Does Semaglutide Raise Pancreatic Cancer Risk? What the Nordic Registry Study Can—and Cannot—Rule Out

A large Nordic registry study found no statistically detected increase in pancreatic cancer among semaglutide initiators with type 2 diabetes compared with matched initiators of other glucose-lowering drugs. That is reassuring evidence against a large short-term increase—not proof that the risk is exactly zero.

The regulator-mandated study included 97,464 people who initiated semaglutide in Denmark, Sweden, or Norway. Researchers counted 131 pancreatic cancers over 167,399 person-years in the semaglutide group and 123 over 140,306 person-years among matched active comparators. The pooled hazard ratio was 0.91 (95% CI 0.71–1.16), and the incidence-rate difference was −0.10 (95% CI −0.31 to 0.10) per 1,000 person-years.

The study compared matched treatment initiators across three national registry systems. Its confidence interval crossed the no-difference line, so the result did not establish either increased risk or a protective effect.

Evidence status: These findings were reported in a late-breaking oral presentation at the European Association for the Study of Diabetes (EASD) annual meeting on September 28, 2026. No full paper is available, the work has not been submitted to a medical journal, and the complete methods and results have not yet received peer-reviewed scrutiny.

Quick answer: does semaglutide raise pancreatic cancer risk?

This study found no increased signal, but it cannot rule out every possible increase. The point estimate was slightly below 1, yet the confidence interval ranged from 0.71 to 1.16. In plain language, the data were compatible with anything from a modestly lower relative hazard to a modestly higher one.

The responsible conclusion is narrower than either headline extreme:

• Supported: no statistically detected increase versus the study's matched diabetes-drug comparators during the observed period.
• Not supported: semaglutide prevents pancreatic cancer.
• Not established: the true long-term risk is exactly zero.
• Not studied here: people using semaglutide only for weight management, people outside the three Nordic countries, or all GLP-1 receptor agonists as a class.

What did the Nordic registry study actually find?

Pancreatic cancer was rare in both groups, and the adjusted relative and absolute estimates both crossed their no-difference values.

| Measure | Semaglutide initiators | Active comparators | Comparison estimate | |---|---:|---:|---:| | Participants after matching | 97,464 | 97,464 | Matched 1:1 | | Pancreatic cancers | 131 | 123 | — | | Follow-up | 167,399 person-years | 140,306 person-years | — | | Crude rate calculated from reported counts | 0.78 per 1,000 person-years | 0.88 per 1,000 person-years | — | | Pooled adjusted hazard ratio | — | Reference | 0.91 (95% CI 0.71–1.16) | | Pooled adjusted incidence-rate difference | — | Reference | −0.10 (95% CI −0.31 to 0.10) per 1,000 person-years |

The crude rates above simply divide reported cancers by reported person-time. The pooled hazard ratio and incidence-rate difference came from country-specific adjusted estimates combined across Denmark, Sweden, and Norway.

Country-level hazard ratios also did not show a statistically detected increase: 1.00 (0.66–1.51) in Denmark, 0.82 (0.56–1.22) in Sweden, and 0.91 (0.55–1.51) in Norway. Conference materials further reported no clear cumulative dose-response or duration-response pattern, although those estimates were imprecise.

How was the study designed?

This was an active-comparator, new-user observational cohort study using nationwide prescription, patient, and cancer registries—not a randomized trial.

Eligible adults had type 2 diabetes and newly initiated either semaglutide or one of three non-incretin comparator categories:

1. sulfonylureas;
2. SGLT2 inhibitors; or
3. insulin.

Participants needed at least two prescriptions for the initiated treatment, with the second filled within the first year. Follow-up for pancreatic cancer began after a one-year lag. That lag reduces the chance of counting cancers already present or diagnosed immediately after treatment initiation as newly caused by the drug.

Researchers used propensity scores to match semaglutide initiators with people who started comparator drugs. National prescription registries supplied exposure information, while cancer and patient registries supplied outcomes and measured covariates.

This design is stronger than comparing semaglutide users with untreated people. Active comparators help align treatment stage and healthcare contact, while a new-user design avoids mixing long-standing users with recent starters. Matching can improve balance on recorded characteristics.

It still does not create random assignment.

Why doesn't “not statistically significant” mean “zero risk”?

A null result means the study did not detect a difference of the size and precision required by its analysis; it does not prove the underlying effect is exactly zero.

The pooled HR of 0.91 is a point estimate. If treated as a relative range, its 95% confidence interval remains compatible with approximately a 29% lower hazard or a 16% higher hazard compared with the active comparators. The absolute interval likewise runs from 0.31 fewer to 0.10 more pancreatic cancers per 1,000 person-years.

That range excludes a large short-term increase more convincingly than it excludes a small one. It also prevents a prevention claim: the interval includes no difference, and an observational point estimate below 1 can reflect bias, comparator choice, or chance.

Pancreatic cancer is uncommon. Rare outcomes generate fewer events, which makes estimates less precise even in a cohort approaching 200,000 matched participants.

What are the study's main limitations?

The biggest limitation is time: mean follow-up after the one-year lag was only 1.40 to 1.85 years across countries. Pancreatic cancer can develop over a long period, so short post-lag observation cannot settle long-latency risk.

Other limits matter too:

Prescription records are not direct adherence measures

Two filled prescriptions improve confidence that treatment was started, but dispensing data cannot prove how consistently a medicine was taken or capture every treatment interruption. Exposure classification may therefore differ from actual biological exposure.

Matching only addresses recorded factors

Propensity-score matching can balance measured characteristics. It cannot balance variables that were absent, misclassified, or poorly captured. Residual confounding may remain from diabetes duration and severity, obesity, smoking, pancreatitis history beyond the study's exclusions, healthcare use, socioeconomic conditions, or other clinical differences.

Comparator choice can change the estimate

Sulfonylureas, SGLT2 inhibitors, and insulin are used in different clinical circumstances. The study chose them because they can be used at a similar treatment stage, but they are not interchangeable populations. Differences in illness severity, contraindications, prescribing preferences, and competing health risks can persist after matching.

Conference reporting limits scrutiny

Without a full manuscript and supplement, readers cannot inspect every model specification, missing-data decision, code list, balance diagnostic, sensitivity analysis, subgroup, or protocol deviation. An oral presentation is useful early evidence; it is not the same evidentiary object as a peer-reviewed paper.

How does this differ from randomized-trial meta-analysis?

The Nordic study asks what happened after real-world treatment initiation; randomized-trial meta-analysis pools cancers observed in trials where treatment assignment was randomized. Each approach solves a different problem.

A 2024 meta-analysis of 90 randomized trials involving 124,791 participants found no significant association between GLP-1 receptor agonists and pancreatic cancer: RR 1.05 (95% CI 0.77–1.43). Randomization reduces confounding by treatment selection. But pancreatic cancer was not the primary outcome of most trials, the analysis covered the GLP-1 class rather than this exact Nordic semaglutide question, and average follow-up was 3.1 years—still limited for a rare cancer with potentially long latency.

The Nordic registry study contributes semaglutide-specific, routine-care evidence and more pancreatic cancer events. The randomized meta-analysis contributes better protection against confounding. Agreement between them is reassuring, but agreement does not convert either into proof of no long-term risk.

How does this differ from spontaneous adverse-event signals?

Spontaneous-report databases can raise safety questions, but they cannot calculate incidence or prove that a drug caused an event. Reports may be incomplete, duplicated, stimulated by publicity, or missing critical clinical context. They also lack a clean denominator showing how many exposed people did not develop the event.

That makes spontaneous reports useful for signal detection—not for estimating semaglutide pancreatic cancer risk. The Nordic study improves on that by defining exposed and comparator cohorts, counting person-time, and linking to national cancer registries. It remains observational, but it can estimate comparative rates in a way a pile of voluntary reports cannot.

For a broader guide to these source types, see GLP-1 Side Effects: What Trials, Labels, and Online Communities Can Each Tell Us and How to Evaluate Peptide Claims Online.

Who do these findings apply to?

The direct evidence applies to adults treated for type 2 diabetes in Denmark, Sweden, and Norway who met the study criteria. It should not be generalized automatically to:

• all people who use GLP-1 receptor agonists;
• people taking semaglutide only for weight management;
• younger, higher-risk, or medically different populations excluded from the cohort;
• healthcare systems with different prescribing, screening, and registry practices; or
• follow-up extending many years beyond the study window.

Semaglutide is a regulated GLP-1 medicine, which is different from an unapproved product merely marketed as a “GLP-1 peptide.” GLP-1 Peptides vs Research Peptides explains why product and regulatory identity matter when reading safety claims.

What is the most defensible conclusion?

The Nordic study adds meaningful evidence against a large short-term pancreatic cancer increase with semaglutide in the studied diabetes population. It does not demonstrate prevention, prove zero risk, or settle long-term risk.

Its strengths are substantial: nationwide registries, a new-user design, active comparators, propensity matching, a one-year lag, country-specific analyses, and both relative and absolute estimates. Its limits are equally important: short follow-up after the lag, a rare and slow-developing outcome, exposure measured through prescription records, residual confounding, comparator differences, and conference-only reporting.

The study was required by regulators and funded by Novo Nordisk. Those facts should be disclosed because they are relevant context. They do not invalidate the analysis, and they do not replace scrutiny of the eventual full methods, results, and peer review.

Frequently asked questions

Did the study prove semaglutide does not cause pancreatic cancer?

No. It found no statistically detected increase compared with matched initiators of other diabetes drugs during relatively short follow-up. Its confidence interval and design do not prove zero risk.

Did semaglutide prevent pancreatic cancer in this study?

No. The point estimate was below 1, but the confidence interval included no difference. An observational null result cannot support a protective-cancer claim.

Was this a randomized clinical trial?

No. It was an observational active-comparator, new-user cohort study using national registry data and propensity-score matching.

Has the study been peer reviewed?

No full paper is available, and the work has not been submitted to a medical journal. The results were presented orally at EASD on September 28, 2026, so full peer-reviewed scrutiny is not yet possible.

Does the study apply to semaglutide used for weight management?

Not directly. The cohort consisted of people treated for type 2 diabetes in Denmark, Sweden, and Norway.

Sources

1. European Association for the Study of Diabetes annual meeting late-breaking oral presentation, September 28, 2026; results and presenter quotations reproduced in the EASD conference media release: “New registry data shows semaglutide does not cause pancreatic cancer”. Conference-only evidence; no full paper available.
2. European Medicines Agency, HMA-EMA Catalogue of Real-World Data Studies: EUPAS37258 / NN9535-4447 administrative details.
3. ClinicalTrials.gov: NCT04572165, Epidemiological Assessment of the Risk for Pancreatic Cancer Associated With the Use of Semaglutide.
4. Figlioli G, et al. “Glucagon-like peptide-1 receptor agonists and risk of gastrointestinal cancers: A systematic review and meta-analysis of randomized controlled trials.” Pharmacological Research. 2024;208:107401. doi:10.1016/j.phrs.2024.107401.
5. U.S. Food and Drug Administration. FDA Adverse Event Monitoring System (AEMS) Public Dashboard.

This article is for general education. It does not provide diagnosis, cancer screening, treatment selection, medication-starting or stopping guidance, dosing, or individualized medical advice.