Prediction of hepatocellular carcinoma recurrence after liver transplantation: Comparison of four explant-based prognostic models.

AIM: Discordance between pre-LT imaging and explanted liver findings have been reported after liver transplantation (LT) for hepatocellular carcinoma (HCC), suggesting the need of reassessing the risk of HCC recurrence post-LT. Our aims were to compare pre-LT imaging and explants features and to test the performances of four explant-based predictive models of recurrence in an external cohort.

METHODS: Staging according to pre-LT imaging and explant features were compared. Four explants-based models were retrospectively tested in a cohort of 372 patients transplanted for HCC in 19 French centres between 2003 and 2005. Accuracies of the scores were compared.

RESULTS: Pre-LT imaging underestimated tumour burden in 83 (22.7%) patients according to Milan criteria. The highest AUCs for prediction of 5-years recurrence were observed in the "Up to seven" (0.7915 [95% CI: 0.7339-0.849]) and Decaens models (0.747 [95% CI: 0.6877-0.806]), with two levels of risk: low (10%) and high (>50%). Chan and Iwatsuki models identified 3 and 4 levels of risk, but had lower AUCs (0.68 and 0.70) respectively. Accuracy of the "Up to seven" model was superior to the Decaens model (P=.034), which was superior to the Chan model (P=.0041) but not to the Iwatsuki model (P=.17).

CONCLUSION: Pre-LT imaging underestimates tumour burden, and prediction of recurrence should be reassessed after LT. The explant-based "Up to seven" and Decaens models provided the best accuracy for prediction of 5-year recurrence, identifying only two levels of risk. New models are needed to further refine the prediction of recurrence after LT.