The slide deck went up at the American Society of Clinical Oncology’s annual meeting last week with the kind of number that books its own news cycle: women in a Penn Medicine cohort who had been prescribed a GLP-1 drug were about thirty percent less likely to have a breast cancer diagnosed during follow-up. The press release went out the same day. By the next morning the figure was riding under brand names, attached to Ozempic, Wegovy, Mounjaro, and Zepbound across the consumer health press, presented as the latest entry in a now-familiar genre: the weight-loss drug that may also do something else for you. The randomized record it walked past says something quieter, and the slide does not.

Start with what the study is. The paper, by Elizabeth McDonald and colleagues in JCO Oncology Practice, is a retrospective look at 111,646 women aged 45 to 80 with a BMI of at least 25 who underwent breast imaging inside the Penn Medicine system between January 2022 and June 2025 and had a documented imaging outcome. About 13.7 percent had at some point received a GLP-1 receptor agonist prescription before the exam date; the remainder had no documented exposure. The headline 30.5 percent comes out of a propensity-matched cohort of just over 30,000 women, paired on age, race, ethnicity, highest recorded BMI, breast density, and diabetes status. No randomization, no protocolized follow-up, no prospective prevention design, and, the authors note, the analysis did not account for which GLP-1 a given patient had taken, the dose, or the duration. McDonald herself told reporters the findings were promising but not yet proof.

That is the floor. The ceiling is where the design starts to creak.

Weight loss in postmenopausal women is itself an established modifier of breast-cancer risk, with its own literature, and it predates the GLP-1 era by years. Propensity matching on highest recorded BMI freezes a number; it does not capture what each woman’s weight was doing across three and a half years of follow-up, which is precisely the variable a GLP-1 prescription is designed to change. The drug class works by altering a trajectory the match cannot see. Whether the signal is a drug effect, a weight-loss effect, or some unresolved combination of the two is a question this design cannot answer, and the paper does not pretend otherwise. The downstream coverage was less careful, which is how a slide ends up under a Wegovy headline.

Then there is the byline. John B. Buse, the Chapel Hill diabetes researcher and one of the most cited names in the field, is a co-author, and his ASCO disclosure runs to consulting and advisory relationships with Novo Nordisk (semaglutide), Eli Lilly (tirzepatide), AstraZeneca, Boehringer-Ingelheim, Amgen, and a long string of other manufacturers across the diabetes and metabolic class. That is not a hidden tie and it does not make the analysis wrong. It does mean a result favorable to the GLP-1 class as marketed entered a brand-name news cycle with one co-author financially tied to the companies that sell the brands. The study’s direct funders are the American College of Radiology, the Pennsylvania Breast Cancer Coalition, and Penn’s Abramson Cancer Center. None of those is a pharmaceutical company. The press cycle is, nonetheless, a product.

The reason the framing matters is that the randomized record already exists, and it is quieter. A 2025 review pooling randomized trial data on GLP-1 receptor agonists looked at the impact and safety of GLP-1 agents in breast cancer patients and found that “current data do not suggest adverse safety signals” but also found no significant protective effect confirmed in prospective data. Those were trials designed to test glucose control and cardiovascular endpoints, not prevention, but they are the closest thing to random assignment the field currently has, and they did not see what the Penn slide describes. The 2024 JAMA Network Open paper that did most of the work pushing the GLP-1-as-cancer-prevention story into the mainstream examined thirteen obesity-associated cancers in a database of type 2 diabetes patients and tied the class to significantly lower risk of ten of them compared with insulin. Postmenopausal breast cancer was among the thirteen studied, and it was not one of the ten; the drugs showed no significant effect on breast-cancer risk in either direction. That detail did not make many of the recaps.

What ASCO got, and what newsrooms then ran with, is a single-system observational signal with a generous relative number, a press-release cycle built around it, and a co-author whose name lends weight to anything he signs. What the underlying science says is that the randomized adverse-event record is unimpressed, the mechanistic story is unsettled, the largest observational tailwind cited for the class did not actually find a breast-cancer signal, and the prospective trial McDonald says is being planned has not been done. Until it has, the number on the slide is a hypothesis with good marketing, dressed for a headline it has not yet earned.

Sources

  1. JCO Oncology Practice – McDonald et al., GLP-1 receptor agonists and breast cancer incidence, Penn Medicine cohort (2026)
  2. Penn Medicine press release – GLP-1 use linked to lower breast cancer incidence in large cohort study (2026)
  3. ASCO Post – GLP-1 RAs may be associated with lower breast cancer incidence (2026)
  4. ScienceDaily – Ozempic and similar weight-loss drugs linked to 30% lower breast cancer risk (2026)
  5. JAMA Network Open – Wang et al., GLP-1 receptor agonists and 13 obesity-associated cancers in type 2 diabetes patients (2024)
  6. PMC – The Impact and Safety of GLP-1 Agents and Breast Cancer: review of RCT and observational data (2025)