Predicting cholangiocarcinoma in primary sclerosing cholangitis: using artificial intelligence, clinical and...

Area of Science:

• Hepatology
• Gastroenterology
• Oncology

Background:

• Primary sclerosing cholangitis (PSC) patients face an elevated risk of developing cholangiocarcinoma (CCA).
• Accurate prediction models for CCA in PSC are essential for timely intervention and improved patient outcomes.

Purpose of the Study:

• To identify clinical and laboratory risk factors for CCA development in PSC patients.
• To develop and validate artificial intelligence (AI)-based predictive models for CCA in PSC.

Main Methods:

• Analysis of a large cohort (1,459 patients) of PSC patients using Cox models.
• Exploration of plasma bile acid (BA) levels for CCA prediction in a subset (300 patients).
• Application of statistical and AI approaches for CCA risk prediction.

Main Results:

• Prolonged inflammatory bowel disease (IBD) duration was a key risk factor for CCA.
• Clinical/laboratory variables predicted CCA with C-indexes of 0.68-0.71, outperforming existing PSC risk scores.
• Bile acids, particularly specific fractions and ratios, also showed predictive power for CCA.

Conclusions:

• AI-based predictive models for CCA in PSC significantly outperform traditional risk scores.
• Identified clinical, laboratory, and bile acid markers contribute to CCA risk prediction in PSC.
• Further data modalities are needed for the clinical implementation of these advanced predictive models.