Nothing in the published record shows that they do, and the honest form of the answer is a bound rather than a reassurance. Across 437,077 patients in three countries, thyroid cancer occurred at 1.33 events per 10,000 person-years on a GLP-1 and 1.46 on the comparator, a hazard ratio of 0.93. The authors put a ceiling on what their data could still be hiding: no more than a 31% increase in relative risk.
For medullary thyroid cancer, the one the boxed warning actually names, the interval runs from 0.37 to 3.86 — too wide to rule much in or out, and saying so is more useful than reassurance. The broader cancer question has been asked separately, in 86,632 matched adults with obesity, where three obesity-associated cancers fell and none rose[2]; that reading is in what the cancer cohort found.
The thyroid warning on these drugs comes from rodent studies, and it has followed the class into every patient leaflet since. Whether it translates to people has taken a decade and several very large datasets to address. And the answer is more interesting than either “yes” or “no” — it is a bound — the same discipline this site applies to a published price in what happens when you read a price twice.
What was compared
A nationwide cohort study used registry data from three countries to compare 145,410 patients treated with GLP-1 receptor agonists against 291,667 treated with DPP-4 inhibitors[1]. Mean follow-up was 3.9 years in the first group and 5.4 in the second.
Seventy-six thyroid cancers occurred in the GLP-1 group, an incidence rate of 1.33 per 10,000 person-years, against 184 in the comparator group at 1.46. The hazard ratio was 0.93 (95% CI 0.66 to 1.31), and the rate difference was −0.13 events per 10,000 person-years (95% CI −0.61 to 0.36). Both groups had type 2 diabetes, which is a different population from the one most sellers here serve — a gap that also limits the results in GLP-1 and cancer risk.
The subgroup the warning is about
Medullary thyroid cancer is the specific tumor the rodent data raised and the labeling names. In this study its hazard ratio was 1.19, with a confidence interval from 0.37 to 3.86.
That interval spans a two-thirds reduction and a near-quadrupling. It is what happens when a very rare cancer is looked for in a cohort that is large but not large enough, and it means the study cannot settle the question the warning exists for. A comparison against SGLT-2 inhibitors gave 1.16 (0.65 to 2.05) for thyroid cancer overall, which is consistent and equally imprecise.
What a buyer does with this
The practical answer is unchanged by the study, because the labeling already says it: anyone with a personal or family history of medullary thyroid carcinoma or multiple endocrine neoplasia type 2 should not take these drugs. That is a contraindication rather than a risk calculation, and it is the sort of thing an intake form should ask about.
Whether it does is a separate question. Most sellers tracked here publish nothing about what their intake asks or who reviews the answers, which we count in what sellers will not tell you and list seller by seller in the questions they leave open. A contraindication only works if somebody checks for it, and the wider pattern of claims that outrun their evidence is in what the semaglutide trials are worth.