Background/Aims: Prognostic models are lacking for patients with recurrent hepatocellular carcinoma (HCC) following surgical resection. This study devised and validated a new hepatoma arterial-embolization prognostic (HAP) score optimized for use in patients undergoing treatment with transarterial chemoembolization (TACE) for recurrence subsequent to surgical resection of HCC. Methods: Training cohort (n=424) and validation cohort (n=350) patients with recurrent HCC after resection treated with TACE between 2003 and 2016 were enrolled. Cox regression and area under the receiver operating characteristic curve (AUC) analyses were used to identify risk factors for survival and to calculate the predictive performance of risk scores, respectively. Results: The median age of the study population was 59.2 years. α-Fetoprotein >400 ng/mL (hazard ratio [HR]=1.815), serum albumin ≤3.5 g/dL (HR=1.966), tumor number ≥2 (HR=1.425), tumor size >5 cm at resection or recurrence (HR=1.356), segmental portal vein invasion at resection or recurrence (HR=2.032), and time from resection to recurrence ≤1 years (HR=1.849) independently predicted survival (all p<0.05). The postoperative HAP (pHAP) model based on the rounded HRs of these variables showed an AUC of 0.723 for predicting survival at 3 years, which was significantly higher than AUCs of other HAP-based models, including HAP, modified HAP, and modified HAP-II scores (0.578-0.621) (all p<0.05). The accuracy of pHAP was maintained in the entire cohort (n=774; AUC=0.776 at 3 years). Conclusions: A new pHAP score optimized for patients treated with TACE due to recurrent HCC after resection showed acceptable accuracy and was externally validated. Further studies of means by which to select treatment options other than TACE for high-risk patients according to pHAP scores are warranted.
Bibliographical noteFunding Information:
This study was supported by the Basic Science Research Program through the National Research Foundation of Korea funded by the Ministry of Science, ICT & Future Planning (grant number: 2016R1A1A1A05005138).
All Science Journal Classification (ASJC) codes