Extra-hepatic sarcoma metastasis surveillance in the liver: is arterial phase imaging necessary?

Abstract

To assess the value of arterial phase imaging (ART) in the detection of liver metastases on CT compared to portal venous phase imaging (PV) alone in patients with primary sarcomas. Multiphasic abdominal computed tomography (CT) images of patients with tissue-proven sarcomas were reviewed by five abdominal radiologists in a staggered fashion. Up to three of the largest or most conspicuous liver lesions were characterized on a four-point confidence level for PV independently, followed by PV+ART. Inter-observer reliability was evaluated with kappa statistics. Change in characterization of lesions by the addition of ART was calculated. Follow-up imaging was used to determine if index lesion characterization was valid. 55 of 149 patients had 470 liver lesion characterizations by the five readers with follow-up. Inter-observer agreement was κ=0.62 on PV and κ=0.58 on PV+ART. The intra-observer agreement between PV and ART interpretations of the same lesion was κ=0.93. 426 lesion characterizations were possible on both PV and ART. Only 6 characterizations were changed after the addition of ART; 4 of the 6 changes were incorrect when compared to follow-up. Only 6 lesion characterizations could be made on ART alone (missed by PV), with all the malignant lesions arising from primary leiomyosarcomas. For the lesions seen on PV alone, the sensitivity, specificity, PPV, NPV, and accuracy were 98.8%, 100%, 100%, 99.3%, and 99.6%, respectively. After the addition of ART, they were 98.8%, 98.7%, 97.5%, 99.4%, and 98.7%, respectively. ART adds marginal value to PV for characterization of metastatic liver lesions in patients with primary sarcomas, except possibly in primary leiomyosarcomas.