科研速览 · Science Skim继续刷下去 · Keep skimming →
◆ Computational biology and chemistry2026-09-04

Construction of efficacy prediction model for T-DXd treatment of HER2-positive and HER2-low expression breast cancer based on hematological indicators: A real-world study.

Yanfang Su, Die Sang, Guoyong Zhang, Haiyan Liu, Yurong Zhang, Shanmin Fan, Jintao Zhang, Shiyu Li, Yanyan Zhang, Huachao Feng, Meiqing Zhao, Longmei Zhao, Peng Yuan, Binsheng He, Man Li

一句话结论 · In one sentence

Hematological toxicity is a recognized concern with T-DXd therapy. Our predictive model demonstrates strong performance in identifying patients suitable for treatment, highlighting that rigorous monitoring of hematological indicators is critical for mitigating risk and upholding safety standards in clinical practice.

原始摘要(英文原文)· Original abstract
BACKGROUND: This study explores hematological changes during trastuzumab deruxtecan (T-DXd) treatment in HER2-positive and HER2-low breast cancer patients, aiming to develop an efficacy prediction model for personalized clinical decisions. MATERIALS: A retrospective analysis of 114 patients treated with T-DXd at Beijing Chaoyang District Sanhuan Cancer Hospital (March 2023-December 2024) was conducted. Data included demographics, treatment history, and laboratory results. Efficacy was assessed via RECIST 1.1, with analyses using the Wilcoxon test and Kaplan-Meier method. RESULTS: HER2-positive patients had a progression-free survival (PFS) of 12.2 months vs. 10.2 months for HER2-low. Common adverse events included Nausea, primarily grade 1 or 2. T-DXd caused significant reductions in red blood cell count, hemoglobin, hematocrit, creatine kinase, and homocysteine (p < 0.05), along with increased red cell distribution width. Responders (CR+PR+SD) and progressors (PD) differed in platelets, plateletcrit, and homocysteine (p < 0.05). Stratification by estrogen receptor (ER) and progesterone receptor (PR) expression and Ki67 revealed distinct PFS outcomes. We aimed to predict the efficacy of T-DXd using pre-treatment hematological parameters and various machine learning models, including Random Forest, SVM, XGBoost, and LightGBM. The performance of the optimal models, as evaluated by leave-one-out cross-validation, yielded AUC values of 0.748 (XGBoost), 0.881 (SVM), and 0.830 (SVM) for models based on Complete Blood Count (CBC) parameters, biochemical markers, and their combination, respectively. CONCLUSION: Hematological toxicity is a recognized concern with T-DXd therapy. Our predictive model demonstrates strong performance in identifying patients suitable for treatment, highlighting that rigorous monitoring of hematological indicators is critical for mitigating risk and upholding safety standards in clinical practice.
读原文 · Read the paper ↗

AI 追问PRO

登录后使用 AI 追问

讨论区

登录后参与讨论

相关论文 · Related

Construction of efficacy prediction model for T-DXd treatment of HER2-positive and HER2-low expression breast cancer based on hematological indicators: A real-world study. — 科研速览 Science Skim