Résumé
AIM: To generate a new score with improved accuracy compared with Milan criteria to select patients. PATIENTS: The training cohort comprised 373 patients transplanted for hepatocellular carcinoma (HCC) between 1988 and 1998 (cohort 1). An algorithm was derived from the analysis of patient data by the proportional hazard Cox regression model. The area under the receiver operating characteristic (AUROC) was used to determine a cut-off value. The validation cohort comprised 140 patients transplanted between 1999 and 2001 (cohort 2). RESULTS: Multivariate analysis identified three predictors of 5-year tumour-free survival: tumour differentiation (P=0.02), diameter (P<0.0001) and number of nodules (P=0.04). A cut-off value of 4 was derived from the AUROC of the final score. Five-year tumour-free survival was 60.2 ± 3.1% in patients with as score <4 and 36.4 ± 4.7% in individuals with a score ≥4, P<0.0001. In the validation cohort, 5-year tumour-free survival was 82.8 ± 3.6% (score <4) and 50.0 ± 10.7% (score ≥4), P=0.0003. In patients with a score <4, there was no significant difference in 5-year tumour-free survival between Milan+ and Milan- patients, either in cohort 1 or 2. Five-year tumour-free survival of Milan- patients was significantly better in individuals with a score <4 compared with those with a score ≥4, both in cohort 1 (61.5 ± 9.1 vs 31.4 ± 4.6%, P=0.009) and in cohort 2 (P=0.02). CONCLUSION: A novel score taking into account tumour differentiation shows higher accuracy than Milan criteria in predicting 5-year tumour-free survival following liver transplantation for HCC. Prospective studies should validate these findings.