ABSTRACT: Background:Early recurrence of hepatocellular carcinoma (HCC) is a major obstacle to improving the prognosis, and no widely accepted adjuvant therapy guideline for patients post-liver resection is available. Currently, all available methods and biomarkers are insufficient to accurately predict post-operation HCC patients' risk of early recurrence and their response to adjuvant therapy. Methods:In this study, we downloaded four gene expression datasets (GSE14520, GSE54236, GSE87630, and GSE109211) from the Gene Expression Omnibus database and identified 34 common differentially expressed genes associated with HCC dysregulation and response to adjuvant sorafenib. Then, we constructed a novel 11-messenger RNA predictive model by using ROC curves analysis, univariate Cox regression analysis, and LASSO Cox regression analysis. Furthermore, we validated the predictive values of the risk model in GSE14520 and TCGA-LIHC cohorts by using Kaplan-Meier survival analysis, multivariable Cox regression analysis, and decision curve analysis, respectively. Results:The risk score model could identify patients with a high risk of HCC recurrence at the early stage and could predict the response of patients to adjuvant sorafenib. Patients with a high risk score had a worse recurrence rate in training cohorts (2-year: p < 0.0001, hazard ratio (HR): 4.658, confidence interval 95% CI [2.895-7.495]; 5-year: p < 0.0001, HR: 3.251, 95% CI [2.155-4.904]) and external validation cohorts (2-year: p < 0.001, HR: 3.65, 95% CI [2.001-6.658]; 5-year: p < 0.001, HR: 3.156, 95% CI [1.78-5.596]). The AUC values of the risk score model for predicting tumor early recurrence were 0.746 and 0.618, and that of the risk score model for predicting the response to adjuvant sorafenib were 0.722 and 0.708 in the different cohort, respectively. Multivariable Cox regression analysis and decision curve analysis also showed that the risk score model was superior to and independent of other clinicopathologic characteristics. Moreover, the risk score model had excellent abilities to predict the overall survival and HCC recurrence of patients with the same tumor stage category. Conclusions:Our risk model is a reliable and superior predictive tool. With this model, we could optimize the risk stratification based on early tumor recurrence and could evaluate the response of patients to adjuvant sorafenib after liver resection.
