Background: Radiation-induced liver disease (RILD) has been tracked when assessing the safety of fractionated liver radiotherapy. However, due to the rarity and severity of events, a more sensitive measure of liver damage is necessary. This study characterizes the time course of changes in generalizable measures of liver function as potential toxicity metrics. Methods: In this IRB-approved retrospective study, the records of 63 patients with 87 HCC tumors treated with SBRT at the University of Michigan between 2006 and 2012 were reviewed. Changes in Child-Pugh (CP), MELD, and MELD-Na were analyzed using the Student t test and chi-squared test. Results: 83% of the patients had cirrhosis, 62% hepatitis (hep) C, 7% hep B, 25% alcoholic, and 13% other. 83% had prior liver-directed therapy. 24% had >1 tumor concurrently treated with SBRT. Median tumor size was 2.3 cm (0.7-10), gross tumor volume was 9.2 cc (0.6-469), and mean liver dose-GTV was 4.4 Gy (0-17.6 Gy). Prior to SBRT, 73% were CP A, 25% CP B, and 2% CP C. Median baseline CP, MELD, and MELD-Na were 5, 9, and 10. Mean CP increases after 3, 6, 9, and 12 months were 0.84, 1.79, 1.72, and 1.33; increases in MELD 1.47, 2.88, 5.38, and 3.91; increases in MELD-Na 1.45, 2.03, 4.24, and 2.48. All changes were significantly increased from baseline. On univariate analysis, >1 tumor and cirrhosis predicted for a 1+ point increase in CP and >1 tumor and higher mean liver dose-GTV for a 2+ point increase in CP. Older age, >1 tumor, smaller GTV, and smaller max tumor dimension predicted for a 10+ point increase in MELD and >1 tumor, hep C cirrhosis, and higher baseline MELD-Na for MELD-Na. Patients treated concurrently to >1 tumor had greater increases in CP (3.60 vs. 0.81), MELD (9.33 vs. 1.97), and MELD-Na (8.47 vs. 2.19), p < 0.0001 for all, compared to those with 1 tumor. No differences were seen with gender, portal vein thrombosis, number of prior treatments and baseline Child-Pugh classifications. Conclusions: We describe a time course and predictive factors for change in CP, MELD, and MELD-Na scores after SBRT for HCCs. These should be investigated further for potential use in toxicity modeling after incorporating dosimetric parameters.