Defining radiation induced liver toxicity in the treatment of hepatocellular carcinoma: Which metric is most predictive for survival?

Abstract

361 Background: Radiation induced liver disease (RILD) is of critical concern in the treatment of hepatocellular carcinoma (HCC) with radiation therapy (RT). Variability exists in metrics used to define RILD with no consensus on which best predict for overall survival (OS) and RILD-specific survival (RILDSS). We examined the correlation between toxicity metrics and clinical outcomes in a heavily pre-treated population that received RT. Methods: The charts of 37 HCC patients treated from 2013 - 2015 were reviewed retrospectively. At baseline, 62% were Child-Pugh (CP)-A, 32% CP-B and 5% CP-C. The majority (59%) had prior liver-directed therapy (LDT), 43% received stereotactic body RT and 49% proton RT. Pre-treatment, toxicity ( ≤ 6 months from treatment) and outcomes data were collected. Deaths from RILD were scored. Pre-treatment factors and toxicity outcomes were assessed by univariate Cox models for association with OS and RILDSS. Statistically significant predictors formed the basis for stepwise multivariate Cox regression to retain independent predictors of survival. Results: At a median follow-up of 8 months, 14 patients had an increase in CP score ( ≥ 2, n = 7) and 3 had ≥ G3 RTOG transaminitis. There were 11 deaths, 5 from RILD. On univariate analysis (UA), tumor size, pre-treatment liver function, prior LDT and 5 toxicity metrics (CP score increases and transaminitis) were significantly associated with OS. An increase of ≥ 1 CP score (HR 22.7, p = 0.005), pre-treatment ALBI grade (HR 6.0, p = 0.02) and tumor size (HR 1.2, p = 0.01) were independent predictors of OS on multivariate analysis (MVA). Similar factors were associated with RILDSS on UA, including ≥ 2 CP score increase and ≥ G3 ALT elevation; however, only pre-treatment CP score (HR 4.0, p = 0.01) and tumor size (HR 1.5, p = 0.03) were independently predictive on MVA. Conclusions: Pre-treatment liver functional status and tumor size were highly predictive of OS and RILDSS, suggesting that baseline functional hepatic reserve is the primary determinant in developing fatal RILD rather than post-RT changes in liver function. Further work is needed to define dosimetric parameters and pre-treatment factors that predict RILD toxicity.