Comparative Validation of Prediction Models for HCC Outcomes in Living Donor Liver Transplantation: Superiority of Tumor Markers to Imaging Study.

Abstract

Background: Living donor liver transplantation (LDLT) offers timely curative treatment for unresectable hepatocellular carcinoma (HCC). This study aims to validate and compare previous prediction models for HCC outcomes in 488 LDLT recipients.

Methods: For 488 patients who underwent LDLT for HCC, pretransplant imaging studies assessed by modified RECSIT criteria, tumor markers such as alpha feto-protein (AFP) and protein induced by vitamin K absence or antagonist-II (PIVKA II), and explant pathology were recruited. C-index of models for the HCC outcomes was compared, followed by further investigation for the predictive performances of the best model.

Results: We found MoRAL (11√PIVKA-II + 2√AFP) demonstrated a higher C-index for HCC recurrence than other models that included radiologically viable tumor number and/or size (MoRAL: 0.709, Milan: 0.537, UCSF: 0.575, Up-to-7: 0.572, French AFP: 0.634, Pre-MORAL: 0.637, HALT-HCC: 0.626, Metroticket2.0: 0.629) and also had the highest C-index for HCC-specific deaths (0.706). Five-year HCC recurrence was well stratified upon dividing the patients into three groups by MoRAL cutoffs (11.9% for MoRAL < 100, 29.6% for MoRAL 100-200, and 48.6% for MoRAL > 200, p < 0.001). However, patients with major vessel invasion or portal vein tumor thrombus showed similarly high HCC recurrence regardless of this grouping (p = 0.612).

Conclusion: The MoRAL, based on tumor markers, showed the best predictive performance for HCC recurrence and HCC-specific death among the validated models, except in cases with major vessel invasion or portal vein tumor thrombus.