A Novel Method for Prognostic Risk Classification After Carbon‐Ion Radiotherapy for Hepatocellular Carcinoma Using Data‐Mining Methods

No classification methods to predict prognosis after carbon-ion radiotherapy for hepatocellular carcinoma have yet been reported. This study aimed to develop risk classification for cancer-specific survival (CSS) after carbon-ion radiotherapy for hepatocellular carcinoma using decision tree analysis as a data-mining method. In this single-center, retrospective study, we analyzed 90 consecutive patients with hepatocellular carcinoma treated with carbon-ion radiotherapy between 2018 and 2022. Liver tumors were irradiated at 60 Gy (relative biological effectiveness [RBE]) in four fractions. If the tumor was close to the gastrointestinal tract, it was irradiated at 60 Gy [RBE] in 12 fractions. Univariate and multivariate analyses of progression-free survival (PFS) and CSS were performed to assess patients' background and treatment-related factors. Decision tree analysis (DTA) was performed to assess prognostic factors for CSS that were significantly different in the multivariate analysis. The median follow-up period was 32.8 months for all patients and 35.6 months for survivors. Multivariate analysis identified dose fractionation and pretreatment alpha-fetoprotein values as significant prognostic factors for PFS and CSS. Moreover, clinical stage and pretreatment protein induced by vitamin K absence or antagonist ΙΙ values were significant prognostic factors for CSS. DTA revealed that the patients could be divided into three groups according to prognosis: low-risk, high-risk, and intermediate-risk. Consequently, the 3-year CSS rates for the low-, intermediate-, and high-risk groups were 100%, 73.3%, and 44.4%, respectively. DTA represents a new method for risk classification for CSS after carbon-ion radiotherapy for hepatocellular carcinoma based on tumor markers and clinical stage.