GLP-1s and Autoimmune Thyroiditis: An Association, and the Reason to Doubt It

A very large cohort study found people taking a GLP‑1 were diagnosed with autoimmune thyroiditis about 30% more often than people taking a comparable diabetes drug. The number is real and the study is a serious one. The explanation most likely to be true, though, is not that the drug causes the disease — it is that everyone taking these drugs is handed a warning about their thyroid, and warnings produce tests.

What was measured

Researchers took a US medical-records network covering 4.8 million adults with type 2 diabetes, identified 412,021 people starting a GLP‑1 receptor agonist and 383,415 starting a DPP-4 inhibitor between 2015 and 2022, and matched them on baseline characteristics to 290,770 in each arm. Followed for up to eight years, the GLP‑1 group had a higher recorded incidence of several autoimmune conditions: autoimmune thyroiditis at a hazard ratio of 1.30 (95% CI 1.24–1.38), ankylosing spondylitis also 1.30, psoriasis 1.17, ulcerative colitis 1.11 and rheumatoid arthritis 1.08. Sensitivity analyses held.^[1]

Those are narrow confidence intervals on very large numbers, which is what makes the association hard to dismiss as noise. It is also, on its own, a statement about how often a diagnosis was recorded, which is not the same thing as how often the disease occurred.

The comparison group is the problem

That is textbook detection bias, and it produces exactly this shape of result: a moderate, statistically solid elevation in a diagnosis that requires somebody to order a test. The study is an active-comparator new-user design, which is the right way to build it and handles a great deal of confounding. It cannot handle this one, because the exposure itself changes the probability of being looked at. The authors present an association and recommend monitoring; they do not claim causation, and neither should anyone quoting them.

It is worth saying plainly that detection bias being the likeliest explanation is not the same as it being the only one. GLP‑1 receptors appear on immune cells and the drugs have measurable immunomodulatory effects, so a real mechanism is not far-fetched. What is missing is any study designed to tell the two apart — one where both groups get the same thyroid testing schedule regardless of which drug they are on.

The second study, and why its number should not be quoted

A 2026 chart review followed 527 patients through twelve months of tirzepatide at 2.5 to 15 mg weekly. Twenty-eight of them — 5.3% — developed or had progression of some thyroid condition, most often nodular or goiter disease and drug-induced thyroiditis. On multivariate analysis, two things predicted it: end-stage renal disease, at an odds ratio of 2.94, and already having thyroid disease at baseline, at 3.78.^[2]

It is useful for one thing, which is what its authors actually propose: identifying who is worth watching. Someone starting tirzepatide who already has a thyroid diagnosis, or end-stage renal disease, is in a higher-risk group than someone who does not, and periodic thyroid testing in those two groups is a modest, cheap suggestion. That is a monitoring recommendation, not a risk estimate.

This is not the boxed warning

Autoimmune thyroiditis and medullary thyroid carcinoma get confused constantly, and they have almost nothing to do with each other. The boxed warning on these labels concerns a thyroid cancer seen in rodents, and what it means for humans is covered separately in what the thyroid cancer boxed warning actually says. Hashimoto’s and drug-induced thyroiditis are immune conditions, usually managed with thyroid hormone replacement, and nothing above bears on cancer risk in either direction.

The practical read is narrow. If you already have a thyroid condition, that is worth your prescriber knowing before you start, and the second study is a reasonable argument for checking your thyroid function periodically once you have. If you do not, an association of this size, in a cohort that could not control for who got tested, is not a reason to change anything.