Abstract

Background and Aims: This study evaluates the performance of various magnetic resonance imaging (MRI) response criteria for the prediction of complete pathologic necrosis (CPN) of hepatocellular carcinoma (HCC) post locoregional therapy (LRT) using explant pathology as a reference. Methods: We included 61 patients (male/female 46/15; mean age 60 years) who underwent liver transplantation after LRT with transarterial chemoembolization plus radiofrequency or microwave ablation (n = 56), or (90)Yttrium radioembolization (n = 5). MRI was performed 90 days before liver transplantation. Three independent readers assessed the following criteria: RECIST, EASL, modified RECIST (mRECIST), percentage of necrosis on subtraction images, and diffusion-weighted imaging (DWI), both qualitative (signal intensity) and quantitative (apparent diffusion coefficient [ADC]). The degree of necrosis was retrospectively assessed at histopathology. Intraclass correlation coefficient (ICC) and Cohen's kappa were used to assess inter-reader agreement. Logistic regression and receiver operating characteristic analyses were used to determine imaging predictors of CPN. Pearson correlation was performed between imaging criteria and pathologic degree of tumor necrosis. Results: A total of 97 HCCs (mean size 2.3 +/- 1.3 cm) including 28 with CPN were evaluated. There was excellent inter-reader agreement (ICC 0.77-0.86, all methods). EASL, mRECIST, percentage of necrosis and qualitative DWI were all significant (p 0.001) predictors of CPN, while RECIST and ADC were not. EASL, mRECIST and percentage of necrosis performed similarly (area under the curves [AUCs] 0.810-0.815) while the performance of qualitative DWI was lower (AUC 0.622). Image subtraction demonstrated the strongest correlation (r = 0.71-0.72, p 0.0001) with pathologic degree of tumor necrosis. Conclusions: EASL/mRECIST criteria and image subtraction have excellent diagnostic performance for predicting CPN in HCC treated with LRT, with image subtraction correlating best with pathologic degree of tumor necrosis. Thus, MR image subtraction is recommended for assessing HCC response to LRT. (C) 2017 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved.