A Practical Nomogram and Risk Stratification System Predicting the Cancer-Specific Survival for Patients With Advanced Hepatocellular Carcinoma

Abstract

Background: Hepatocellular carcinoma (HCC) has the highest cancer-related mortality rate. This study aims to create a nomogram to predict the cancer-specific survival (CSS) in patients with advanced hepatocellular carcinoma.

Methods: Patients diagnosed with advanced HCC (AJCC stage III and IV) during 1975 to 2018 were obtained from the Surveillance, Epidemiology, and End Results (SEER) database. Qualified patents were randomized into training cohort and validation cohort at a ratio of 7:3. The results of univariate and multivariate Cox regression analyses were used to construct the nomogram. Consistency index (C-index), area under the time-dependent receiver operating characteristic (ROC) curve [time-dependent area under the curve (AUC)], and calibration plots were used to identify and calibrate the nomogram. The net reclassification index (NRI), integrated discrimination improvement (IDI), and C-index, and decision curve analysis DCA were adopted to compare the nomogram's clinical utility with the AJCC criteria.

Results: The 3,103 patients with advanced hepatocellular carcinoma were selected (the training cohort: 2,175 patients and the validation cohort: 928 patients). The C-index in both training cohort and validation cohort were greater than 0.7. The AUC for ROC in the training cohort was 0.781, 0.771, and 0.791 at 1, 2, and 3 years CSS, respectively. Calibration plots showed good consistency between actual observations and the 1-, 2-, and 3-year CSS predicted by the nomogram. The 1-, 2-, and 3-year NRI were 0.77, 0.46, and 0.48, respectively. The 1-, 2-, and 3-year IDI values were 0.16, 0.15, and 0.12 (P < 0.001), respectively. DCA curves in both the training and validation cohorts demonstrated that the nomogram showed better predicted 1-, 2-, and 3-year CSS probabilities than AJCC criteria.

Conclusions: This study established a practical nomogram for predicting CSS in patients with advanced HCC and a risk stratification system that provided an applicable tool for clinical management.

Keywords: AJCC (TNM) staging system; advanced hepatocellular carcinoma; cancer-specific survival; nomogram; risk stratification.

Figure 1

Flow diagram of the advanced hepatocellular carcinoma patients with training and validation cohorts.

Figure 2

A nomogram for advanced hepatocellular carcinoma patients. *P < 0.05; **P < 0.01; ***P < 0.001.

Figure 3

ROC of the nomogram for 1-, 2-, and 3-year prediction. (A) Training cohorts based on the nomogram. (B) Validation cohorts based on the nomogram.

Figure 4

Calibration plots of 1-year, 2-year, and 3-year CSS for advanced hepatocellular carcinoma patients. (A, C, E) Calibration plots of 1-year, 2-year, and 3-year CSS in the training cohort. (B, D, F) Calibration plots of 1-year, 2-year, and 3-year CSS in the validation cohort.

Figure 5

Decision curve analysis of CSS-associated nomogram and AJCC criteria. (A, C, E) DCA curves of 1-year, 2-year, and 3-year CSS in the training cohort. (B, D, F) DCA curves of 1-year, 2-year, and 3-year CSS in the validation cohort.

Figure 6

C-index analysis. (A) The nomogram related C-index. (B) AJCC staging criteria related C-index.

Figure 7

Cutoff point for risk stratifications selected using X-tile.

Figure 8

Kaplan–Meier CSS curves of patients with advanced hepatocellular carcinoma based on different criteria. (A, B) Kaplan–Meier CSS curves of training and validation cohorts based on the new risk stratification system. (C, D) Kaplan–Meier CSS curves of training and validation cohorts based on AJCC staging criteria.