Women on GLP-1 drugs were diagnosed with breast cancer about 30 percent less often than women who were not. Matching the two groups on the obvious confounders did not erase the gap.
That is the headline from a retrospective study published June 2 in JCO Oncology Practice ↗ by a team at the University of Pennsylvania. The researchers pulled electronic health records from a major academic center and its affiliated sites, covering January 2022 through June 2025, and built a cohort of women between 45 and 80 with a body mass index of 25 or higher who had undergone breast imaging. After filtering, that left 111,646 women, median age 61. The question was whether a prescription for a GLP-1 receptor agonist ↗, the drug class that includes semaglutide ↗, the molecule in Ozempic and Wegovy, tracked with fewer breast cancers.
The number, and then the harder number
In the full cohort, women who had filled a GLP-1 prescription before their imaging exam had lower odds of a breast cancer diagnosis, with an odds ratio of 0.649 (95 percent confidence interval 0.569 to 0.741). In plain terms, their odds of being diagnosed were roughly a third lower than women who had not taken the drugs.
The obvious objection is that GLP-1 users differ from non-users in ways that themselves change cancer risk. They tend to be heavier to start, more likely to have type 2 diabetes, and they differ by age, race, and ethnicity. Excess weight is one of the few modifiable risk factors for breast cancer, so any honest analysis has to account for it.
So the authors did a one-to-one match. Using propensity scores built from age, race, ethnicity, highest recorded BMI, breast density, and diabetes history, they paired each GLP-1 user with a comparable non-user, leaving 30,528 matched observations and 600 cancer cases. The protective association held: odds ratio 0.695 (95 percent confidence interval 0.590 to 0.819). Both results cleared statistical significance at p below 0.0001. The matched estimate is the one that matters, and it still points to roughly 30 percent lower odds.
What a cohort like this can and cannot say
This is an association, not a verdict. The study cannot prove that GLP-1 drugs prevent breast cancer, and the authors do not claim it does. The primary outcome was breast cancer detection in women who were already getting imaging, which means the cohort is selected for women engaged with screening, and weight loss itself changes breast density and how easy a tumor is to see. Residual confounding, the influence of variables the match did not capture, is always live in a records-based study.
What the matching does is take the easy dismissal off the table. If the entire signal were driven by the fact that GLP-1 users are heavier or more diabetic, adjusting for highest BMI, density, and diabetes history should have collapsed it. It did not. The effect size barely moved, from 0.649 to 0.695, which is the opposite of what a confounded result usually does under matching.
The cancer story is becoming a recurring one for this class. Peptidemodel earlier covered a separate cohort in which GLP-1 use tracked with roughly half the rate of lung cancer metastasis ↗ versus an older diabetes drug, with signals in three other tumor types. None of these are trials. They are large observational hints pointing the same direction, which is exactly the setup that justifies the expensive thing the authors ask for at the end: a prospective trial that gives some women the drug, some a placebo, and counts the cancers ↗ that follow.