Abstract
BACKGROUND: The preoperative prediction of vessels encapsulating tumor cluster (VETC) status in hepatocellular carcinoma (HCC) patients is essential for both prognostic assessment and treatment decision-making. This study aims to systematically evaluate the diagnostic performance of various imaging techniques for non-invasive prediction of VETC presence in HCC, through a comprehensive meta-analysis. METHODS: An extensive search was conducted across PubMed, Embase, Web of Science, and the Cochrane Library to locate studies assessing the diagnostic effectiveness of imaging techniques for anticipating VETC status in individuals with HCC. The quality of the studies included was assessed utilizing the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool. A bivariate random-effects meta-analysis was utilized to derive pooled estimates of sensitivity, specificity, positive likelihood ratio (PLR), and negative likelihood ratio (NLR) for every imaging technique. Comprehensive evaluations of heterogeneity and publication bias were also performed. To delve deeper into the sources of variability, both meta-regression and subgroup analyses were executed. The Z test was employed to analyze differences in sensitivity, specificity, PLR, and NLR, while the DeLong test was implemented to compare the area under the receiver operating characteristic curve values. RESULTS: A total of 18 studies involving 3,615 patients with HCC were included in the meta-analysis. The pooled diagnostic performance of imaging methods for predicting VETC status demonstrated a sensitivity of 0.79 [95% confidence interval (CI): 0.73-0.84], specificity of 0.83 (95% CI: 0.78-0.87), area under the summary receiver operating characteristic (SROC) curve of 0.88 (95% CI: 0.85-0.91), PLR of 4.64 (95% CI: 3.44-6.26), and NLR of 0.25 (95% CI: 0.19-0.33). However, significant heterogeneity was observed (I(2)=73.18% for sensitivity, I(2)=83.56% for specificity). Subgroup analysis revealed that artificial intelligence (AI)-based imaging methods in validation cohorts exhibited significantly better diagnostic performance compared to non-AI methods, with higher area under the curve (AUC) (0.88 vs. 0.82, P=0.01) and specificity (0.84 vs. 0.70, P=0.043). CONCLUSIONS: Preoperative imaging methods demonstrate moderate diagnostic accuracy for the non-invasive prediction of VETC status in HCC. Notably, AI-based imaging approaches show superior diagnostic performance compared to conventional imaging techniques in identifying VETC patterns. Nevertheless, because of the substantial heterogeneity observed among the studies included in the analysis, caution is warranted in interpreting these results.