Genetic biomarkers predicting sensitivity to atezolizumab plus bevacizumab in resected hepatocellular carcinoma (HCC): A SNP prediction signature for HCC.

e16014 Background: Current prognostic scoring systems based on clinicopathologic variables are inadequate for identifying patients with hepatocellular carcinoma (HCC) who would benefit from adjuvant therapy with atezolizumab plus bevacizumab following surgery. We developed a single-nucleotide polymorphisms (SNP)-based classifier to improve postoperative risk stratification and prediction of adjuvant therapy benefit for these patients. Methods: In this multicentre study, we developed a 12-SNP classifier derived from blood-based SNP profiles correlated with recurrence-free survival from 453 patients with resected HCC in training set. We assessed intratumour heterogeneity by analysing two regions of paraffin-embedded specimens from the same tumours in the training set. We validated the classifier’s prognostic and predictive performance in an internal testing set (n = 195), two independent external validation sets (n = 238 and n = 308), and TCGA dataset (n = 278). we conducted a nested case-control dataset (n = 388) to evaluate classifier’s ability for identifying patients likely to benefit from atezolizumab plus bevacizumab adjuvant therapy. Additionally, we performed in vitro analyses of the functional relevance of the twelve SNPs. Results: The twelve-SNP-based classifier demonstrated consistent predictive accuracy in blood samples and two different regions of the training set. The twelve-SNP-based classifier precisely predicted recurrence-free survival of patients in training and four validation sets. In the nested case-control analysis, atezolizumab-bevacizumab adjuvant therapy was associated with improved recurrence-free survival (HR 2.120, 1.246-3.599; p = 0.0054) and overall survival (HR 2.870, 1.194-6.900; p = 0.0184) in the high-risk subgroup as defined by the twelve-SNP-based classifier. In vitro analyses showed that the high-risk group, defined by a twelve-SNP classifier, exhibited upregulation of pathways involved in angiogenesis, tumor invasion, metastasis, and tumor immunosuppression compared to low-risk patients. Conclusions: The twelve-SNP classifier is a practical and reliable predictor that can complement the current classification system for identifying patients most likely to benefit from atezolizumab-bevacizumab adjuvant therapy.