The Prognostic Value of Preoperative Serum Markers and Risk Classification in Patients with Hepatocellular Carcinoma.

ABSTRACT

BACKGROUND: Complex hepatocellular carcinoma (HCC) prognostic biomarkers have been reported in various studies. We aimed to establish biomarkers that could predict prognosis, and formulate a simple classification using non-invasive preoperative blood test data. METHODS: We retrospectively identified 305 patients for a discovery cohort who had undergone HCC-related hepatectomy at four Japanese university hospitals between January 1, 2011 and December 31, 2013. Preoperative blood test parameter optimal cut-off values were determined using receiver operating characteristic curve analysis. Cox uni- and multivariate analyses were used to determine independent prognostic factors. Risk classifications were established using classification and regression tree (CART) analysis. Validation was performed with 267 patients from three other hospitals. RESULTS: In multivariate analysis, α-fetoprotein (AFP, p < 0.001), protein induced by vitamin K absence or antagonist-II (PIVKA-II, p = 0.006), and C-reactive protein (CRP, p < 0.001) were independent prognostic factors for overall survival (OS). AFP (p = 0.007), total bilirubin (p = 0.001), and CRP (p = 0.003) were independent recurrent risk factors for recurrence-free survival (RFS). CART analysis results formed OS (CRP, AFP, and albumin) and RFS (PIVKA-II, CRP, and total bilirubin) decision trees, based on machine learning using preoperative serum markers, with three risk classifications. Five-year OS (low risk, 80.0%; moderate risk, 56.3%; high risk, 25.2%; p < 0.001) and RFS (low risk, 43.4%; moderate risk, 30.8%; high risk, 16.6%; p < 0.001) risks differed significantly. These classifications also stratified OS and RFS risk in the validation cohort. CONCLUSION: Three simple risk classifications using preoperative non-invasive prognostic factors could predict prognosis.