ABSTRACT
Objective:To construct and validate prognostic nomograms predicting overall survival (OS) and cancer-specific survival (CSS) of patients with late-stage hepatocellular carcinoma (HCC).Methods:A retrospective cohort study was used in this report. Screened 2382 late-stage HCC patients obtained from Surveillance, Epidemiology, and End Results (SEER) database (2010—2015), were randomly classified into the training cohort and the internal validation cohort by using the function in R software according to the ratio of 1∶1. Chi-square test was applied to verify the comparability of data between two groups. The external validation cohort ( n=62) were collected from the Affiliated Zhangjiagang Hospital of Soochow University. Based on univariate and multivariate COX regression analyses in the training cohort, this study constructed nomograms for 6- and 12- month OS and CSS. Concordance index (C-index), calibration plots, the receiver operating characteristic (ROC) curves and Kaplan-Meier survival curves were applied to measure the performance of nomograms in the training cohort and to validate nomograms in two validation cohorts. The clinical utility was measured by decision curve analysis (DCA). Results:Two nomograms were constructed. The identified risk factors included sex, Edmondson-Steiner grade, T stage, N stage, M stage, tumor size, bone metastasis, Alpha-fetoprotein (AFP), surgery of primary site, radiation and chemotherapy. The C-index for OS in the training and two validation cohorts was 0.729(95% CI: 0.711-0.747), 0.721(95% CI: 0.705-0.737) and 0.860(95 CI: 0.831-0.889), respectively. The C-index for CSS in the training and two validation cohorts was 0.732(95% CI: 0.714-0.750), 0.725(95% CI: 0.707-0.743) and 0.862(95% CI: 0.829-0.895), respectively. Afterwards, for nomograms in the training and two validation cohorts, C-index and calibration plots expressed great predictive accuracy and concordance. ROC curves and Kaplan-Meier survival curves demonstrated good prognostic ability. Furthermore, nomograms performed superior to other models. DCA showed substantial clinical utility. Conclusion:This study has developed and validated nomograms predicting 6- and 12- month OS and CSS of patients with late-stage HCC, which may be useful to develop the individualized treatment.