Accuracy of cross-sectional imaging in predicting tumor viability using the LI-RADS treatment response algorithm after image-guided percutaneous ablation with radiologic-pathologic explant correlation.

Abstract

Background: Accurate assessment of viable HCC on pre-transplant cross sectional imaging is important for correct organ allocation and overall patient outcome following liver transplantation.

Purpose: Determine accuracy of LI-RADS TRA compared to explant pathology in patients treated with thermal ablation, using contrast enhanced multiphase CT and MRI.

Materials and methods: Imaging studies for 119 consecutive adult HCC patients treated with thermal ablation and liver transplantation from March 2001 to September 2019 at a single tertiary care hospital were retrospectively studied by three Board-certified radiologists. LI-RADS TRA categories for each tumor were compared with explant pathology. Cohens Kappa test and inter-reader agreement by Fleiss κ test, with 95% confidence intervals obtained with bootstrap technique were used.

Results: Of the 119 patients (median age 59 years, 95 [80%] male) with 165 HCCs treated with percutaneous thermal ablation, 68% were completely necrotic and 32% were viable on pathologic analysis. Tumors viable on explant were larger on pre-treatment imaging (median 2.4 vs. 2.1 cm; p = 0.02) with no difference in pre-transplant ablation cavity sizes between groups (4.0 vs. 3.9 cm, respectively; p = 0.58). NPV of LI-RADS TRA for viable tumor was 71% (68-74); PPV of 62.5% (39-81) (p = 0.01) with a sensitivity of 19% (9.4-32), specificity of 95% (89-98), and accuracy of 70% (63-77). On explant, 55 incidental treatment naïve viable tumors, not visible on pre-transplant imaging, were found in 33 patients.

Conclusion: The "non-viable" category of LI-RADS TRA even when applied to a relatively uniform percutaneous ablation cohort, demonstrated low sensitivity in predicting absence of viable tumor. MRI had more accuracy than CT in predicting tumor viability when compared to explant pathology.