Acute pancreatitis is the side effect that made this drug class frightening before most people had heard of it. It carries a boxed-warning reputation it does not actually hold, it is the reason prescribers ask about gallstones and alcohol, and it is the thing a reader most wants a number for.
The labels warn about it in every case. Six of them, read on 14 September 2026 through openFDA: Zepbound (effective 2026-08-28), Mounjaro (2026-07-29), Wegovy (2024-04-23), Ozempic (2026-07-30), the shared Rybelsus and Ozempic tablets label (2026-01-30) and Saxenda (2026-02-25). Every one carries acute pancreatitis in Warnings and Precautions, every one names the fatal haemorrhagic and necrotising forms, and every one tells the prescriber to stop the drug if it is suspected.
Not one of those warnings contains a number.
The warning and the count live in different sections
This is worth stating precisely, because it is the shape of the whole page. In all six labels the Warnings and Precautions entry for acute pancreatitis says the event "has been observed" and then describes what to watch for. The incidence, where there is any, sits further down in Adverse Reactions — a section a reader reaches only by scrolling past the tables of nausea and diarrhoea.
Compare that with the gallbladder. The same labels quantify cholelithiasis to a tenth of a percentage point, drug against placebo, in the warning itself — we set those figures out in full on the gallbladder page, where they turned out to run backwards to the weight loss. Two adverse events, the same documents, one counted where the reader will find it and one not.
The same molecule, two labels, two different answers
Tirzepatide is sold as Zepbound for weight and as Mounjaro for type 2 diabetes. Both labels report adjudicated pancreatitis. They do not report the same thing.
| Label | Population | Drug | Comparator |
|---|---|---|---|
| Zepbound | weight reduction (Studies 1 and 2) | 0.2%, 0.14 per 100 patient-years | 0.2%, 0.15 per 100 patient-years |
| Zepbound | obstructive sleep apnea (Studies 5 and 6) | 0.84 per 100 patient-years | 0 |
| Mounjaro | type 2 diabetes | 0.23 per 100 patient-years (13 patients, 14 events) | 0.11 per 100 patient-years (3 patients, 3 events) |
In the weight-reduction pool the placebo rate is fractionally higher than the drug rate. In the diabetes studies the drug rate is roughly double the comparator. In the sleep apnea studies the drug rate is six times what the same drug produced in the weight trials, against a placebo arm that recorded none at all.
The most likely explanation is the one the numbers themselves suggest: these are three different populations, and type 2 diabetes independently raises the background rate of acute pancreatitis. That is not a criticism of either label. It is a warning about how these figures get quoted. A page that reports "tirzepatide doubles pancreatitis risk" and a page that reports "tirzepatide shows no pancreatitis excess over placebo" can both be citing a current FDA label, accurately, and neither is telling the reader which population produced the number.
Semaglutide's labels are thinner on this point but point the same way. Wegovy, in obesity, records four adjudicated cases against one on placebo — 0.2 against under 0.1 cases per 100 patient-years. The shared Rybelsus and Ozempic tablets label, in diabetes, records pancreatitis as a serious adverse event in six patients against one on comparator, at 0.1 events per 100 patient-years against under 0.1.
The largest number in these labels came from the one trial that did not adjudicate
Saxenda's adult figure is 9 of 3,291 treated patients (0.3%) against 2 of 1,843 on placebo (0.1%) — the clearest excess anywhere in the set, and still nine people.
Its paediatric figure is 0.8%.
That is the biggest pancreatitis percentage printed in any of these documents, and the label gives the reason in the same sentence it gives the number: "In a SAXENDA pediatric clinical trial, pancreatitis was not independently adjudicated." One patient, in one trial, with no adjudication committee deciding whether the event met the definition. Every other figure on this page survived a blinded committee; that one did not.
Adjudication is not a formality here. Acute pancreatitis is diagnosed from a combination of pain, enzyme levels and imaging, and in a population being given a drug that causes abdominal pain and vomiting in a third of patients, the raw reported-event count and the confirmed-event count are not the same quantity. A number that has passed adjudication and a number that has not should never appear in the same column, and on this page they do not.
What the reporting database adds, and what it cannot
FDA's adverse event system holds every report anyone chose to file. Read on 14 September 2026, acute pancreatitis appears in:
| Molecule | Reports naming acute pancreatitis | All reports | Share |
|---|---|---|---|
| liraglutide | 747 | 50,704 | 1.47% |
| exenatide | 745 | 52,313 | 1.42% |
| semaglutide | 283 | 73,001 | 0.39% |
| dulaglutide | 323 | 88,160 | 0.37% |
| tirzepatide | 324 | 160,663 | 0.20% |
The two molecules at the top are the two oldest in the class, and they are the two that were on the US market through FDA's pancreatitis investigations of 2013. Clinicians were being asked, in print, to consider exactly this diagnosis in exactly those patients. A reporting share that tracks the publicity as closely as this one does is not measuring incidence, and cannot be converted into a risk for any individual — the database has no denominator, no verification and no control group. It is included here because its shape is informative in the opposite direction from the way it is usually quoted: the drugs everyone is currently taking sit at the bottom of the table, not the top.
What none of this settles
The honest summary is that the trials were not built to answer this question. Acute pancreatitis in the general population runs at roughly the same order of magnitude as the rates in these tables, the trials excluded people with a history of it, and the confirmed-case counts run from four to thirteen patients per programme. Wegovy's label says as much outright: there is limited trial experience in patients with a history of pancreatitis, and it is unknown whether they are at higher risk.
Two things are nonetheless firm, and they are the two worth carrying away. The warning is real enough that every manufacturer in the class writes the same stop instruction, and the class's own documents record it as a reason to discontinue rather than to monitor — the same posture the labels take on surgery and anaesthesia, where they also decline to convert a warning into a protocol. And the numbers that do exist are small, denominated inconsistently, and dependent on which population was studied. Anyone quoting one of them should say which label it came from.
Sources
Every figure above was read on 14 September 2026 from FDA prescribing information published through openFDA — Zepbound (effective 2026-08-28), Mounjaro (2026-07-29), Wegovy (2024-04-23), Ozempic injection (2026-07-30), Rybelsus and Ozempic tablets (2026-01-30), Saxenda (2026-02-25), Byetta (2025-09-02) — and from FDA's adverse event reporting system via the same interface on the same day.