product's labelling are the ones the system is best placed to surface, because they are the ones nobody was looking for.

4. Change over time. A shift in the pattern of reports, interpreted cautiously in light of the attention effects above, can prompt a closer look.

In each case the output is a question. The answer comes from somewhere else: a review of the individual case reports by agency safety staff, a study using linked health records, or an active surveillance system that does have a denominator, such as the FDA's Sentinel system.

Why compounded products are harder to read

For an approved medicine there is a manufacturer with a legal duty to collect and report serious adverse events. For compounded medicines the obligations differ by type of compounder:

• 503B outsourcing facilities, which compound at scale, are required to report serious, unexpected adverse events to the FDA.
• 503A pharmacies, the state-licensed pharmacies that compound for individual prescriptions, are not required by federal law to report adverse events to the FDA. Some states impose their own rules.

That gap matters for how a total is read. When a large share of the relevant dispensers has no reporting duty, the total is more likely to be too low than too high. It also means the product behind a report is less well characterised: compounded preparations vary between pharmacies and are not FDA-approved, so a cluster of reports may reflect a specific preparation or presentation rather than the molecule. The practical counterweight is monitoring at the level of the individual: Vivre's semaglutide and tirzepatide are offered with baseline and follow-up bloodwork recommended, so a problem is looked for in you rather than waiting to surface in a national database. How compounding is regulated is set out in what 503A compounding is, and the legal position for GLP-1s specifically in is compounded GLP-1 legal.

Disproportionality analysis, and why it is not a measure of risk

Researchers often analyse FAERS using disproportionality methods. The idea is to ask whether a particular event is reported more often, relative to all reports for a drug, than it is for other drugs in the database. The results are expressed as ratios, such as a reporting odds ratio.

These analyses are useful for finding signals. They are not measures of risk, and published papers based on them are frequently over-read. A high reporting ratio means an event makes up an unusually large share of the reports for a drug; it does not mean the event is common in people taking it, or that the drug causes it. The same attention effects, duplicate reports and missing denominators apply. A disproportionality result is a stronger question, not an answer.

How to read the number you saw quoted

The FDA's own statement about compounded GLP-1s is the right model. As of 31 May 2026 it had received 990 adverse event reports associated with compounded semaglutide and more than 730 associated with compounded tirzepatide, over 1,700 in total, and it noted that reports do not establish causation and that the totals are likely to be undercounts, in part because state-licensed pharmacies are largely not required to report. Many of the reports involved dosing errors, including large errors made when measuring from multi-dose vials, some requiring hospital care.

Four questions turn that paragraph into something you can use:

1. What system produced it? Passive and voluntary, with a large exempt group of dispensers. So: a floor, not a ceiling.
2. Is there a denominator? No. So it cannot be quoted as a rate or compared with another product's rate.
3. What else was happening? 2026 has seen heavy coverage and enforcement around compounded GLP-1s, which is likely to have raised reporting.
4. Is there a specific, consistent pattern? Yes: the measurement-error cluster. That part does not depend on resolving causation for every report, because the mechanism is visible in the reports themselves.

Put together: the number does not show that compounded GLP-1s are dangerous in general, and it does not show that they are safe. It shows a real signal, partly about how products are presented and supplied, that warrants attention, and it shows how little visibility there is into this part of the market.

A recall is a different kind of signal (October 2026)

In October 2026 a compounded tirzepatide recall made headlines at the same time as the 1,700 figure was being quoted again. The two are easy to blur together. They should not be.

The facts: Apothecary Pharma, a North Carolina company registered with the FDA as a 503B outsourcing facility (a compounder that makes larger batches under federal oversight), voluntarily recalled 22 lots of compounded tirzepatide injection and one lot of NAD+ injection because it could not assure the products were sterile. The recall began on 24 September 2026, and on 8 October 2026 the FDA classified it as Class II. FDA's enforcement report separately lists a Class II recall of compounded semaglutide syringes from Carolina Infusion, LLC, also over sterility assurance. The FDA notice did not report any illnesses or injuries linked to the recalled products.

What "Class II" means, in plain words: the FDA judges that using the product may cause temporary or medically reversible health problems, and that serious harm is unlikely. Class I is the most serious grade (a reasonable chance of serious harm or death); Class III is the least (unlikely to cause any harm).

Why the two signals are different:

• An adverse event count is a pile of suspicions from people who used products, with no denominator and no proof of cause.
• A recall is a decision about specific batches from a specific facility, based on how they were made. "Lack of sterility assurance" is a finding about a process, not a test result showing that any particular vial was contaminated.

So a recall does not add to the 1,700, and the 1,700 does not explain the recall. Each answers a different question. If a medicine you use is on a recall list, that is a conversation with your own prescriber, today.

The same blind spot, now for peptides (October 2026)

On 13 October 2026 KFF Health News (reporter Amy Maxmen; the story also ran in The Washington Post) published individual accounts of people who say experimental peptides harmed them, including a man who reported months of anxiety and twitching after using BPC-157 bought online, and cited two FAERS reports describing hospitalisation. The piece gives no aggregate FAERS count for peptides, and none should be inferred from it. The 1,700+ figure above is about compounded semaglutide and tirzepatide only.

The structural point is the one this page has been making. FAERS is passive: reports are voluntary, brief, unverified, and cannot prove causation, and many harms are never reported at all. For products that move largely outside the approved-drug system, such as research-use-only peptides, there is often no manufacturer with a reporting duty, so the count is likely to be even further below reality. That gap matters most when access is expanding: on 14 October 2026 HHS/FDA sent an interim peptide compounding policy to the White House Office of Management and Budget for review, which has not been issued and whose text is not public. A system that under-counts harms is weakest exactly when more people may be starting to use a product. Absence of reports is not evidence of safety, and individual stories are not evidence of a rate: both leave the question unestablished until it is studied properly.

Frequently asked questions

Does an adverse event report prove a drug caused the side effect? No. A report records that an event happened to someone using a product and that the reporter suspected a link. It does not establish causation, and the FDA does not require causation to be shown before accepting a report. Establishing whether a product causes an event takes further investigation, such as case review or a study with a comparison group.

Why can't FAERS numbers be turned into a risk percentage? Because FAERS has no denominator. It counts reports, not the number of people who used the product. It also misses most events and can be inflated by publicity. A rate needs both a reliable count of events and a reliable count of people exposed, and FAERS has neither.

Are reports for compounded drugs less reliable than for approved drugs? They are harder to interpret. 503A pharmacies are not federally required to report adverse events, so totals are likely to undercount, and compounded preparations vary more between sources than an approved product does. That makes the numbers more likely to understate what is happening and less able to pin it to a specific product.

If the data is so limited, why does the FDA use it? Because it catches things trials cannot: rare events, events in people excluded from trials, and events that appear after years of use. It is an early-warning system. When it raises a signal, the FDA investigates with other tools, including active surveillance systems that do know how many people were exposed.

---

Related reading

• What a study actually measured: reading the trial side of the evidence
• Peptide side effects: what is known, and how well
• What 565 studies actually found: what the published literature does and does not establish
• Is compounded GLP-1 legal?: where compounded GLP-1s stand in 2026
• What 503A compounding is: how compounding is regulated, and who has to report what

Part of The Complete Guide to Peptides

Monitoring is the part a reporting system cannot do for you. A surveillance database finds problems across a population after the fact; baseline and follow-up bloodwork, read with your own doctor, look for them in you, as they happen. Vivre Labs offers semaglutide and tirzepatide, including within a metabolic protocol, and we recommend starting with baseline bloodwork.

Educational content, not medical advice. This page explains how post-market adverse event reporting works and how to interpret figures drawn from it. It makes no claim that any product is safe or unsafe, and it does not assess any seller, pharmacy or product. If you think you have had a side effect from a medicine, speak to your prescriber, and you can report it to your national regulator. Claims for the Malaysian market are subject to MAB requirements. Last reviewed: October 2026.

References

• US Food and Drug Administration. Questions and Answers on FDA's Adverse Event Reporting System (FAERS). FAERS supports post-marketing safety surveillance; reports do not establish causation; FDA does not require that a causal relationship be proven; reports may lack detail; the system cannot be used to calculate incidence. (fda.gov)
• Potter et al. FDA Adverse Event Reporting System (FAERS) Essentials: A Guide to Understanding, Applying, and Interpreting Adverse Event Data Reported to FAERS. Clinical Pharmacology & Therapeutics. 2025. Causality, denominator, reporting bias, duplication and appropriate use of disproportionality methods. (ascpt.onlinelibrary.wiley.com/doi/10.1002/cpt.3701)
• 21 CFR 310.305 and 314.80: manufacturer adverse drug experience reporting requirements for marketed prescription drugs.
• FD&C Act section 503B(b)(5) and FDA guidance, Adverse Event Reporting for Outsourcing Facilities Under Section 503B: outsourcing facilities must report serious, unexpected adverse events. (fda.gov/media/90997/download)
• National Academies of Sciences, Engineering, and Medicine. Compounded Topical Pain Creams: Gaps in Regulation, Oversight, and Surveillance. 503A compounding pharmacies are not required to collect or share adverse event data with FDA. (ncbi.nlm.nih.gov/books/NBK560345)
• US Food and Drug Administration, statement on compounded GLP-1 adverse events: as of 31 May 2026, 990 reports associated with compounded semaglutide and more than 730 with compounded tirzepatide; reports do not establish causation and totals are likely undercounts, in part because state-licensed pharmacies are largely not required to report; many reports involve dosing errors with multi-dose vials, some requiring hospitalisation. (fda.gov; figures as carried across this library)
• Maxmen A. People Say These Peptides Hurt Them. Their Stories Are Muffled in MAHA's Push for Medical Freedom. KFF Health News, 13 October 2026 (also published by The Washington Post). Individual harm accounts and two cited FAERS hospitalisation reports; no aggregate peptide count; FAERS reports are voluntary, unverified and cannot prove causation. (kffhealthnews.org/public-health/peptides-experimental-drugs-patient-harms-maha-medical-freedom/)
• Holland & Knight. Peptides at a Crossroads: What Wellness Providers Need to Know. October 2026. HHS/FDA submitted an interim peptide compounding access policy to OIRA/OMB for review on 14 October 2026; not issued, text not public. (hklaw.com)
• Becker's Hospital Review. Compounded GLP-1 injections recalled nationwide: 4 notes. 20 October 2026. Apothecary Pharma voluntary recall of compounded tirzepatide and one NAD+ lot; Class II; no illnesses reported. (beckershospitalreview.com)
• BioPharma Dive. 22 October 2026. Apothecary compounded tirzepatide recall, 22 lots, Class II. (biopharmadive.com)
• US Food and Drug Administration. Warning letter 717972 to Apothecary Pharma LLC, 1 December 2025: registered 503B outsourcing facility; insanitary conditions in sterile production. (fda.gov)
• Peptide Briefs. A sterility recall, a regulatory signal. 25 October 2026. FDA enforcement report entry: Class II recall of compounded semaglutide syringes, Carolina Infusion, LLC, lack of sterility assurance. (peptidebriefs.substack.com)
• US Food and Drug Administration. Recalls background and definitions: Class I, II and III. (fda.gov)
• US Food and Drug Administration. Sentinel Initiative: active surveillance using electronic health data with known populations, used alongside FAERS to evaluate signals. (fda.gov)