Toward consensus reporting of radiation-induced liver toxicity in the treatment of primary liver malignancies: Defining clinically relevant endpoints

Background: Our purpose was to define the most clinically relevant “nonclassic” radiation-induced liver disease (RILD) endpoints in cirrhotic patients receiving stereotactic body radiation therapy or proton beam therapy for primary liver cancer. Methods and materials: We retrospectively collected pretreatment, detailed toxicity (≤6 months posttreatment), and outcomes data from 48 patients. Deaths were examined for association with RILD. Univariate and multivariate Cox models defined significant predictors of overall survival (OS)/RILD-specific survival (RILD-SS). Results: With median follow-up of 13 months, 23 patients (48%) had an increase in Child-Pugh (CP) score (≥2, 25%) and 3 (6%) had ≥G3 transaminase elevation. Of 18 deaths, 6 were potentially ascribed to RILD. Univariate analysis showed that CP score increases of ≥1 and ≥2 and CP class change predicted OS, as did ≥G3 aspartate transaminase (AST) elevation and ≥1 Common Terminology Criteria for Adverse Events (CTCAE) AST toxicity grade change. On multivariate analysis, CP score increase of ≥2 and ≥1 CTCAE AST toxicity grade change were the strongest independent nonclassic RILD predictors of OS. For RILD-SS, CP score increases of ≥2, ≥grade 3 CTCAE alanine transaminase, and ≥grade 2 bilirubin elevations were predictive. Conclusions: Increased CP score ≥2 strongly predicts for both OS and RILD-SS and should be reported in future studies along with transaminase elevations, which are also predictive of outcomes.