A new scoring model predicting macroscopic vascular invasion of early-intermediate hepatocellular carcinoma

Abstract

Macroscopic vascular invasion cannot be properly predicted in advance in hepatocellular carcinoma patients based on clinical characteristics and imaging features.To develop a predictive scoring model of macroscopic vascular invasion in hepatocellular carcinoma patients after transcatheter arterial chemoembolization combined with radiofrequency ablation based on specific laboratory and tumor indicators.A predictive scoring model, which estimates the incidence of macroscopic vascular invasion at 1-year follow-up, was constructed based on a derivation cohort of 324 patients with hepatocellular carcinoma; a validation cohort of 120 patients was prospectively included. The prognostic value of the scoring model was determined by concordance index, time-dependent receiver operating characteristics, and calibration curves.Cox multivariate analysis of the derivation cohort identified prothrombin time, aspartate aminotransferase, and Barcelona clinic liver cancer (BCLC) staging as independent predictive factors of macroscopic vascular invasion. The areas under the receiver operating characteristic curves of the predictive scoring model were 0.832 and 0.785 in the derivation and validation cohorts, respectively, and the calibration curves fitted well. Kaplan-Meier analysis showed that the incidence of macroscopic vascular invasion was significantly higher in the high-risk group (score 0-2) than in the low-risk group (score 3-4) in both the derivation and validation cohorts (P < .0001 and P = .0008, respectively).The predictive scoring model enables the accurate prediction of macroscopic vascular invasion incidence 1 year in advance in hepatocellular carcinoma patients who undergo transcatheter arterial chemoembolization combined with radiofrequency ablation.

Figure 1

Kaplan–Meier analysis for the incidence of macroscopic vascular invasion (MVI). The incidence of MVI in the derivation cohort at 1-year follow-up was similar to that in the validation cohort (P = .4702). MVI = macroscopic vascular invasion.

Figure 2

Receiver operating characteristic (ROC) curves of the predictive scoring model for the (A) derivation cohort and (B) validation cohort. ROC = receiver operating characteristic.

Figure 3

The bars show the proportion of macroscopic vascular invasion (MVI) at 1-year follow-up for each score category in the derivation cohort (A). The incidence of MVI in the high-risk group (score 3–4) was significantly higher than that in the low-risk group (score 0–2) over the 12-month follow-up period in the derivation (B) and the validation (C) cohorts (P < .0001 and P = .0008, respectively). MVI = macroscopic vascular invasion.

Figure 4

Calibration plot of the scoring model for macroscopic vascular invasion (MVI) rates in the derivation cohort (A) and validation cohort (B), in which the predicted probability of survival was compared with actual survival, MVI = macroscopic vascular invasion.

Figure 5

ROC analysis of MVI incidence at 12 months in the derivation cohort. The area under the curve (AUC) of the scoring model developed in the study was greater than that of previously reported indicators. AUC = area under the curve, MVI = macroscopic vascular invasion, ROC = receiver operating characteristic.

Figure 6

The MVI incidences of BCLC A and B stage. The incidence of MVI in the high-risk group was significantly higher than that in the low-risk group over the 12-month follow-up period for patients with BCLC A stage (A, P < .0001) and B stage (B, P = .0123). BCLC = Barcelona clinic liver cancer, MVI = macroscopic vascular invasion.