Reproducibility of mRECIST in Measurement and Response Assessment for Hepatocellular Carcinoma Treated by Transarterial Chemoembolization

Abstract

To evaluate the reproducibility of Modified Response Evaluation Criteria in Solid Tumors (mRECIST) for hepatocellular carcinoma (HCC) after transarterial chemoembolization (TACE) using multiphasic computed tomography. The institutional review board approved this retrospective study. We evaluated 97 patients who underwent TACE (60 conventional TACE [cTACE] and 37 drug-eluting bead TACE [DEB-TACE]) for HCC from 2010 to 2014. Four radiologists evaluated pairs of dynamic liver CTs scanned within 2 months before and after TACE based on mRECIST. Assessment of intra- or interobserver reproducibility for response categorization and sum of long diameter were evaluated using weighted kappa statistics (κ) and intraclass correlation coefficients, respectively. The relationship between concordance of target lesion selection and agreement of target lesion response was evaluated using Fisher exact test. Intraobserver reproducibility for overall response was moderate to excellent (κ = 0.525-0.865). Interobserver reproducibility was improved on the second review compared to the first review and it was good in both treatment groups (κ = 0.627 for cTACE and 0.602 for DEB-TACE). Between the two treatment methods, intra- or interobserver reproducibility was better after cTACE than DEB-TACE. Intraclass correlation coefficients for sum of long diameter measurement showed excellent intra- or interobserver reproducibility. The concordance rate of target lesion selection was significantly higher for patients with radiologists' agreement for target lesion response than patients with disagreed response (P = .003). The intra- and interobserver reproducibility of mRECIST in patients with HCC after TACE was moderate to excellent, and the reproducibility was slightly better after cTACE than DEB-TACE.