Pancreatic cyst guidelines agree on high risk, and stable low-risk cysts rarely turn

Design

Retrospective single-centre cohort using language-model extraction from reports

Population

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

Primary outcome

Five-year pancreatic malignancy; risk after five years' stability

Effect

AUROC 0.655 (Kyoto); 8-year incidence after stability 0.46% meeting Kyoto stop criteria

This single-centre study used a validated language model to extract pancreatic cyst features from 25 years of abdominal imaging reports, then applied the Kyoto, ACR and European guidelines to 13,269 patients. Over five years, 209 (1.6%) developed pancreatic cancer or high-grade dysplasia.

All three guidelines predicted malignancy only modestly from imaging features alone (AUROC 0.66 Kyoto, 0.64 ACR, 0.62 European). They agreed completely on which lesions were high risk, but disagreed widely on 'worrisome' features (308 to 507 patients), with similar cancer rates in each worrisome group (10–12%). In 1,830 patients whose cysts had been stable for five years, eight-year cancer incidence was 0.71%, falling to 0.46% in those meeting Kyoto criteria to stop surveillance and 1.57% in those who did not.

This is retrospective and single-centre, but it supports ending imaging surveillance for low-risk cysts after five years of stability where the patient meets criteria — a decision that saves repeated MRI for many older patients.

• Consider ending surveillance for low-risk pancreatic cysts stable for five years when Kyoto criteria are met.
• Continue surveillance for stable cysts not meeting those criteria; their risk was about three times higher.
• Any guideline will identify high-risk features consistently — act on them promptly.
• Worrisome-feature categories differ by guideline; follow one guideline consistently within a service.
• Include the stop-surveillance criteria in the report recommendation when they apply.

Why it matters

Indefinite cyst surveillance costs scanner time and patient anxiety for very little yield after five stable years.

Don't overread it

A retrospective single-centre study using report text; features were not re-read from images.

The statistics, in plain English

An AUROC of 0.5 is a coin toss and 1.0 is perfect; 0.62–0.66 is modest. An eight-year incidence of 0.46% means fewer than 1 in 200 patients developed malignancy after five stable years.

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.