Validation of US Liver Imaging Reporting and Data System Version 2017 in Patients at High Risk for Hepatocellular...

Background The 2017 version of the Liver Imaging Reporting and Data System (LI-RADS) recently included standardized interpretation, reporting, and management guidelines for US (US LI-RADS); however, this system has not yet been validated. Purpose To evaluate the diagnostic performance of US LI-RADS version 2017 for detecting hepatocellular carcinoma (HCC) in participants at high risk and to determine the clinical factors associated with a poor visualization score. Materials and Methods This study included 407 prospectively recruited participants (mean age, 56 years; age range, 28-76 years) with cirrhosis at high risk for HCC who underwent US surveillance from November 2011 to August 2012. Two radiologists retrospectively analyzed US images, assigning a LI-RADS category (US-1 = negative, US-2 = subthreshold, US-3 = positive) and a visualization score (A = no or minimal limitations, B = moderate limitations, C = severe limitations). The sensitivity and specificity for diagnosing HCC were calculated on a per-patient and per-lesion basis, using pathologic results and typical CT or MRI as reference standards. The risk factors for a poor visualization score were determined by using univariable and multivariable analyses. Results Of 429 lesions in 407 participants, there were 32 HCCs in 28 participants. In the per-lesion analysis, the specificity for US-3 was 366 of 397 (92%; 95% confidence interval [CI]: 89%, 95%) and the sensitivity was 11 of 32 (34%; 95% CI: 20%, 52%). In the per-patient analysis, the specificity for US-3 was 352 of 379 (93%; 95% CI: 90%, 95%) and the sensitivity was 11 of 28 (39%; 95% CI: 24%, 58%). Visualization score C (114 of 407 [28%] participants) had the highest false-negative rate (six of seven [86%] participants). High body weight (adjusted odds ratio [OR], 2.1 [95% CI: 1.2, 3.6]; P = .01), Child-Pugh class B disease (OR, 2.9 [95% CI: 1.7, 4.9]; P < .001), and moderate to severe fatty liver (OR, 1.7 [95% CI: 1.0, 2.8]; P = .047) were associated with a poor visualization score of C. Conclusion The US-3 category demonstrated high specificity but low sensitivity for diagnosing hepatocellular carcinoma. The visualization score C had a higher false-negative rate than scores A or B, and patients with high body weight, Child-Pugh class B disease, and moderate to severe fatty liver may present limitations for US surveillance. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Milot in this issue.