# Pilot study of an arterial enhancement fraction-based model for progression prediction in HCC undergoing chemoembolization

## TL;DR

A new model called ADMAN uses arterial enhancement fraction and other factors to predict cancer progression in HCC patients after a specific treatment.

## Contribution

The ADMAN model, incorporating arterial enhancement fraction of residual tumor, shows improved prediction of progression-free survival in HCC patients.

## Key findings

- The ADMAN model outperformed existing models with a C-index of 0.75 in the training cohort.
- High-risk patients had 4.69 times greater progression risk than low-risk patients in the training cohort.
- The model showed significantly lower Brier scores at 6 and 12 months in both training and validation cohorts.

## Abstract

To develop a prognostic model including arterial enhancement fraction of residual tumor (AEF-RT) for predicting progression-free survival (PFS) in hepatocellular carcinoma (HCC) patients after drug-eluting beads transarterial chemoembolization (DEB-TACE).

Between March 2019 and March 2024, 111 HCC patients undergoing DEB-TACE were randomly allocated to a training cohort and a validation cohort in a 7:3 ratio. LASSO regression was applied in the training cohort to identify risk factors for recurrence, which were subsequently used to construct the Cox model. Model performance was assessed using the concordance index (C-index, where 0.5 indicates non-informative discrimination and 1 represents perfect discrimination) and Brier score (ranging from 0 to 1, 0 indicating higher calibration) and was compared with five existing prognostic models.

The final model, termed ADMAN model, incorporated AEF-RT, Diameter, Margin appearance, Aspartate transaminase, and Neutrophil-to-lymphocyte ratio. High-risk patients defined by ADMAN had 4.69 times greater progression risk than low-risk ones in the training cohort (p < 0.001) and 3.52 times greater in the validation cohort (p = 0.005). The C-index of ADMAN (0.75) was significantly higher than that of other models in the training cohort (p < 0.05 for all) and remained significantly higher than three of them in the validation cohort [0.71 vs. 0.55 (p = 0.041), 0.54 (p = 0.033), 0.53 (p = 0.004)]. The ADMAN model showed a significantly lower Brier score than that of other models at 6 months and 12 months in the training cohort (p < 0.05 for all). In the validation cohort, the ADMAN model remained to have significantly lower Brier score than the four models (p < 0.05) at 6 months, while it had significantly lower score than one model at 12 months.

The AEF-based model may be a promising tool for progression risk stratification in HCC patients undergoing DEB-TACE. Further external validation in independent cohorts with larger sample sizes is necessary to confirm the robustness of the ADMAN model.