Optimal criteria for hepatocellular carcinoma diagnosis using CT in patients undergoing liver transplantation

Objective: To compare the diagnostic performance of various guidelines for hepatocellular carcinoma (HCC) diagnosis using computed tomography (CT) in patients undergoing liver transplantation (LT). Methods: In total, 216 patients who underwent preoperative CT and subsequent LT were included. Two radiologists retrospectively evaluated focal hepatic lesions independently according to various guidelines and allocated patients according to the Milan criteria. The diagnostic performance of the guidelines was compared using alternative free-response receiver-operating characteristics (AFROC) analysis with bootstrapping. Comparisons of sensitivity, specificity, and accuracy of patient allocation based on the Milan criteria between guidelines were performed using logistic regression with generalized estimating equations (GEE). Results: Fifty-two of 216 patients had 87 HCCs. The reader-averaged figure of merit obtained using AFROC analysis was 0.738 for the AASLD/EASL or KLCSG-NCC guidelines and 0.728 for the LI-RADS v2014 or OPTN/UNOS (bootstrapping, p = 0.005). The per-lesion sensitivity for HCCs (all and 1–2-cm lesions) was significantly higher with the AASLD/EASL (37.9–41.4% and 30.8–41.0%) than with LI-RADS (28.7% and 15.4–18.0%) (logistic regression with GEE, p = 0.008 and 0.030 for reader 1 and p = 0.005 for reader 2). The per-patient specificity (98.8–99.4%) was the same for all guidelines. The accuracy of the Milan criteria was 81.5–83.3% without significant differences among the four guidelines (logistic regression with GEE, p > 0.05). Conclusion: AASLD/EASL showed higher diagnostic performance and sensitivity, particularly for 1–2-cm HCCs, and the same specificity with LI-RADS. All guidelines are comparable for patient allocation based on the Milan criteria for LT. Key Points: • The overall diagnostic performance of CT for HCC diagnosis was highest with AASLD/EASL. • AASLD/EASL showed higher sensitivity for diagnosis of 1–2-cm HCCs than LI-RADS. • The accuracy of the Milan criteria using CT was comparable among the four guidelines.