Multi-institutional validation of novel models for predicting the prognosis of patients with huge hepatocellular carcinoma.

Abstract

The population of patients with huge hepatocellular carcinoma (H-HCC diameter > 10.0 cm) is an odd group that is not well adjudicated in the current staging systems, whose prognosis after curative resection varies. We aimed to develop novel models to predict the long-term outcomes of patients with H-HCC without portal vein tumor thrombus after hepatectomy. There were 1076 H-HCC patients enrolled who underwent curative liver resection in five institutions in China. In total, 670 patients were recruited from our center and randomly divided into the training cohort (n = 502) and internal validation (n = 168) cohorts. Additionally, 406 patients selected from other four centers as the external validation cohort. Novel models were constructed based on independent preoperative and postoperative predictors of postsurgical recurrence (PSR) and postsurgical mortality (PSM) determined in multivariable cox regression analysis. The predictive accuracy and discriminative ability of the model were measured using Harrell's concordance index (C index) and calibration curve and compared with five conventional HCC staging systems. PSR model and PSM model were constructed based on tumor number, microscopic vascular invasion, tumor differentiation, preoperative alpha-fetoprotein level, albumin-bilirubin grade, liver segment invasion, neutrophil-to-lymphocyte ratio or platelet-to-neutrophil ratio, and surgical margin or intraoperative blood transfusion. The C-indexes were 0.84 (95% CI, 0.78-0.90) and 0.85 (95% CI, 0.78-0.91) for the PSR and PSM models, respectively, which were substantially higher than those of the five conventional HCC staging systems (0.63-0.75 for PSR; 0.66-0.77 for PSM). The two novel models achieved more accurate prognostic predictions of PSR and PSM for H-HCC patients after curative liver resection.