Across 100,038 matched pairs of veterans with diabetes, GLP-1 drugs were tied to no more and no fewer blood cancers than an older diabetes pill. The odds were flat across all four cancers the researchers checked, and every confidence interval crossed the line that means no difference.
The finding, published online June 18 in Clinical Lymphoma, Myeloma and Leukemia ↗, fills a quiet gap in a fast-growing safety literature. GLP-1 receptor agonists, the drug family that includes semaglutide ↗ (sold as Ozempic and Wegovy), have been studied hard for links to thyroid and pancreatic cancer. The cancers of blood and bone marrow (lymphoma, leukemia, and multiple myeloma, grouped together as hematologic malignancies) had barely been examined.
The design is the point
A null result is only as good as the comparison behind it. The researchers pulled Veterans Affairs national records from 2006 to 2021. They compared people starting a GLP-1 drug against people starting a DPP-4 inhibitor, a different pill for the same disease. Both groups had diabetes. Both were beginning a new drug on the same day of their care, so the contrast isolates the drug rather than the diabetes or the weight that usually travels with it. Then they matched the two arms on 71 baseline characteristics before counting a single cancer, from an eligible pool of 409,334 patients down to 100,038 pairs.
The counts came out almost on top of each other. Hodgkin lymphoma appeared in 32 GLP-1 users and 36 comparator users. Non-Hodgkin lymphoma, 253 against 260. Leukemias, 248 against 258. Multiple myeloma, 115 against 126. Every odds ratio landed between 0.89 and 0.97. Every confidence interval spanned 1.0, the statistical way of saying the small gaps could easily be chance. The pattern held when the authors sliced the data by how long people took the drug, their starting weight, and how much weight they lost.
A clean null across 100,000 pairs is not a non-event. In a class that millions now take for weight loss, the absence of a blood-cancer signal is exactly the kind of result that should travel as far as the scares do.
Mixed by cancer type
This lands next to a different GLP-1-and-cancer result pointing the other way. In May, a large cohort tied GLP-1 drugs to far fewer cancer metastases ↗ than DPP-4 inhibitors, using the same comparator drug. So the emerging picture is not one story. There is a protective hint for the spread of some solid tumors, and a flat nothing for the blood cancers. The honest read is that GLP-1 effects on cancer, if they exist, are specific to tissue and mechanism, not a blanket property of the class.
The caveats are the usual ones for this kind of study, and they matter. This is observational, not a randomized trial, so an unmeasured difference between the groups could still hide in the numbers. Veterans skew older and male, so the counts may not carry to a younger or more female population. Blood cancers develop slowly, and even a 2006 start date leaves a limited follow-up window. And a null cannot prove that no effect exists. It can only say that if one does, it was too small to see in roughly 200,000 people.
For a platform that hosts a live GLP-1 receptor ↗ page and cards for every drug in the class, the useful version of this result is boring on purpose. One more cancer question asked cleanly, and answered with a shrug, is worth more to a nervous reader than another headline that outruns its data. ↯