Inclusion of tumor markers improves the correlation of the Milan criteria with vascular invasion and tumor cell differentiation in patients with hepatocellular carcinoma undergoing liver resection (#JGSU-D-07-00462).

The currently used criteria, such as the Milan criteria, to select a candidate of liver transplantation for HCC consists of size and number of tumors because vascular invasion and poor differentiation, the strongest prognostic factors, are difficult to be assessed preoperatively. We hypothesized that inclusion of two tumor markers (alpha-fetoprotein and des-gamma-carboxy prothrombin) into the criteria would increase the prediction accuracy of these factors. Our hypothesis was tested in 478 HCC patients undergoing liver resection. The models with or without markers, constructed at predicting vascular invasion (n=150) or poor differentiation (n=49), were compared. The model including markers was superior at predicting the absence of vascular invasion to either the Milan criteria alone [at 81.2% sensitivity; specificity, 52.4 vs 43.3%; difference, 9.1%(95% CI, 1.3-14.2%)] or a model in which size and number varied freely [AUCs of receiver operating characteristic curves, 75.2 vs 69.1%; difference, 6.1%(2.33-10.7%)]. The model incorporating markers was also superior at predicting well to moderate differentiation to either the Milan criteria [at 74.5% sensitivity; specificity, 57.1 vs 38.8%; difference, 18.3%(2.4-32.7%)] or a model with size and number [AUCs, 71.5 vs 59.0%; difference, 12.5%(5.84-21.4%)]. In conclusion, the tumor marker levels should be considered when selecting patients with HCC for liver transplantation.