Evaluation of the LI-RADS treatment response algorithm in hepatocellular carcinoma after trans-arterial chemoembolization

Purpose: To evaluate the diagnostic performance of LI-RADS treatment response algorithm (LR-TRA) and modified RECIST (mRECIST) for the detection of viable hepatocellular carcinoma (HCC) on MRI after trans-arterial chemoembolization (TACE).

Materials and methods: This retrospective study includes cirrhotic patients that underwent trans-arterial chemoembolization prior to liver transplantation from 2013 to 2017 with a pre- and post-treatment MRI available. Three blinded readers assigned a LR-TRA and mRECIST category to each lesion. Lesions on MRI and explant pathology were matched and characterized as complete (100% necrosis) or incomplete necrosis (≤99% necrosis). Diagnostic performance of LR-TRA and mRECIST were calculated with a generalized estimating equation.

Results: A total of 52 patients with 71 lesions were included, 47 with incomplete and 24 with complete necrosis. In consensus, 45 lesions were categorized as LR-TR Nonviable, of which 62.2% (28/45) had incomplete and 37.8% (17/45) had complete necrosis. Six lesions were categorized as LR-TR Equivocal, of which 33.3% (2/6) had incomplete and 66.7% (4/6) had complete necrosis. Twenty lesions were categorized as LR-TR Viable of which 85.0% (17/20) had incomplete and 15.0% (3/20) had complete necrosis. The sensitivity of LR-TR Viable for detecting incompletely necrotic tumor when LR-TR Equivocal was considered as viable, in consensus was 40.4%; specificity 70.8%; accuracy 50.7%. The sensitivity of mRECIST for detecting incompletely necrotic tumor was 37.0%; specificity 79.2%; accuracy 51.4%. There was no significant difference in diagnostic performance between mRECIST and LR-TRA (p = 0.14-0.33). Agreement for LR-TRA category was moderate (k = 0.53 [95% CI: 0.45, 0.67]).

Conclusion: LI-RADS treatment response algorithm demonstrates high specificity and low to moderate sensitivity for the detection of viable HCC after TACE in a North American cirrhotic cohort, without significant difference in diagnostic performance between LR-TRA and mRECIST.