Abstract PS3-07-26: Precision neoadjuvant treatment with Artificial Intelligence assisted subtyping in HR+/HER2- breast cancer: a randomized, open-label, phase 2 trial FASCINATE-N

Abstract Identification of effective neoadjuvant personalized treatment in hormone receptor-positive/human epidermal growth factor receptor 2-negative (HR+/HER2−) breast cancer remains a topic of ongoing research. We previously performed next-generation sequencing combined with metabolomics and proteomics on HR+/HER2- breast cancer sample. Using large-scale multi-omics data and similarity network fusion (SNF), we classified HR+/HER2- breast cancer into four novel subtypes: SNF1 (canonical luminal), SNF2 (immunogenic), 126 SNF3 (proliferative), and SNF4 (RTK-driven). Then we developed a digital pathology based on artificial intelligence (AI) assisted- classification approach employing convolutional neural networks derived from pathology whole-slide images (WSIs) to classify molecular subtypes. In this phase 2 trial, patients with stage II-III HR+/HER2- breast cancer were categorized into endocrine-based and targeted-based group based on the proposed similarity network fusion (SNF) subtyping by AI assisted digital pathology classification and randomly assigned (1:1) to receive precision or control treatment. Precision treatments included six cycles of dalpicilib (CDK4/6 inhibitor) plus endocrine therapy every 4 weeks for endocrine-based group and six cycles of targeted therapy (SHR-1316 [PD-L1 inhibitor] for SNF2, fuzuloparib [PARP1 inhibitor] for SNF3 and apatinib [VEGFR inhibitor] for SNF4) plus nab-paclitaxel and carboplatin every 4 weeks for targeted-based group. Control treatment was nab-paclitaxel plus carboplatin every 4 weeks for six cycles. The primary endpoint was pathological complete response (pCR). Among 251 randomized patients, SNF subtyping classified 49 patients (19.5%) into the endocrine-based group and 202 (80.5%) into the targeted-based group. Patients receiving precision treatment had significantly higher pCR rate (11.1% [14/126; 90% CI, 6.8-16.8] vs. 4.0% [5/125; 90% CI, 1.6-8.2]; P = 0.033). In the endocrine-based group, no pCR events were observed with endocrine treatment, compared with one pCR event in the control group (P = 0.490). In the targeted-based group, the pCR rate was significantly higher with precision treatment than with control (13.9% vs. 4.0%; P = 0.026). Grade 3 or higher treatment-related adverse events occurred similar in both treatment groups. No treatment-related deaths occurred. Our findings showed precision therapy could improve the benefit of neoadjuvant therapy for HR+/HER2- breast cancer patients. NET plus CDK4/6 inhibitor was verified to replace chemotherapy with better tolerance in the endocrine-base group while precision therapy was proved promising clinical activity and manageable safety profile in the targeted-based group. (ClinicalTrials.gov: NCT05582499). Citation Format: J. Li, L. Ma, P. Ji, L. Chen, J. Wu, G. Liu, Y. Hou, G. Di, Z. Hu, Y. Jiang, K. Yu, L. Fan, Z. wang, Z. Shao. Precision neoadjuvant treatment with Artificial Intelligence assisted subtyping in HR+/HER2- breast cancer: a randomized, open-label, phase 2 trial FASCINATE-N [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-07-26.