A propensity-matched cohort of 22,340 patients with early-stage cancer found that the group on GLP-1 receptor agonists was 38 to 50 percent less likely to develop metastatic disease than the group on a different class of type 2 diabetes drug. The effect reached statistical significance in four of seven cancer types studied: non-small cell lung, breast, colorectal, and hepatocellular carcinoma. Prostate, pancreatic, and renal cell carcinoma trended in the same direction but did not pass the significance bar.
The analysis is ASCO 2026 abstract 3143, led by Mark D. Orland of the Cleveland Clinic Taussig Cancer Institute with senior author Jaroslaw Maciejewski. It used the TriNetX research network to identify 10,225 adults who started a GLP-1 RA (semaglutide ↗, tirzepatide ↗, liraglutide ↗, or dulaglutide ↗) after a stage I to III cancer diagnosis, and 12,115 adults who started a dipeptidyl peptidase-4 inhibitor (gliptin) in the same setting. Patients were propensity-matched on age, race, BMI, comorbidities, cancer treatments, and concomitant medications. The endpoint was progression to stage IV over five years. The findings were reported by Healio ↗ and picked up by the Wall Street Journal ↗ on May 21, ahead of the in-person presentation at the ASCO Annual Meeting May 29 to June 2 in Chicago.
What was significant
In non-small cell lung cancer, 22.3 percent of gliptin patients had progressed to stage IV by the end of follow-up, versus 10.0 percent of GLP-1 patients. That is a hazard ratio of 0.50 (95% confidence interval 0.43 to 0.59, p less than 0.001), or a roughly 50 percent reduction in the relative risk of metastatic progression. In breast cancer, 20.1 percent versus 10.2 percent (HR 0.57, 95% CI 0.46 to 0.71). In colorectal cancer, 22 percent versus 13 percent. In hepatocellular carcinoma, 28 percent versus 19 percent. Each of those is a 32 to 50 percent relative reduction on an absolute base rate that ran in the high teens to mid-twenties in the gliptin arm. The risk reduction is not subtle.
Why the active comparator is the design point
Earlier observational cancer signals for GLP-1 drugs have run into a confounder the field knows well. Patients who get on the newest, most desirable T2D drug class tend to be patients with better access to care, better adherence, and more attention from their oncologists, all of which lower the odds of progressing to stage IV for reasons that have nothing to do with the drug. A comparison against no treatment, or against the general population, cannot separate those effects from a true drug effect. A comparison against another T2D drug class started in the same diagnostic moment can, at least in principle. DPP-4 inhibitors (sitagliptin, linagliptin, alogliptin, saxagliptin) are an older, cheaper, no-weight-loss class with no established cancer-direction signal, which makes them the cleanest active comparator a TriNetX analysis can pull. This is a stronger design than most of the prior GLP-1-and-cancer literature has reached.
What did not reach significance
Three of the seven cancer types studied (prostate, pancreatic, and renal cell carcinoma) showed protective point estimates but did not pass the conventional significance threshold. The authors did not over-interpret. Maciejewski told Healio that the results "need further confirmation and must be viewed with proper dose of skepticism." Orland flagged the standard caveat about residual confounding in the database.
Mechanism, with caveats
The mechanistic hypothesis the authors offered focuses on the anti-inflammatory and immune-modulating effects of GLP-1 receptor agonism, plus the substantial weight-loss-driven reduction in obesity-related cancer biology. The receptors are expressed beyond the islet and the gut. Multiple animal and human signals now suggest GLP-1 agonism dampens the chronic low-grade inflammation implicated in the microenvironments that allow metastatic seeding. None of this rises to the level of a worked-out mechanism. It is a plausible direction.
Platform note
The cancer story for this class is moving fast. The May 11 piece on GLP-1 RAs cutting breast cancer mortality ↗ showed a separate breast-cancer survival signal that an SGLT2-inhibitor active comparator largely matched, which argued the breast effect may not be GLP-1-specific. The May 18 GLP-1 thyroid landmark analysis ↗ found the boxed-warning thyroid signal did not survive landmark analysis. Today's piece adds a metastasis-specific signal across four cancers with a cleaner active-comparator design than either of those.
The next thing to watch is whether the same dataset can be split by GLP-1 agent. If the effect concentrates in semaglutide or tirzepatide versus the older liraglutide and dulaglutide, that argues for a weight-loss-driven mechanism rather than a class-wide receptor effect. If the effect is flat across all four agents, the inflammation hypothesis gets more support. The TriNetX dataset is large enough to support that breakout. ASCO 2026 starts May 29.