Evidence review

Does semaglutide cause cancer?

By David Chen·Editor·Reviewed September 2026

What the warning actually says¹

Both semaglutide labels open with the same boxed warning, and it is worth reading for what it does not claim. It reports an animal finding, then states the human question is open:

In rodents, semaglutide causes dose-dependent and treatment-duration-dependent thyroid C-cell tumors at clinically relevant exposures. It is unknown whether WEGOVY® causes thyroid C-cell tumors, including medullary thyroid carcinoma (MTC), in humans as human relevance of semaglutide-induced rodent thyroid C-cell tumors has not been determined.

The second half of the warning is a rule rather than a finding: semaglutide is contraindicated in anyone with a personal or family history of MTC, or with Multiple Endocrine Neoplasia syndrome type 2. The Ozempic label carries the same two paragraphs with its own brand name.¹¹ The full text, verbatim, is on the boxed warning page.

What happened in the rodents¹

Section 13.1 is where the numbers are. Two lifetime studies, one in mice and one in rats:

• Mice, two years. Doses up to 7-fold the maximum recommended human dose in males and 2-fold in females, measured by drug exposure rather than milligrams. The label reports “a statistically significant increase in thyroid C-cell adenomas and a numerical increase in C-cell carcinomas… in males and females at all dose levels.”
• Rats, two years. Doses up to 0.7-fold human exposure — below the human dose. A significant increase in C-cell adenomas at every dose level in both sexes, and in carcinomas in males at the higher doses.
• Then the sentence that matters. “Human relevance of thyroid C-cell tumors in rats is unknown and could not be determined by clinical studies or nonclinical studies.”

The label also records that semaglutide “was not mutagenic or clastogenic in a standard battery of genotoxicity tests” — it does not damage DNA in the usual assays. Whatever happened in those rodent thyroids, it was not that.

What the label says about people¹

Section 5.1 adds three human-facing points, and none of them is a rate:

• A different drug’s reports. “Cases of MTC in patients treated with liraglutide, another GLP-1 receptor agonist, have been reported in the postmarketing period; the data in these reports are insufficient to establish or exclude a causal relationship between MTC and GLP-1 receptor agonist use in humans.” That is liraglutide, not semaglutide, and the label draws no conclusion from it.
• Screening does not help. Routine monitoring of serum calcitonin or thyroid ultrasound “is of uncertain value for early detection of MTC,” and “may increase the risk of unnecessary procedures” because the test is not specific enough and thyroid disease is common anyway.
• What to report. The label asks prescribers to inform patients of the symptoms of thyroid tumors: a mass in the neck, difficulty swallowing, shortness of breath, persistent hoarseness.

Neither label’s postmarketing section — the list of things reported after approval — contains a thyroid or cancer entry. What it does contain is pancreatitis, ileus, obstruction, severe constipation, hypersensitivity, aspiration and acute kidney injury.

What the semaglutide trials recorded²

SELECT is the longest randomized record of semaglutide 2.4 mg: 17,604 adults, mean follow-up 39.8 months. Malignant neoplasms were a prespecified adverse event of special interest, which means they were collected deliberately rather than incidentally.

The authors’ own summary sentence is narrower than a headline but clear: “serious adverse events related to gastrointestinal disease, acute kidney failure, pancreatitis, cancers, or psychiatric disorders were not more frequent with semaglutide than with placebo.” One caveat travels with every SELECT figure: the trial collected only serious adverse events, events that led to stopping, and events of special interest — not everything that happened. The rest of what it recorded is on long-term safety.

STEP 1, the pivotal obesity trial, is shorter — 1,961 adults, 68 weeks — and reports the same shape of result: malignant neoplasms in 14 of 1,306 on semaglutide (1.1%) and 7 of 655 on placebo (1.1%), with the authors writing that “there was no difference between groups in the incidence of benign and malignant neoplasms.”³ Sixty-eight weeks is far too short for a tumor to begin and be found, which is why neither trial answers the label’s question. The trial itself is on the STEP 1 page.

Thyroid cancer in the trials⁴

Thyroid cancer is rare enough that no single trial can measure it, so the manufacturer pooled its own: 93 trials of liraglutide or semaglutide, 101,732 participants, 207,000 patient-years. Across all of them, thyroid cancer was reported in 0.08% of people on a GLP-1 drug and 0.06% on placebo — a hazard ratio of 1.70, with a confidence interval of 0.99 to 3.03 that just touches no-difference. Medullary thyroid cancer, the type the boxed warning names, was reported in three people on the drugs and four on placebo.

Its table of the six long cardiovascular and kidney outcome trials is the closest thing to a semaglutide-specific count, so here are the semaglutide ones, participants with at least one malignant thyroid neoplasm:

Single digits either way, in trials of thousands. Two things follow. Counts this small cannot distinguish a real effect from chance — and they also cannot rule one out. The analysis was funded by Novo Nordisk, which makes both drugs, and the trials in it were too.

The observational studies⁶

This is where the public argument lives, and the first column is the one to read. Most of these studies are about the GLP-1 class, and in the biggest of them semaglutide is a minority of the drug exposure.

How to read that table

• The one study that is only about semaglutide found nothing. Its confidence interval runs from 0.33 to 12.61, which is another way of saying the trials it pooled contained almost no thyroid cancers.⁵
• The study that found a class increase was mostly not semaglutide. The French analysis covers prescriptions from 2006 to 2018⁷, and the semaglutide-only review describes the drugs behind that signal as exenatide, liraglutide and dulaglutide.⁵
• The two neutral cohorts are short. Mean follow-up on semaglutide in the Scandinavian study was 1.1 years; the multisite study’s median was under three⁸. Both sets of authors say so, and the second adds that the evidence was “insufficient to rule out excess risk with long-term use.”
• A meta-analysis of trials found a fragile class signal. Odds ratio 1.52, and a fragility index of 1: a single reclassified event would erase it.⁹
• Other cancers point the other way, and prove nothing either. A 2026 analysis of 161,798 matched adults with obesity and no diabetes reported fewer obesity-associated cancers among users of GLP-1 drugs as a class, semaglutide and tirzepatide included, hazard ratio 0.59 — and its authors write that “prospective trials are needed to confirm causality.” People prescribed these drugs are not like people who are not, and nothing in that design fixes it.¹⁰

What none of it settles

Cancer takes years to appear and years to find. Semaglutide has been marketed since 2017, the longest randomized follow-up is under four years, and the drug’s heaviest use began in the 2020s. A study that could settle the thyroid question does not exist yet, and will not for some time. What exists is a warning that says so.

This page reports the approved labels and the published studies, quoted and attributed, and takes no position of its own on whether anyone should start, continue or stop the drug — that is a prescriber’s judgment, and this is not medical advice. Every study was located on PubMed and its PMID resolved before it was cited, and each row above names the drugs the study actually included. See the medical disclaimer.