1175 – HER2MtGx score predicts high risk of recurrence and death in luminal breast cancer

Abstract

Introduction: Molecular profiling has been demonstrated to be a powerful tool for identifying breast cancer patients at
high risk of recurrence and death. A metagene score (HER2MtGx), derived from a 31-gene signature across human epidermal growth factor receptor 2 (HER2), Luminal, and Basal subspaces, was recently proposed and shown to identify
patients who do not benefit from HER2-targeted therapy.
Objective: To investigate how this model can identify high-risk patients with estrogen receptor-positive (ER+)/HER2- tumors.
Methods: Exploratory analyses were performed to assess HER2MtGx as a prognostic tool for ER+/HER2- breast cancer
in the METABRIC (n=85) and TCGA (n=558) datasets. Prognostic factors (age, clinical stage, and histological grade) and
the HER2, Basal and Luminal metagene scores estimated by the HER2MtGx model were used as independent variables.
Survival data were right-censored at 7.5 years, and disease-specific survival was analyzed using Kaplan–Meier and Cox
regression model. Time-dependent receiver operating characteristic (ROC) curve estimation was applied to survival data
to estimate the best metagene score cutoff to discriminate between low- and high-risk patients.
Results: Multivariate analysis showed age (hazard ratio [HR] 1.01; 95% confidence interval [CI] 1.0–1.02), stage (stage
2: HR 2.0; 95%CI 1.4–2.9; stage 3: HR 4.5; 95%CI 2.8–7.3), grade 3 (HR 1.5; 95%CI 1.1–2.0), and the Luminal score (HR 0.7;
95%CI 0.5–0.8) as significant prognostic factors. The Luminal score classified 217 (14%) patients as high risk. The metagene
risk score was not associated with stage (p=0.50). A strong association was observed with histological grade (p<0.001).
Disease-specific survival was 82.0% (95%CI 80.2–85.3) in low-risk patients compared with 70.1% (95%CI 63.3–77.7) in highrisk patients (p=0.002).
Conclusion: The HER2MtGx model can identify patients at high-risk with ER+/HER2- invasive breast cancer.
Ethics: Public data.
Funding: National Council for Scientific and Technological Development (CNPq), Grant No. 307801/2025-0.