Peptide basics

What people actually report to the FDA about semaglutide

This page reports what one public query returns and what that return can and cannot support. It does not say semaglutide is safe, does not say it is dangerous, and does not tell you what to do about any symptom — that requires someone who can examine you.

The query, stated in full

FDA's Adverse Event Reporting System is exposed through a free, keyless API. The single request behind this page asks for every report naming SEMAGLUTIDE as a product and counts the reported reaction terms: search=patient.drug.medicinalproduct:"SEMAGLUTIDE" with count=patient.reaction.reactionmeddrapt.exact. The extract we read was dated 28 April 2026 and returned 6,618 reports in total under that name.¹³

The top of that ranked list, in the same extract, was NAUSEA 943, VOMITING 835, DIARRHOEA 589, OFF LABEL USE 426 and FATIGUE 396 — and every one of those numbers is a count of reports mentioning the term, not a count of people, not a count of confirmed events, and not a share of anything.¹

• NAUSEA 943 reports mentioning the term — not 943 verified cases.
• VOMITING 835 reports mentioning the term.
• DIARRHOEA 589 reports mentioning the term.
• OFF LABEL USE 426 reports mentioning the term, which describes how a product was used rather than what happened to anyone.
• FATIGUE 396 reports mentioning the term.¹

One report can carry several reaction terms, so these counts overlap and do not sum to the number of reports. That is a property of the data structure, not an error in the query.³

Why this is not a rate

A rate needs two numbers: how many events, and how many people were exposed. FAERS supplies neither reliably. It holds reports that someone chose to submit, and it holds no record at all of how many people took the drug and had nothing to report. FDA states the consequence plainly on the dashboard's own limitations page: the information in these reports cannot be used to estimate the incidence of the events reported.²

The same page lists three further limitations before that one: duplicate and incomplete reports are in the system, the existence of a report does not establish causation, and the information in reports has not been medically verified. FDA also states that the data by themselves are not an indicator of a product's safety profile. Those are the publisher's words about its own dataset, not an outside critique of it.²

This is why 943 nausea reports for semaglutide cannot be compared with, say, 220 for a different product and read as a fourfold difference in nausea. The two products have different numbers of users, different indications, different reporting histories and different manufacturers' reporting obligations. The counts are not measured on a shared scale.²⁹

Why causality is not established

A report records that someone experienced something while taking a product. It does not record that the product caused it, and no reviewer confirms causation before the report is stored and published. FDA's own limitations note says that for any given report there is no certainty a suspected drug caused the event, and that the event may relate to the underlying disease, to another drug taken at the same time, or to something else entirely.²

The reports also arrive from mixed sources. In the same extract, reports naming semaglutide were filed by other health professionals (1,974), consumers or non-health professionals (1,922), physicians (1,624), pharmacists (1,034) and lawyers (9) — categories that differ in how they establish what happened, and none of which constitutes an adjudicated diagnosis.¹

In the same extract, semaglutide appeared as a suspect product in 6,519 drug entries and as a concomitant product in 3,792 — meaning a large block of these reports concern people who were taking something else as well, with the reaction attributed elsewhere or nowhere.¹

Why the counts move for reasons that are not risk

Under-reporting is the normal state of spontaneous surveillance. Hazell and Shakir pooled 37 studies from 12 countries and calculated a median under-reporting rate of 94% (interquartile range 82–98%) — that is, most adverse reactions detected by intensive data collection never reached the spontaneous reporting scheme at all. For studies of specific serious drug–event pairs the median was still 85%.⁵

That gap is not a fixed multiplier that can be applied to a count to recover the true number. Its size varies by setting, by seriousness and by drug, and it is unknown for any particular product, which is precisely why a count cannot be scaled up into an estimate.⁵

Reporting also has a shape over time. Hoffman and colleagues examined 62 drugs approved between 2006 and 2010 and found that the classic pattern of reports peaking in the first years after launch does not hold consistently in modern FAERS data: the common shape was a rise over the first three quarters and then a broadly constant count. Either way, the curve is a curve of reporting behaviour.⁶

And the name you query changes the answer more than anything about the drug does. In the same extract, the generic term SEMAGLUTIDE returned 6,618 reports while OZEMPIC alone returned 57,104 — the same molecule, filed under a different product string. A count is a count of a name, not of a substance.¹⁷

What the counts are actually good for

FAERS is built as a signal generator, not a measuring instrument. Its designed purpose is to surface events worth investigating with a study that has a denominator — a rare event nobody expected, a cluster that arrived faster than prescribing did. FDA describes the system in exactly those terms.²

It is also a record of what people chose to report, which is a legitimate object of interest in its own right. That OFF LABEL USE ranks fourth in the semaglutide extract says something about how this product entered the population; it says nothing about anyone's physiology.¹

Where a rate does exist

The approved labelling is the place where denominators are published. The prescribing information for semaglutide in weight management tabulates adverse reactions against a placebo column with the number of participants in each — a structure FAERS does not have and cannot be made to have.⁸

Semaglutide has an approved US product, a large registered trial record and a labelled adverse-reaction table. Where a question can be answered from any of those, the FAERS extract is the weakest available source and should not be the one quoted.

Common questions