LI-RADS CT/MRI Radiation Treatment Response Algorithm Version 2024: Category Redistribution and Short-Term Outcomes in Patients Undergoing Y-90 Radioembolization for HCC

Background: LI-RADS CT/MRI Radiation Treatment Response Algorithm (TRA) version 2024 (v2024) addresses a pitfall of the earlier algorithm relating to expected persistent enhancement of hepatocellular carcinoma (HCC) responding to radiation. v2024 removes LR-TR Equivocal, introduces LR-TR Nonprogressing (stable or decreasing masslike enhancement), and more narrowly defines LR-TR Viable (new or increasing masslike enhancement). Objective: To evaluate in patients with HCC treated by Y-90 radioembolization the redistribution of categories in LI-RADS CT/MRI Radiation TRA v2024 versus LI-RADS CT/MRI TRA version 2018 (v2018) and to assess short-term outcomes of v2024 categories. Methods: This retrospective study included 242 patients (57 women, 185 men; median age, 65 years) with 319 HCCs treated by Y-90 radioembolization from February 2011 to March 2022 and evaluated by initial 3-month posttreatment CT or MRI. Two radiologists assigned v2018 and v2024 categories; a third radiologist resolved discrepancies. The radiologists also assessed available second posttreatment CT or MRI examinations using v2024. Overall survival (OS) was determined. Results: On initial follow-up, by v2018, 18 (5.6%) lesions were LR-TR Nonviable, 21 (6.6%) LR-TR Equivocal, and 280 (87.8%) LR-TR Viable; by v2024, 18 (5.6%) were LR-TR Nonviable, 182 (57.1%) LR-TR Nonprogressing, and 119 (37.3%) LR-TR Viable. All LR-TR Equivocal and 161 (57.5%) LR-TR Viable lesions by v2018 were recategorized as LRTR Nonprogressing by v2024. Of 96 LR-TR Nonprogressing lesions with second follow-up, 63 (65.6%) remained LRTR Nonprogressing, 19 (19.8%) transitioned to LR-TR Nonviable, and 14 (14.6%) transitioned to LR-TR Viable. Of 29 LR-TR Viable lesions by v2024 with second follow-up, 23 (79.3%) remained LR-TR Viable, and 6 (20.7%) transitioned to LR-TR Nonprogressing. By Kaplan-Meier analysis using initial categories, OS showed no significant difference between LR-TR Equivocal and LR-TR Viable for v2018 (p=.05) but was significantly worse for LR-TR Viable than LRTR Nonprogressing for v2024 (p<.001). Conclusion: For v2024, LR-TR Viable was substantially less frequent versus v2018, and the majority of lesions were assigned LR-TR Nonprogressing. Using v2024, most LR-TR Viable lesions and the majority of LR-TR Nonprogressing lesion on initial follow-up remained as such on later imaging. Initial v2024 categories were associated with OS. Clinical Impact: The findings support the revisions in v2024.