Machine learning improves post-transplantation HCC recurrence prediction.

Area of Science:

• Hepatobiliary Surgery
• Transplantation Medicine
• Oncology

Background:

• Hepatocellular carcinoma (HCC) recurrence after liver transplantation (LT) remains a significant clinical challenge.
• Accurate prediction of post-transplant recurrence is crucial for patient management and resource allocation.

Purpose of the Study:

• To develop and validate advanced machine learning (ML) models for enhanced prediction of HCC recurrence post-LT.
• To identify novel risk factors contributing to post-transplant HCC recurrence.

Main Methods:

• Utilized the United Network for Organ Sharing (UNOS) database for adult HCC patients undergoing LT (2015-2018).
• Evaluated over 50 clinical, radiographic, laboratory, and explant pathology variables.
• Employed recursive feature elimination and Gradient Boosting Survival and Random Survival Forest algorithms.
• Compared ML model performance against the established Risk Estimation of Tumor REcurrence After Transplant (RETREAT) Score.

Main Results:

• The Gradient Boosting Survival model achieved a C-index of 0.73, outperforming the RETREAT Score (C-index 0.70).
• Key predictors included explant tumor burden score, pre-transplant alpha-fetoprotein (AFP), AFP slope, microvascular invasion, and tumor differentiation.
• A preoperative ML model (C-index 0.69) incorporating AFP, tumor burden, and Albumin-Bilirubin (ALBI) Grade changes also demonstrated predictive capability.

Conclusions:

• Developed a novel ML model that surpasses current clinical tools in predicting post-LT HCC recurrence.
• The model enables improved risk stratification for post-transplant surveillance and adjuvant therapy decisions.
• Pre-transplant ML models can aid in refining eligibility for LT, complementing existing criteria like the Milan Criteria.