New criterion for patient selection for transplantation in hepatocellular carcinoma against the background of liver cirrhosis

Aim . To develop and justify a new criterion for selecting patients for liver transplantation in cases of hepatocellular carcinoma against the background of cirrhosis in order to improve the accuracy of recurrence prediction and ensure clinically acceptable five-year disease-free and overall survival rates. Materials and methods . A retrospective analysis was conducted on the results of liver transplantation in 69 patients with cirrhotic hepatocellular carcinoma. The size of the largest tumor, the number of tumors, alpha-fetoprotein level and the maximum LI-RADS score were considered to determine the new selection criteria for patients. New criteria termed “5-6-250/LI-RADS” were developed and compared with regard to their effectiveness with the Milan criteria and the 5-5-500 rule. Sensitivity, specificity, F1-score, C-index, and overall and disease-free survival were calculated. Results . The proportion of patients meeting the criteria of 5-6-250/LI-RADS, 5-5-500, and the Milan criteria accounted for 65.2%, 60.9%, and 36.2%, respectively. The sensitivity, specificity, C-index, and F1-score for the 5-6-250/LI-RADS were found to be 78%, 100%, 0.89, and 0.87; for the 5-5-500 rule – 78%, 88%, 0.83, and 0.82; for the Milan criteria – 89%, 65%, 0.77, and 0.79, respectively. The five-year disease-free survival rates amounted to 85%, 86%, and 91%, and the five-year overall survival rates accounted for 76%, 79%, and 78%. Conclusion . The study confirmed the potential of the 5-6-250/LI-RADS rule in enhancing patient selection for liver transplantation in cases of hepatocellular carcinoma. The integration of LI-RADS with traditional parameters demonstrated high sensitivity, specificity, F1-score, and C-index values. This contributes to a more accurate assessment of recurrence risk and expands indications for transplantation. The accuracy and automation of the LI-RADS classification can be improved through increasing the sample, applying machine learning methods, and validation on real data.