GLP-1 drugs cut cancer spread in 4 tumour types. ASCO 2026 data, honest reading.

What was presented at ASCO 2026

The 2026 ASCO Annual Meeting runs May 29 through June 2 in Chicago. Among the late-breaking abstracts presented during the obesity-and-cancer session, a research team led by oncology researchers using the TriNetX federated electronic-health-records network presented a retrospective cohort study of cancer patients who initiated either a GLP-1 receptor agonist or a DPP-4 inhibitor after their cancer diagnosis.

Both drug classes are antidiabetic. Both are commonly prescribed to cancer patients with type 2 diabetes. The two classes differ in mechanism: GLP-1 receptor agonists — liraglutide, dulaglutide, exenatide, semaglutide, tirzepatide — directly activate the GLP-1 receptor across multiple tissues including pancreas, gut, brain, heart, and as it turns out many tumours. DPP-4 inhibitors — sitagliptin, linagliptin, saxagliptin — block the enzyme that degrades endogenous GLP-1, raising endogenous GLP-1 levels modestly but to a fraction of what direct GLP-1 receptor agonists achieve.

The two classes thus form a natural-experiment comparison. Both populations have diabetes plus cancer. Both are receiving glucose-lowering therapy after diagnosis. The clinical decision to use a GLP-1 instead of a DPP-4 inhibitor is driven by efficacy, weight effects, cost, and prescriber preference — but the underlying diabetes severity is roughly similar. If GLP-1s have a cancer-specific effect, the comparison should surface it.

Study design — TriNetX cohort, GLP-1 vs DPP-4

The investigators queried the TriNetX network — a global federated network of de-identified electronic health records from over 120 healthcare organizations — for adults with a documented stage I, II, or III cancer diagnosis who initiated either a GLP-1 receptor agonist or a DPP-4 inhibitor after diagnosis. The pre-match screen identified 10,225 patients in the GLP-1 group and 12,115 in the DPP-4 group (22,340 total). After propensity-score matching, the analysis was run on a balanced cohort of roughly 6,000 patients per arm.

Seven cancer types were examined: lung, breast, colorectal, liver, prostate, pancreatic, and kidney. The primary outcome was subsequent metastasis or progression to stage IV disease, with overall survival as a secondary outcome. The propensity-score matching balanced baseline variables including age, sex, body mass index, stage at diagnosis, and key comorbidities. Hazard ratios for stage IV progression were estimated within each cancer type on the matched cohort.

The 4-of-7 headline result and the magnitudes

For four of the seven cancer types — lung, breast, colorectal, and liver — the GLP-1 cohort was 38 to 50 percent less likely to progress to stage IV disease compared with the DPP-4 cohort. The hazard ratios in those four cancers were approximately 0.50 to 0.62 with confidence intervals that excluded the null.

Forty to fifty percent reduction in metastasis risk is a large effect size — in the ballpark of (though larger than) the recurrence reduction from adjuvant chemotherapy in early breast cancer, which the EBCTCG meta-analyses put closer to 20 to 35 percent relative risk reduction. If this GLP-1 signal holds up in a properly designed prospective trial, the class would join a small set of pharmaceuticals with measurable anti-metastatic activity in epithelial cancers.

The four cancers that showed the signal share a feature: all four are associated with obesity. Obesity is an established risk factor for lung adenocarcinoma, postmenopausal breast cancer, colorectal cancer, and hepatocellular carcinoma. GLP-1 receptor agonists produce substantial weight loss — typically 10 to 22 percent of body weight depending on the agent and dose — and weight loss itself reduces cancer recurrence risk in observational data. The challenge is disentangling the direct receptor-mediated effect from the indirect weight-loss-mediated effect.

Forty to fifty percent reduction in metastasis is the largest cancer-prevention signal ever reported for a class of drugs already on the market for diabetes and weight loss.

The three tumour types that did not reach significance

For three of the seven cancer types — prostate, pancreatic, and kidney — the direction of effect was the same. The GLP-1 cohort had numerically fewer instances of metastasis than the DPP-4 cohort. But the differences did not reach statistical significance. Whether this reflects a true biological difference (some cancers do not respond to GLP-1 signaling) or simply lower statistical power (smaller per-cancer sample sizes) is unanswerable from this study.

The prostate-cancer null is worth noting because prostate is the most common cancer in men and would have been a high-priority finding if positive. The pancreatic-cancer result is worth noting because GLP-1 receptor agonists have a complex historical association with pancreatic events — early concerns about pancreatitis and pancreatic cancer in the 2010s were ultimately not borne out in large outcome trials, and the ASCO 2026 data add reassurance that GLP-1 use does not appear to accelerate pancreatic cancer progression.

Tumour GLP-1 receptor expression and survival

The investigators also examined tumour-level GLP-1 receptor expression using paired pathology and outcome data. Overall, high tumour GLP-1 receptor expression was associated with approximately 33 percent lower risk of death compared with low expression. The signal was strongest in breast cancer, where high tumour GLP-1 receptor expression was associated with a 45 percent lower risk of death.

This is the mechanistic complement to the cohort result. If GLP-1 receptor agonism slows metastasis, you would expect tumours with more receptors to be more susceptible — and tumours with more receptors should have better outcomes when the patient is on a GLP-1 drug. Both predictions are consistent with the receptor-expression data.

Receptor expression is not destiny. Many tumours have low or absent GLP-1 receptor expression, which would predict no direct anti-tumour effect of GLP-1 agonism in those tumours. The cohort-level result and the receptor-level result together suggest a mixed mechanism: a direct receptor-mediated effect in receptor-positive tumours plus an indirect effect mediated through weight loss, insulin sensitivity, and systemic inflammation reduction in all tumour types.

Why the DPP-4 comparator matters

The choice of DPP-4 inhibitor as the comparator is the most important methodological feature of this study. A naive comparison would have been GLP-1 users versus all other antidiabetics, or GLP-1 users versus no-antidiabetic. Both would have been heavily confounded.

DPP-4 inhibitors are the cleanest comparator class. They are prescribed in roughly the same clinical setting — type 2 diabetes that requires escalation beyond metformin. They have weight-neutral effects, modest glycemic efficacy, and a safety profile that makes them a common choice for older or frailer patients with diabetes plus cancer. The propensity-score matching balanced baseline severity reasonably well across the two groups.

That said, the comparison is still observational. Clinicians may be more likely to choose GLP-1 over DPP-4 in patients with better functional status, higher BMI, or younger age — and those characteristics independently predict better cancer outcomes. Propensity-score matching reduces but does not eliminate this kind of confounding. The ASCO 2026 result is hypothesis-generating, not hypothesis-confirming.

The mechanistic story — what could be happening

Several mechanisms could explain a GLP-1 anti-metastatic effect, and most are not mutually exclusive.

First, weight loss. Body fat is metabolically active tissue that produces estrogens, insulin, IGF-1, and inflammatory cytokines — all of which promote tumour growth in receptor-positive cancers. Losing twenty percent of body weight reduces all four. The magnitude of the GLP-1 weight effect is in the range that produces measurable changes in breast cancer recurrence in surgical and lifestyle studies.

Second, direct GLP-1 receptor signaling in tumours. The receptor is expressed at variable levels across cancer types and is functional in cell culture — activation can reduce proliferation, alter glucose metabolism, and modulate the local immune environment in some tumour models. The receptor-expression-survival correlation supports this mechanism.

Third, systemic inflammation reduction. GLP-1 receptor agonists reduce C-reactive protein, IL-6, and other inflammatory markers — partly through weight loss and partly through direct effects on immune cells. Chronic inflammation supports a pro-metastatic microenvironment.

Fourth, insulin sensitization. Hyperinsulinaemia drives proliferation in IGF-1-receptor-positive tumours. GLP-1s reduce insulin requirements and improve sensitivity. This mechanism would predict the largest benefit in tumours expressing the insulin receptor or IGF-1 receptor.

None of these are proven as the dominant mechanism. The honest answer is "probably some combination, with different mechanisms dominating in different cancers."

The limits of retrospective claims data

The ASCO 2026 result is a retrospective TriNetX cohort presented as a late-breaking abstract. It is not a randomised controlled trial. Three specific limits matter.

First, confounding by indication. Clinicians choose GLP-1 over DPP-4 for reasons that correlate with prognosis. Propensity matching captures measured confounders. It cannot capture unmeasured ones — performance status, patient motivation, household income, dietary patterns. All of these correlate with both drug choice and cancer survival.

Second, claims-data outcome adjudication. TriNetX records what was billed, not what was clinically confirmed. Stage IV progression coded in claims may misclassify some patients. The misclassification is non-differential — likely to affect both arms equally — but it adds noise.

Third, the abstract-stage limitation. ASCO meeting abstracts present preliminary results. The full peer-reviewed paper, when published, will include sensitivity analyses, additional adjustments, and the chance for the broader oncology community to scrutinise the methods. The pattern in oncology is that abstract effect sizes often shrink when the full paper is published. The 38 to 50 percent metastasis reduction may attenuate.

What this changes — and what it does not

The ASCO 2026 result does not change clinical practice tomorrow. No oncologist will start prescribing GLP-1 receptor agonists for cancer treatment based on a retrospective abstract. Adjuvant therapy decisions are made based on randomised trial evidence, and there are no randomised trials of GLP-1 receptor agonists as adjuvant cancer therapy yet.

What it does change: it accelerates the case for prospective trials. Several investigator-initiated trials of GLP-1 receptor agonists in early-stage cancer are already enrolling. The ASCO 2026 data provide the effect-size estimate those trials need to power their primary endpoints, and they suggest the highest-yield populations are breast, lung, colorectal, and liver cancer. Expect more trial activity in those four cancers over the next 24 to 36 months.

It also changes the conversation for patients already on GLP-1 receptor agonists who are diagnosed with cancer. Pre-ASCO, the conversation was: "We will pause the GLP-1 if needed for surgery or chemotherapy and reassess." Post-ASCO, the conversation is: "There may be cancer-specific reasons to continue the GLP-1, especially if you have one of the four cancers that signaled strongly." Many oncologists will now consult endocrinology rather than discontinue GLP-1s reflexively at cancer diagnosis.

The headline reading of this study — "GLP-1 drugs cut cancer spread" — is defensible. The careful reading is that an obesity-driven class of drugs reduces metastasis in obesity-associated cancers in a retrospective cohort, by an amount that probably reflects a mixture of weight loss, direct receptor signaling, and reduced systemic inflammation. The signal is real. The magnitude needs prospective replication. The mechanism is plural.