Imaging Features of Biphenotypic Primary Liver Carcinoma (Hepatocholangiocarcinoma) and the Potential to Mimic Hepatocellular Carcinoma: LI-RADS Analysis of CT and MRI Features in 61 Cases.

Abstract

The purpose of this study was to determine the frequency with which biphenotypic primary liver carcinoma (also called hepatocholangiocarcinoma) may be misclassified as hepatocellular carcinoma (HCC) when only Liver Imaging Reporting and Data System (LI-RADS) major features are used and after consideration of ancillary features. A review of all pathologically proven biphenotypic primary liver carcinomas diagnosed at one institution from 2006 to 2014 was performed. Two subspecialized abdominal imagers independently reviewed cases using LI-RADS version 2014 and assigned major features, ancillary features, and additional findings. The number of lesions meeting imaging criteria for HCC was determined after assessment of major features alone and after the addition of ancillary features. Sixty-one patients (30 men, 31 women; mean age, 62 years; range, 22-89 years) with biphenotypic primary liver carcinomas who underwent pretreatment multiphasic contrast-enhanced MRI (48 patients) or CT (13 patients) were included. According to LI-RADS major features alone, 33 (54.1%) lesions met criteria for HCC and therefore might have been misclassified. Thirteen had arterial phase hyperenhancement, washout, and a capsule. Twenty had arterial phase hyperenhancement with either washout (15 lesions) or a capsule (five lesions). After evaluation of ancillary features, 29 of these potential mimics exhibited at least one ancillary feature favoring non-HCC malignancy, possibly leading to appropriate reclassification. Of the four carcinomas that met criteria for HCC by major features and did not have ancillary features favoring non-HCC malignancy, two (3.3% of all tumors) fell within the Milan criteria. Most biphenotypic primary liver carcinomas have features of non-HCC malignancy and can be correctly categorized as such. Addition of ancillary features to major features may improve diagnostic accuracy over systems in which only major features are used.