GLP-1 Receptor Agonists and Cancer Risk: What a 106,000-Patient Study Found

TL;DR: In a propensity-score-weighted cohort of 106,088 adults with type 2 diabetes, GLP-1 receptor agonist use was associated with a 53% lower hazard of liver cancer (HR 0.47) and a 77% lower hazard of pancreatic cancer (HR 0.23) compared with insulin, while the incidence of 14 other cancer types was statistically indistinguishable between groups.

Abstract

Study at a Glance

Study Snapshot: Key Statistics

py = person-years; HR = hazard ratio; CI = confidence interval. Rates reflect post-OPSW estimates.

Study Facts Table

What Researchers Actually Did

Rashid and colleagues used the IBM MarketScan commercial claims database to identify adults aged 18 to 64 diagnosed with type 2 diabetes between 2013 and 2021. They constructed two mutually exclusive groups: patients who initiated a GLP-1RA medication (and took no insulin) and patients who initiated insulin (and took no GLP-1RA). Pharmaceutical exposure was confirmed through national drug codes, and continuous use for at least six months from the index prescription was required. A three-month lead-in window was applied after that six-month confirmation period before cancer events were counted, and any individual with a new cancer diagnosis in the first nine months after the index prescription was excluded to reduce detection bias. Patients with a cancer diagnosis in the year preceding the index date were also excluded.

To address the substantial baseline differences between patients who receive GLP-1RAs versus insulin in clinical practice, the investigators applied Overlap Propensity Score Weighting (OPSW) across approximately 50 covariates, including age, sex, body mass index, adverse social determinants of health, personal or family cancer history, diabetes-related complications, genetic cancer predisposition, concurrent antidiabetic medications, and prior bariatric surgery. Following OPSW, standardized differences across measured covariates were reduced to less than 0.1. Cox proportional hazards models were then used to estimate hazard ratios for 16 specific cancer types, with Kaplan-Meier cumulative incidence curves generated for each. Participants were followed until cancer diagnosis, loss to follow-up, death, five years after index prescription, or end of the study period, whichever came first. Sensitivity analyses excluded patients using DPP-4 inhibitors and stratified by sex.

Key Findings: Primary Outcomes

• Of 106,088 patients, 1.9% (n = 1,594) developed cancer during a median follow-up of 30 months (IQR 20–46).
• Liver cancer: GLP-1RA was associated with a 53% lower hazard compared with insulin (HR 0.47, 95% CI 0.27–0.82; p < 0.05). Post-OPSW incidence rates were 0.34 per 1,000 person-years (GLP-1RA) versus 0.71 per 1,000 person-years (insulin).
• Pancreatic cancer: GLP-1RA was associated with a 77% lower hazard compared with insulin (HR 0.23, 95% CI 0.11–0.51; p < 0.001). Post-OPSW incidence rates were 0.14 per 1,000 person-years (GLP-1RA) versus 0.61 per 1,000 person-years (insulin).
• Overall combined cancer incidence was lower in the GLP-1RA group (6.5%, 95% CI 6.1–7.0) compared with insulin (7.7%, 95% CI 7.2–8.2).
• No statistically significant difference in cancer risk was identified for thyroid, lung, breast, esophageal, gastric, biliary, small intestinal, renal, bladder, colorectal, prostate, ovarian, endometrial, or neuroendocrine cancers (all p > 0.05).

Key Findings: Secondary Outcomes and Subgroup Analyses

• A sex-stratified sensitivity analysis was performed to account for sex-specific cancers; results were reported as consistent with the primary analysis, though granular sex-specific hazard ratios are not provided in the published tables.
• A secondary sensitivity analysis excluding patients who received DPP-4 inhibitors produced consistent findings (Supplementary Table 2), suggesting that concurrent incretin-based therapy was not driving the observed associations.
• Among cancer types with numerically lower point estimates that did not reach statistical significance, biliary cancer showed an HR of 0.32 (95% CI 0.06–1.59) and gastric cancer an HR of 0.57 (95% CI 0.21–1.54). The wide confidence intervals for these low-incidence cancers preclude any conclusion.
• Small intestinal cancer showed an HR of 1.85 (95% CI 0.46–7.41) favoring insulin, though the confidence interval is extremely wide and the event count is low (n = 23 total), rendering this numerically elevated estimate uninformative.
• The median duration of GLP-1RA use was 21 months (IQR 15–31) and insulin use was 23 months (IQR 15–38), indicating broadly comparable exposure durations between groups.

Adverse Events and Safety Profile

This study was designed to assess cancer incidence as an outcome, not to capture adverse events from GLP-1RA therapy. Accordingly, no structured adverse event data are reported. Notably, the study found no increased cancer signal for any of the 16 cancer types evaluated, including thyroid and neuroendocrine tumors, which have carried regulatory concern based on preclinical animal data. The absence of a thyroid cancer signal in this cohort is consistent with the study authors’ observation that clinicians may already be excluding patients with familial thyroid cancer or multiple endocrine neoplasia risk from GLP-1RA prescriptions, per FDA guidance, which would attenuate any detectable signal in a real-world claims dataset.

Statistical Approach and Rigor

The investigators used Overlap Propensity Score Weighting, a causal inference technique that assigns greater weight to individuals near the region of propensity score overlap between treatment groups. The authors cite its superiority over Inverse Probability of Treatment Weighting for achieving covariate balance without extreme weight truncation. Post-weighting standardized differences below 0.1 were achieved across all measured covariates, which is a standard threshold for balance adequacy. Cox proportional hazards models provided time-to-event estimates, and Kaplan-Meier survival analysis generated cumulative incidence curves adjusted for baseline characteristics. Cancer outcomes required either one inpatient or two outpatient ICD-9/10 diagnoses, a two-source confirmation criterion that reduces false-positive outcome misclassification. The study followed STROBE reporting guidelines. The primary statistical limitation is that OPSW, however well-executed, balances only measured covariates; residual confounding from unmeasured variables (e.g., HbA1c, smoking status, race, alcohol use) cannot be excluded. Multiple outcomes were tested across 16 cancer types without a formal correction for multiplicity, which means some null results may be underpowered and some statistically significant results may warrant additional scrutiny.

Clinical Takeaway

For the clinician managing a patient with type 2 diabetes who has elevated baseline risk for hepatocellular carcinoma or pancreatic cancer, particularly in the context of metabolic-associated steatotic liver disease or obesity, these data add to a growing body of evidence suggesting that GLP-1RA therapy may confer oncologic benefit beyond glycemic control. The association with reduced liver cancer risk is biologically plausible and directionally consistent with prior work from Wang et al. The pancreatic cancer finding is the more striking result numerically and mechanistically; the reported HR of 0.23 is large enough to warrant serious attention, though the absolute event counts (22 cases in the GLP-1RA group vs. 65 in the insulin group) are modest. Neither finding should yet redirect oncologic surveillance protocols or override individualized prescribing decisions. These are hypothesis-generating signals from an observational dataset that demand prospective confirmation.

Clinical Bottom Line: GLP-1 receptor agonists were associated with significantly lower incidence of liver and pancreatic cancer compared with insulin in a large, propensity-weighted cohort of commercially insured adults with type 2 diabetes — an observational signal that warrants prospective investigation but does not yet change screening or treatment guidelines.

Why This Matters Clinically

GLP-1 receptor agonists are now prescribed to tens of millions of patients worldwide for type 2 diabetes, obesity, and cardiovascular risk reduction. Any meaningful cancer signal in this drug class, whether protective or harmful, carries population-level implications that dwarf most single-drug oncology trials. The prior anxiety about thyroid cancer risk, which drove FDA boxed warning language for this class, was grounded largely in rodent C-cell hyperplasia data and has not been replicated in large human cohort studies, including this one. Conversely, the emerging protective signal for liver and pancreatic cancer is biologically anchored: GLP-1RAs reduce hepatic steatosis, modulate the PI3K/Akt pathway, suppress tumorigenic adipokines such as leptin, and may enhance natural killer cell cytotoxicity against hepatocellular carcinoma through IL-6/STAT3 pathway suppression. Pancreatic cancer cells also appear to have reduced or absent GLP-1 receptor expression compared with normal pancreatic cells, which may explain why receptor agonism does not promote pancreatic carcinogenesis in this dataset. These mechanistic threads do not prove causality, but they provide a rational biological framework for the observed associations. The clinical significance of a 77% hazard reduction for pancreatic cancer, a malignancy with a five-year survival rate below 15%, would be extraordinary if replicated in prospective trials.

CED Clinical Relevance

At CED Clinic, patients with type 2 diabetes often present with overlapping metabolic comorbidities, including obesity, dyslipidemia, hypertension, and metabolic-associated steatotic liver disease. These are precisely the patients at elevated baseline risk for both hepatocellular carcinoma and pancreatic adenocarcinoma. When a patient with this metabolic phenotype requires an antidiabetic agent, these data introduce a legitimate secondary consideration: the drug class choice may have cancer-related implications over a multi-year treatment horizon. This does not mean GLP-1RA therapy should be prescribed specifically for cancer prevention — the data do not support that framing. What the data do support is that patients already using GLP-1RAs for established indications can be counseled that the available evidence does not indicate an increased cancer risk for the 16 tumor types evaluated, and may

Physician-Led, Whole-Person Care

A doctor who takes the time to truly understand you.

Personal care that starts with listening and is guided by experience and ingenuity.

Health, Longevity, Wellness

One-on-One Cannabis Guidance

Metabolic Balance

Leave a Message
Metabolic Care
Medical Consulting
Cannabis Care