Contrast enhanced ultrasound combined with serology predicts hepatocellular carcinoma recurrence: a retrospective observation cohort study

• Published in: Frontiers in oncology 2023-07, Vol.13, p.1154064-1154064
• Main Authors: Tu, Haibin, Feng, Siyi, Chen, Lihong, Huang, Yujie, Zhang, Juzhen, Wu, Xiaoxiong
• Format: Article
• Language: English
• Subjects: contrast-enhanced ultrasound; early recurrence; hepatocellular carcinoma; Oncology; prediction; serological markers
• ISSN: 2234-943X; 2234-943X
• DOI: 10.3389/fonc.2023.1154064

To construct a novel model based on contrast-enhanced ultrasound (CEUS) and serological biomarkers to predict the early recurrence (ER) of primary hepatocellular carcinoma within 2 years after hepatectomy. A total of 466 patients who underwent CEUS and curative resection between 2016.1.1 and 2019.1.1 were retrospectively recruited from one institution. The training and testing cohorts comprised 326 and 140 patients, respectively. Data on general characteristics, CEUS Liver Imaging Reporting and Data System (LI-RADS) parameters, and serological were collected. Univariate analysis and multivariate Cox proportional hazards regression model were used to evaluate the independent prognostic factors for tumor recurrence, and the Contrast-enhanced Ultrasound Serological (CEUSS) model was constructed. Different models were compared using prediction error and time-dependent area under the receiver operating characteristic curve (AUC). The CEUSS model's performances in ER prediction were assessed. The baseline data of the training and testing cohorts were equal. LI-RADS category, α-fetoprotein level, tumor maximum diameter, total bilirubin level, starting time, iso-time, and enhancement pattern were independent hazards, and their hazards ratios were 1.417, 1.309, 1.133, 1.036, 0.883, 0.985, and 0.70, respectively. The AUCs of CEUSS, BCLC,TNM, and CNLC were 0.706, 0.641, 0.647, and 0.636, respectively, in the training cohort and 0.680, 0.583, 0.607, and 0.597, respectively, in the testing cohort. The prediction errors of CEUSS, BCLC, TNM, and CNLC were 0.202, 0.205, 0.205, and 0.200, respectively, in the training cohort and 0.204, 0.221, 0.219, and 0.211, respectively, in the testing cohort. The CEUSS model can accurately and individually predict ER before surgery and may represent a new tool for individualized treatment.