Abstract Background: Bevacizumab, an anti-vascular endothelial growth factor (VEGF) monoclonal antibody, inhibits tumor angiogenesis and is used to treat several advanced-stage cancers, such as glioblastoma, ovarian, lung, and colorectal cancers. More recently, it has been incorporated into treatment regimens for locally advanced nasopharyngeal carcinoma (NPC). Despite its widespread adoption, no reliable biomarker has been developed to predict the treatment outcome. Moreover, bevacizumab treatment carries risks such as bleeding, hypertension, and proteinuria. Therefore, selecting patients for bevacizumab addition to already intensive chemotherapy regimens remains a clinical challenge. Methods: We obtained 19 formalin-fixed paraffin-embedded (FFPE) biopsies from a completed phase 2 clinical trial (NCT01309633) for locally advanced NPC, where patients received bevacizumab prior to standard concurrent chemoradiation. Based on haematoxylin and eosin (H0.05), with 44 upregulated in PR. These genes were related to extracellular matrix remodelling, vascular endothelia, and immune response. GSEA using cell-type signature gene sets or Curated Cancer Cell Atlas gene sets from MSigDB showed enrichment of inflamed fibroblasts, endothelial, and stromal cell types in PR, and cell-cycle programs in CR. Therefore, the 44-gene panel was used as a gene signature to predict reduced response to bevacizumab. Both ssGSEA and GSVA scores robustly separated PR from CR (p0.001). These findings suggest that stromal/vascular inflammation in the tumor epithelial regions is associated with reduced bevacizumab response. Conclusions: In this study, we derived a 44-gene tumor epithelial signature that predicts reduced response to bevacizumab in locally advanced NPC and confirmed its performance with ssGSEA and GSVA. The signature captures stromal, endothelial, and immune-inflammatory programs within tumor epithelial regions and informs patient selection for anti-angiogenic therapy. Our current work includes validation in independent patient cohorts across multiple cancer types to improve clinical utility and generalizability. Citation Format: Yi Ren, Chee Yit Lim, Yaw Chyn Lim, Joseph W. Foley, Raymond Tsang, Wan Qin Chong, Boon Cher Goh, Joshua K. Tay. Development of a gene expression predictor for bevacizumab response abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 3731.
Ren et al. (Fri,) studied this question.