Authors
Serkan Sucu, Aditya V Kotla, Ethan Angle, Andy Tran, Hassan Aziz
Published in
Surgery. Pages 110650. Sep 10, 2026. Epub Sep 10, 2026.
Abstract
Patients undergoing liver transplantation for hepatocellular carcinoma beyond the Milan criteria represent an expanding population with heterogeneous outcomes. We developed and validated a point-based score (post-transplant hepatocellular carcinoma score) to stratify 5-year mortality risk using readily available clinical and pathologic variables analyzed via machine learning.
We analyzed hepatocellular carcinoma liver transplant recipients from the United Network for Organ Sharing database (2012-2019) with determinable 5-year outcomes whose explant pathology exceeded Milan criteria. A gradient-boosting classifier identified key predictors of mortality. Based on feature importance, we developed a 0-5 point score using the post-transplant-hepatocellular carcinoma components. Risk categories were low (score 0), moderate (score 1-2), and high (score 3-5).
Among 2,952 patients, 812 (27.5%) died within 5 years of transplantation. The post-transplant hepatocellular carcinoma score stratified patients into 3 risk groups, with 5-year survival of 82.7% (low), 72.2% (moderate), and 50.2% (high). Survival decreased monotonically with score, from 82.7% at score 0 to 39.0% at score 4. In the temporally independent 2019 validation cohort, the score demonstrated an area under the curve of 0.619 (95% confidence interval, 0.554-0.678), with 5-year survival of 84.0%, 72.2%, and 47.1% in the low-, moderate-, and high-risk groups, respectively. The score performed similarly to the Risk Estimation of Tumor Recurrence After Transplant score (area under the curve, 0.619 vs 0.629; P = .72).
The post-transplant hepatocellular carcinoma score is a simple, validated tool for stratifying 5-year mortality risk in hepatocellular carcinoma transplant recipients beyond the Milan criteria and may assist in surveillance planning, patient counseling, and clinical decision-making pending external validation.
PMID:
42816225
Bibliographic data and abstract were imported from PubMed on 01 Oct 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 22
- Comments 0