Much-hyped research linking GLP-1 drugs with a significant reduction in death risk after breast cancer diagnosis is unpackedDr F Perry Wilson, from the Yale School of Medicine, who cautions against the study's dramatic claims.
Wilson writes in Medscape:
A recent study suggests that GLP-1 drugs like Ozempic might reduce the risk of death after a breast cancer diagnosis by 91%.
While these drugs may significantly reduce breast cancer mortality, data quality issues and potential biases cast doubt on the findings, with the study highlighting the need for robust evidence to support such dramatic claims.
I hate breast cancer – my wife is a breast cancer surgeon, after all. And I’ve been pretty pro-GLP-1 from the beginning. I want to believe this frankly unbelievable number. But I’m afraid this is one of those exciting headlines that basically falls apart once you start reading the actual paper.
We’re looking at a new study that has appeared in JAMA Network Open, which examines a huge cohort of 841 831 women with breast cancer across 68 US healthcare organisations. Before we dig into the data, let’s look at the biological plausibility here.
Breast cancer is an obesity-associated cancer. This isn’t to say that you have to be overweight to get breast cancer, but higher BMI is a clear risk factor. And there’s a reason for that. Fat produces oestrogen, which can promote certain types of breast cancer, and the metabolic complications of obesity may reduce immune surveillance that weeds out pre-cancerous cells before they become a problem.
GLP-1s, by reducing body weight and body fat, certainly seem as if they might be able to influence this system, and thereby reduce the risk of recurrence after a breast cancer diagnosis. Independent of recurrence, there are certainly mechanisms by which GLP-1s might improve overall survival; for example, by reducing rates of cardiovascular disease or chronic kidney disease.
The researchers wanted to quantify how much benefit one might expect from these drugs. Enter the TriNetX database. This is an administrative database of patients from health centres around the United States.
Its main strength is the sheer number of individuals in the data – 841 000 women with breast cancer, for example.
The main weaknesses? Well, it’s administrative data, which means you’re depending primarily on billing codes to try to figure out comorbidities and interventions. Also, it doesn’t provide individual level data; you have to do your analyses on their proprietary platform which aggregates data for you.
I suppose this is understandable, but I have to tell you as a researcher who works with electronic health record data, if I couldn’t drill down to the patient level, I’d be very unsure what was actually going on.
The structure of the analysis works like this. Identify patients with a new stage I-III breast cancer diagnosis, flag the group who were put on a GLP-1, match them with similar people who did not get a GLP-1, and follow those pairs forward in time until a cancer recurrence or death occurs.
It’s easier said than done. You might already be suspecting that people who receive GLP-1s are quite different from those who do not, and those differences may play into their risk of breast cancer recurrence or mortality completely independently of the drug they receive. So how do you match apples to oranges?
The authors use propensity scores. Basically, you make a multivariable model that predicts whether someone is going to receive a GLP-1 or not. You then run all the patients through the model, which generates an individual probability. For every person who actually received a GLP-1, you match with someone with the closest probability score who didn’t receive a GLP-1, and in that way, you’re trying to compare apples to apples.
Once matched, the authors look at a couple of outcomes, and the results are pretty dramatic. Here we’re looking just at GLP-1 users vs non-users. A 65% reduction in all-cause mortality and a 55% reduction in cancer recurrence.
The authors also compared GLP-1 users with diabetes to people with diabetes receiving either metformin or insulin. The GLP-1 users had a staggering 91% lower risk of all-cause mortality and 67% lower risk of recurrence.
The only slightly less jaw-dropping effects were when GLP-1 users were compared with users of SGLT2 inhibitors – newer diabetes drugs with fairly profound cardioprotective effects. Here there was no difference in all-cause or recurrence-free survival.
I think the best way to think through data like these is to consider what possibilities could be true.
One: GLP-1s truly and dramatically improve survival after breast cancer. It has been shown that some breast cancers express the GLP-1 receptor, so perhaps it is a disease-modifying therapy here. Or the weight loss associated with GLP-1 reduces oestrogen production.
But that wouldn’t explain why we see similar benefit with SGLT2 inhibitors. And, of course, people with oestrogen receptor-positive breast cancer tend to get estrogen blocking drugs like tamoxifen, so I’m not sure if the marginal contribution of a little less oestrogen from fat makes a biological difference here.
Option 2: We’re not seeing a modification in breast cancer risk, we’re seeing benefits from reduction of risk of other diseases. Most women with breast cancer don’t die of breast cancer, they die of heart disease and it is clear that both GLP-1s and SGLT-2 inhibitors reduce that risk. Of course, that doesn’t explain the lower risk of recurrence of breast cancer with these drugs.
Which leaves us option 3, which is what I think is actually going on here. It’s our old nemesis, bad observational data.
For one thing, this study clearly suffers from immortal time bias. Individuals who received a GLP-1 any time after their breast cancer diagnosis were classified in the GLP-1 group. This is a problem, because the longer you survive, the more opportunity you have to be prescribed a GLP-1.
To be fair, the authors address this in a secondary analysis, but the results of that analysis are not as impressive as the primary one reported in the paper.
That’s not the biggest problem though.
We all know there is confounding here. People who get GLP-1s are different from those who don’t. These are expensive drugs and relatively new to the marketplace – at least for obesity treatment.
Propensity score matching is supposed to address that, but the quality of the match is dependent on the quality of the underlying data, and, frankly, I don’t trust this data quality at all.
For example, let’s look at the treatments these women got for this stage I-III breast cancer: 4% mastectomy? 2% partial mastectomy? 3% radiation therapy? 22% chemo? These are clearly radical underestimates.
There is no way that less than 10% of a broad sample of women are getting surgery for breast cancer. That’s just not plausible. It suggests that this dataset has a lot of missing data in it.
Another example – according to their table 1, 14% of women had ER+ cancers, and 3.5% ER- cancers. I’m no mathematician, but I don’t think that maths works unless there is some third type of ER receptor status secretly affecting 80% of women with breast cancer.
You can’t make good matches with this much missing data. If you’re looking at breast cancer recurrence, it’s a good idea to match on the type of surgery the woman had initially or her oestrogen receptor status.
But according to this, 90% of women had no surgery at all, and 80% had neither ER+ or ER- cancer, so you can’t match on that. Which means your matches are bad.
Which means this study has not fixed the underlying confounding problem which is that women with access to GLP-1s are less sick than women who don’t have access. Hence the dramatic improvement in survival vs women taking metformin or insulin for diabetes. Hence the similar effects of SGLT2-inhibitors, despite a completely different mechanism of action.
It’s another new, expensive drug, which flags women with better access to care.
This is not to say that GLP-1s will have no effect on breast or other cancer. Indeed, other studies have suggested as much. But this study is a good reminder that extreme results require extreme amounts of evidence, and this paper falls short of that threshold for me.
F. Perry Wilson, MD, MSCE, is an associate professor of medicine and public health and director of Yale’s Clinical and Translational Research Accelerator.
Study details
Survival and Recurrence with GLP-1 Receptor Agonists in Breast Cancer
Kristina Tatum, Bassam Dahman, Aniyah Stevenson et al.
Published in JAMA Network on 11 May 2026
Key Points
Question Is there a potential benefit associated with the use of glucagon-like peptide-1 receptor agonists (GLP-1 RAs) in female breast cancer survival and cancer recurrence among patients with obesity or with type 2 diabetes (T2D)?
Findings In this cohort study using propensity score–matched data from 841 831 patients with breast cancer, GLP-1 RA use vs non-use was significantly associated with improved survival and reduced recurrence among patients with obesity. Similar benefits were seen among patients with T2D when compared with other antidiabetic medications.
Meaning This study suggests that GLP-1 RAs may offer protective benefits beyond glycaemic and weight control, potentially improving survival and recurrence risk in some female patients with breast cancer.
Abstract
Importance
Patients with breast cancer (BC) with comorbid obesity or type 2 diabetes (T2D) experience poorer survival. Glucagon-like peptide-1 receptor agonists (GLP-1 RAs) are approved to treat these comorbidities; however, their associations with BC survival and recurrence remain unclear.
Objective
To evaluate the association between GLP-1 RA use and 10-year all-cause mortality and recurrence-free survival (RFS) over the 10-year follow-up period, as well as 5- and 10-year all-cause mortality and RFS probabilities among patients with BC.
Design, Setting, and Participants
This retrospective cohort study used TriNetX US Collaborative Network data from women (≥18 years) with BC from 68 health care organisations who received a diagnosis between April 1, 2006, and April 1, 2023. Propensity score matching balanced characteristics. Data were analysed between September 16 and October 3, 2025.
Exposures
GLP-1 RA use (≥2 prescriptions) during the 6 months before and any time after the index diagnosis; nonuse (0 entries).
Main Outcomes and Measures
The primary outcome was all-cause mortality, and the secondary outcome was RFS. Cox proportional hazards regression model–estimated hazard ratios (HRs) were restricted to 10 years. Kaplan-Meier estimators were used to calculate 5- and 10-year all-cause mortality and RFS probabilities. Prespecified subgroup (postmenopausal) and landmark (6- and 12-month) analyses were conducted.
Results
The study comprised 841 831 eligible patients with BC (mean [SD] age, 69.1 [12.2] years). After exclusions and 1:1 propensity score matching, 3 cohorts were identified: 1610 patients for GLP-1 RA use vs nonuse (patients with obesity [body mass index ≥30]), 2323 patients for GLP-1 RA use vs insulin or metformin (patients with T2D), and 4052 patients for GLP-1 RA use vs sodium-glucose cotransporter 2 inhibitors (patients with T2D). Among patients with obesity, GLP-1 RAs were associated with lower hazard of all-cause mortality (HR, 0.35; 95% CI, 0.21-0.58; P < .001) and RFS (HR, 0.44; 95% CI, 0.30-0.64; P < .001) over a 10-year follow-up period. Among patients with T2D, GLP-1 RAs vs insulin or metformin were associated with lower hazard of all-cause mortality (HR, 0.09; 95% CI, 0.06-0.15; P < .001) and RFS (HR, 0.33; 95% CI, 0.21-0.50; P < .001). No significant differences were observed between GLP-1 RA and sodium-glucose cotransporter 2 inhibitor groups. Subgroup and landmark analyses yielded similar findings.
Conclusions and Relevance
In this cohort study of patients with BC, findings suggested a potential association between GLP-1 RA use and improved outcomes among patients with BC who have obesity and related metabolic conditions. These findings support further evaluation of GLP-1 RA therapy in randomised clinical trials.
Medscape article – ‘Too Good to Be True?’ GLP-1s and Breast Cancer Survival (Open access)
See more from MedicalBrief archives:
Weight-loss jabs could halve obesity-linked cancer risks – Israeli study
Boundless benefits of weight-loss drugs under scrutiny
Almost 40% of cancers tied to modifiable risk factors – global report