Prognostication and risk factor stratification for survival of patients with hepatocellular carcinoma: a nationwide big data analysis

Abstract

This study was conducted to identify risk factors affecting overall survival (OS) and provide prognostication for patients with hepatocellular carcinoma (HCC) using nationwide big data. Between January 2008 and December 2014, 10,573 adult patients with new HCC were registered in a nationwide database. Among them, 6830 patients without missing data were analyzed to construct a prognostication system. A validation cohort of 4580 patients was obtained from a tertiary hospital. All patients were assumed to have received the best treatment. A conditional inference tree analysis was performed to establish a prognostic system. The C-index and calibration plot for 5-year survival were estimated for validation. As a result, the tumor burden (TB) grade was the most significant factor in determining OS, and the cutoff was TB3 (TB1‒3 versus TB4). The patients were ultimately divided into 13 prognosis groups. The C-indexes were 0.714 and 0.737 (95% confidence interval, 0.733–0.742) in the nationwide (derivation) and hospital (validation) cohorts, respectively. In the calibration plot, the 5-year survival of the validation cohort largely matched the 45-degree line. In conclusion, the proposed prognostication system with a simple tree structure enabled the detailed stratification of patient prognosis and visualized the strata of risk factors affecting OS.