Skip to main content

Sources last read

Report-stream question · eight GLP-1 brands

How serious are the reports on GLP-1 drugs?

It varies enormously, and not for the reason you would guess. FDA marked 57% of what was filed about Saxenda serious, against 8.9% for Zepbound. The ordering does not track how risky the drugs are.

Source: FDA side-effect reports, read 2026-08-17

Serious has a fixed meaning here. FDA marks a report serious when it records a death, a stay in hospital, or a life-threatening event. Lasting damage, a birth defect, or anything a doctor judged as grave counts too.

It is a property of the report, not of the drug. One product can look five times as serious as another because different people file about it. This page is mostly about how that happens.

Across FDA reports

Products in the table

8

With a seriousness split

8

Reports counted

401,158

Highest — Saxenda

57%

What share of each product’s FDA reports were marked serious

Each row shows one brand’s FDA reports, split by whether FDA marked them serious. Deaths and hospital stays are shown beside the total because percentages can hide large differences in report volume.

◆ Serious reports — share of each product’s own reportsShare of that product’s reports
8Exact values — 8 rows in the data table
Serious reports — share of each product’s own reports
ProductShare of that product’s reports
Saxendaliraglutide57% of Saxenda reports
Rybelsussemaglutide56% of Rybelsus reports
Ozempicsemaglutide49% of Ozempic reports
Victozaliraglutide47% of Victoza reports
Wegovysemaglutide45% of Wegovy reports
Trulicitydulaglutide26% of Trulicity reports
Mounjarotirzepatide26% of Mounjaro reports
Zepboundtirzepatide8.9% of Zepbound reports

Every row is measured, so there are no ceilings on this chart. Read from FDA on 2026-08-17.

Each product’s total reports, the count FDA marked serious, the count recording a stay in hospital, and the count recording a death.
ProductAll reportsMarked seriousStay in hospitalDeath recorded
Saxendaliraglutide10,040 reports57% of Saxenda reports2,456 reports290 reports
Rybelsussemaglutide7,379 reports56% of Rybelsus reports1,690 reports174 reports
Ozempicsemaglutide66,161 reports49% of Ozempic reports12,871 reports1,114 reports
Victozaliraglutide44,086 reports47% of Victoza reports9,836 reports1,694 reports
Wegovysemaglutide21,044 reports45% of Wegovy reports2,645 reports123 reports
Trulicitydulaglutide88,332 reports26% of Trulicity reports10,807 reports1,696 reports
Mounjarotirzepatide93,975 reports26% of Mounjaro reports8,667 reports772 reports
Zepboundtirzepatide70,141 reports8.9% of Zepbound reports2,430 reports176 reports

A death column is the most misread thing on this site. It records that a person died and that the drug was named in the report. It is not a count of people the drug killed, and FDA states plainly that a report does not establish cause.

Every share above is a share of that product’s own reports. It is never a share of the people who took it, because nobody knows how many that is. Two products can differ here for reasons that have nothing to do with either drug.

The same drug, twice, at two very different numbers

Mounjaro and Zepbound are the same drug substance from the same company. Mounjaro’s serious share is more than twice Zepbound’s, which settles what this measure is really about.

One molecule cannot be two different levels of dangerous. So whatever moves this number between those two rows is not the drug. Something that can move a figure that far between twins can move it anywhere in the table.

The full spread is more than fivefold across the eight brands. That is far wider than anything the differences between these drugs could produce. It is the clearest sign on this site that a raw report count is not a safety measure.

Who files decides what the share looks like

Drug companies must send FDA the serious reports fast and the rest on a slower schedule. People at home file about whatever bothered them, and most of that is not serious.

So a product with a big consumer following collects a flood of non-serious reports, and its serious share falls. A product mainly reported through doctors and drug companies collects proportionally more serious ones, and its share rises. Neither movement says anything about the drug.

Age of the product does the same thing from another direction. A drug sold for a decade has had time to collect hospital admissions and deaths. Its users include older and sicker people than a newer drug’s.

The death column needs the most care of anything on this site. A death recorded in a report means the person died, not that the drug killed them. FDA states plainly that a report does not establish cause.

What the labels say about it

Labels carry warnings and frequencies from trials, where the denominator is known. That is the document to read for how often something happened, and this one is not.

Trials and FDA reports answer different questions. A trial follows selected people for a fixed period; voluntary reports show what was submitted after wider use, with all the reporting bias that implies.

This section summarizes reports people filed. The label guides show what the FDA-approved prescribing information lists. The two answer different questions and they often disagree, which is the most useful thing about having both.

What this page does and does not establish

This is the page most easily misread on the site, so the boundary is spelled out row by row.

Each claim a reader might take from this page, and what the sources behind it actually support.
The claimWhat the evidence shows
The serious share varies hugely between GLP-1 brandsEstablished. The top row is more than five times the bottom one.
The measure is not tracking the drugEstablished by the twins. Two products built on one drug substance sit more than twice as far apart as each other.
The brands at the top are more dangerousNot established. The ordering tracks who files and how long the product has been sold. There is no denominator to turn either share into a risk.
A death count is a count of people killed by the drugNot established, and FDA says so itself. A report records that a death occurred while the drug was named.
You can compare these numbers to a trial’sNot established. A trial has a denominator and this does not, so the two are different kinds of object.

Common questions

Can these drugs kill you?

Every product here carries reports where a death was recorded, and every one of those is a real person. None of them is a finding that the drug caused the death.

The labels carry the warnings that come from trials, where cause can be studied. That is the document worth reading with a clinician, and it is linked at the foot of this page.

Why does the newest brand look safest?

Because it has had the least time to collect serious reports. A lot of what gets filed about it comes from people at home rather than hospitals.

A low share only describes the submitted FDA reports. It is not proof that a product is safer.

Should this change what I do?

No page can answer that, and this one has no denominator, so it is especially unqualified to. What it is good for is reading news stories that quote a raw count.

Where to go next

This page is one question asked of one snapshot. The pages below answer the neighboring ones from the labels and the trials, which is a different kind of evidence.

Every question in this section is listed on the report-stream questions index.

Read this before quoting anything above

These are counts of reports filed with the FDA, not rates. Nobody knows how many people took the drug and did not file a report, and a report is not a finding that the drug caused what was reported.

  • There is no denominator. FAERS counts reports that were filed. It does not count the people who took the drug, so there is no bottom half to the fraction. Any figure on this page that looks like a percentage is a share of this drug’s reports — never a share of the people who took it.
  • A report is not proof the drug did it. FDA does not require a reporter to establish cause. A report says somebody noticed something and thought the drug was worth naming. Some of these events were caused by the drug, some were caused by the illness it was prescribed for, and some were coincidence — FAERS does not separate them.
  • A bigger number can just mean a more-prescribed drug. More reports can simply mean that more people use the drug, that it is newer, or that it receives more news and legal attention. Raw report totals do not tell you which drug is safer.
  • The same event can be counted twice. One event can reach FAERS through more than one channel — the patient, the prescriber and the manufacturer can all file. FDA does not guarantee that the public files are de-duplicated.
  • FDA’s own words. Do not rely on openFDA to make decisions regarding medical care. While we make every effort to ensure that data is accurate, you should assume all results are unvalidated. We may limit or otherwise restrict your access to the API in line with our Terms of Service.
  • Retrieved 2026-08-17 from https://api.fda.gov/drug/event.json · FDA data last updated 2026-07-30 · license https://open.fda.gov/license/

How everything on this site is sourced →

10primary sources, each with the day we read it — open the documents used on this page