Abstract

Explant-based models for assessing hepatocellular carcinoma (HCC) recurrence post-liver transplantation (LT) serve as the gold standard, guiding post-LT screening and immunosuppression adjustment. Incorporating alpha-fetoprotein (AFP) levels into these models, such as the novel R3-AFP score, has notably enhanced risk stratification. However, validation of these models in high-evidence data is mandatory. Therefore, the aim of the present research was to validate the R3-AFP score in a randomized clinical trial. We analyzed the intention-to-treat (ITT) population from the 2-arm SiLVER trial (NCT00355862), comparing calcineurin-based (CNI-Group A) versus mammalian target of rapamycin inhibitors (mTOR)-based (Sirolimus-Group B) immunosuppression for post-LT HCC recurrence. Competing risk analysis estimated sub-hazard ratios (SHR), with testing of discriminant function and calibration. Overall, 508 patients from the ITT analysis were included (Group A, n=256; Group B, n=252). The R3-AFP score distribution was as follows: 42.6% low risk (n=216), 35.7% intermediate risk (n=181), 19.5% high risk (n=99), and 2.2% very high risk (n=11) groups. The R3-AFP score effectively stratified HCC recurrence risk, with increasing risk for each stratum. Calibration of the R3-AFP model significantly outperformed other explant-based models (Milan, Up-to-7, and RETREAT), while discrimination power [0.75 (95% CI 0.69;0.81)] surpassed these models, except for the RETREAT model (p=0.49). Sub-group analysis showed lower discrimination power in the mTOR group versus the CNI group (p=0.048). In conclusion, the R3-AFP score accurately predicted HCC recurrence using high-quality evidence-based data, exhibiting reduced performance under mTOR immunosuppression. This highlights the need for further research to evaluate surveillance schedules and adjuvant regimens.