Pancreatic cysts stable for five years rarely become cancer

Design

Single-centre retrospective cohort using language-model extraction from reports

Population

13,269 patients with pancreatic cystic lesions, 2000-2025

Primary outcome

5-year malignancy prediction by guideline; risk after 5 years of stability

Effect

AUROC 0.62-0.66; 8-year incidence 0.46% when Kyoto stopping criteria met

This single-centre US study analysed abdominal radiology reports from 2000 to 2025 for 13,269 patients with pancreatic cystic lesions, median age 69. A previously validated large language model extracted cyst features and assigned risk categories under ACR, Kyoto and European guidelines. Patients with a cancer within six months were excluded.

Over five years, 209 (1.6%) developed pancreatic cancer or high-grade dysplasia. All three guidelines predicted this only modestly (AUROC 0.655 for Kyoto, 0.639 for ACR, 0.623 for European). They agreed completely on which cysts were high risk, but varied widely in how many they labelled worrisome (308 to 507), with similar cancer rates in each worrisome group. Among 1,830 patients whose cysts were stable for five years, eight-year cancer incidence was 0.71%, 0.46% in those meeting Kyoto criteria for stopping surveillance and 1.57% in those who did not.

This supports discussing an end to surveillance for low-risk cysts after five years of stability, especially in older patients. It is retrospective, single-centre and based on report text rather than re-review of images. Current guidelines differ on stopping, and none changes on this study alone.

• Report cyst size, main duct calibre, mural nodules and growth explicitly, so risk can be assigned.
• Consider recommending an end to surveillance for low-risk cysts stable for five years, where local guidance allows.
• Continue surveillance when Kyoto stopping criteria are not met; risk was about three times higher.
• Weigh age and comorbidity: surveillance has least value where life expectancy is limited.

Why it matters

Lifelong cyst surveillance consumes imaging capacity and causes anxiety for a very low yield.

Don't overread it

Retrospective single-centre data extracted from reports by a language model, not independently re-reviewed images.

The statistics, in plain English

An AUROC of 0.65 means the guidelines ranked a future cancer above a non-cancer only about two times in three, little better than chance at 0.5. Eight-year incidence of 0.46% means about 1 in 220 stable, low-risk patients developed cancer.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.