# Long-Term Prognostic Value of AFP and PIVKA-II in HCC After Living Donor Liver Transplantation: A Single-Center Retrospective Study

## TL;DR

This study compares the usefulness of tumor biomarkers AFP and PIVKA-II with other models in predicting outcomes for liver transplant patients with hepatocellular carcinoma.

## Contribution

The study evaluates the long-term prognostic value of AFP and PIVKA-II in HCC patients post-liver transplant compared to six existing models.

## Key findings

- AFP and PIVKA-II achieved iAUCs of 0.68–0.75 for recurrence and C-indices of 0.66–0.77 for mortality.
- Combining AFP and PIVKA-II improved performance with iAUCs up to 0.78 and C-indices up to 0.80.
- The R3-AFP model showed the most consistent performance with iAUC 0.76–0.81 and C-index 0.78–0.82.

## Abstract

Despite the development of numerous prognostic models for hepatocellular carcinoma (HCC) recurrence and mortality after liver transplantation, tumor biomarkers such as alpha-fetoprotein (AFP) and protein induced by vitamin K absence-II (PIVKA-II) remain widely used in clinical practice. This study evaluated the performance of AFP and PIVKA-II compared with six prognostic models (RETREAT, SNAPP, MoRAL, R3-AFP, METROTICKET 2.0, and SALT) in a retrospective cohort of 707 adults who underwent living donor liver transplantation (LDLT) for HCC between 2003 and 2018. Patients were stratified into Milan and Beyond Milan groups. Time-dependent receiver operating characteristic curve analysis was conducted using integrated area under the curve (iAUC) and concordance index (C-index) to assess recurrence and mortality. AFP and PIVKA-II (continuous) achieved iAUCs of 0.68–0.75 for recurrence and C-indices of 0.66–0.77 for mortality. Their combination reached iAUCs up to 0.78 and C-indices up to 0.80. Threshold models (AFP ≥200, PIVKA-II ≥400) showed modest predictive performance. Among multivariable models, R3-AFP demonstrated the most consistent performance (iAUC 0.76–0.81; C-index 0.78–0.82). SNAPP, MoRAL, and SALT also performed well. AFP and PIVKA-II may offer practical utility in resource-limited settings. However, multivariable models remain the preferred approach where comprehensive diagnostics are available.