Development and validation of a model to predict the risk of distant metastases from hepatocellular carcinoma: a real-world retrospective study.

PURPOSE

This study aimed to construct a novel clinical prediction model to predict the risk of distant metastases (DM) in hepatocellular carcinoma (HCC).

METHODS

We included 3869 HCC patients, comprising 3076 patients from the Surveillance, Epidemiology, and End Results (SEER) database and 793 patients from a hospital in China. Variables with a P-value < 0.05 in the univariate logistic analysis were entered into the multivariate analysis to determine the independent predictive factors for DM in HCC. A nomogram was created based on the independent predictive factors. The predictive performance of the model was assessed using the receiver operating characteristics (ROCs) curve, decision curve analysis (DCA), calibration curves, and clinical impact curve analysis (CIC). Additionally, we developed a user-friendly web-based calculator based on the model.

RESULTS

The multivariate logistic regression analysis revealed that tumor size (P < 0.001), type of treatment (P < 0.001), T stage (P = 0.001), N stage (P < 0.001), and grade (P = 0.043) were identified as independent predictive factors. A nomogram was constructed based on these factors. The area under the ROC curves (AUC) value was 0.845 (95% CI 0.815-0.874) for the training set, 0.818 (95% CI 0.774-0.863) for the internal validation set, and 0.823 (95% CI 0.770-0.876) for the external validation set. Moreover, DCA analysis, calibration curves, and CIC analysis demonstrated the favorable predictive performance of the nomogram. Finally, a more user-friendly web-based calculator was developed.

CONCLUSION

We developed a nomogram and showed its favorable predictive performance in predicting DM in HCC. Furthermore, we developed a more user-friendly web-based calculator, which has the potential to aid clinicians in individualized diagnosis and make better clinical decisions for HCC patients.