Purpose: The prognosis of patients with intermediate-stage hepatocellular carcinoma (HCC) treated by conventional TACE (cTACE) is greatly heterogeneous. This study aimed to develop a new survival prediction model to help select patients who would benefit better from cTACE treatment. Methods: We collected data of 848 treatment-naïve patients with BCLC B HCC who received cTACE as first-line therapy. The prognostic model's variables were derived from univariate and multivariate Cox regression analyses. The concordance index (C-statistic) calculated through cross-validation and bootstrap resampling was used for the model selection. The calibration of our final prediction model was also assessed. Results: The model showed a better discrimination ability than Bolondi's BCLC B1-B4 subclassification to predict the prognosis of BCLC B patients (C-statistic, 0.66 vs. 0.60; difference, 0.05, 95% CI, 0.03-0.07). In cross-validation, bootstrap resampling demonstrated that the model maintained sufficiently discriminant (an average of C-statistic, 0.66; 95% CI, 0.65-0.68). The model calibration was accurate in predicting survival of patients matched well with the observed outcomes. On the basis of the improved survival of 18 months or more as the responding patient, the observations of patients in each response category (responder and non-responder) were fair-moderately matched with those predicted by the model (κ=0.40, P<0.001). Conclusions: Based on clinically available features of patient, tumor and liver function, we developed an alternative prediction model with better performance than the Bolondi's substaging system for intermediate HCC patients after cTACE, which could help define the distinct subgroup of BCLC B patients who are suitable for cTACE treatment.
Keywords: Carcinoma, hepatocellular; Chemoembolization, therapeutic; Prognosis.
© The author(s).