2842 - Dosimetric and Clinical Predictors of Severe Radiation-Induced Lymphopenia In Breast Cancer: A Multi-Organ Predictive Model Analysis
Presenter(s)
Z. Qiu, G. Sun, H. Fang, Y. Tang, Y. Zhai, H. Jing, S. Chen, X. Zhao, Y. Song, Y. Song, Y. Liu, B. Chen, S. Qi, Y. Tang, N. Lu, W. Zhang, X. Liu, Z. Yang, Y. LI, and S. Wang; Department of Radiation Oncology, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
Purpose/Objective(s): To identify clinical and dosimetric predictors of = Grade 3 radiation-induced lymphopenia (RIL) in patients with breast cancer undergoing postoperative radiotherapy and to develop a predictive model.
Materials/Methods: Clinical and dosimetric data from 80 breast cancer patients were analyzed. Organs at risk (OARs), including the lungs, heart, thymus, great vessels, and bone within the radiation field, were delineated. Dosimetric parameters, including mean dose (Dmean), maximum dose (Dmax), and relative/absolute volumes receiving 1-5Gy by 1-Gy increments and 5–45 Gy by 5-Gy increments, were extracted from dose-volume histogram (DVH). To address the inherent multicollinearity among dosimetric variables, a sequential approach of clinical-adjusted screening followed by stepwise Variance Inflation Factor (VIF) elimination was utilized. A multivariable logistic regression model was constructed using the identified core predictors and internally validated via Bootstrap resampling.
Results: The incidence of G3 RIL was 31.3% (25/80). Pre-radiotherapy peripheral lymphocyte count (Pre-PLC) showed a strong positive correlation with nadir PLC during radiotherapy(R = 0.55, P < .001). Univariate analysis revealed significant associations between G3 RIL and dosimetric parameters for the bilateral lung (V1–5, V25–45), bone (V1–5,V10,V40 and Dmean), thymus (V10, V15 and Dmax), and great vessels(V3-5, V10, V15, AV4, AV5-15 and Dmean). The final multivariable model identified three independent predictors: Pre-PLC (OR = 0.02, P<0.001), Bone V5 (OR = 1.27, P=0.01) and Lung V40 (OR = 1.72, P=0.037). Restricted cubic spline (RCS) analysis confirmed a non-linear relationship for Bone V5, identifying an optimal cut-off value of 35.47% for risk stratification. The model demonstrated excellent discriminative performance with an initial Area Under the Curve (AUC) of 0.866 and a bootstrap-corrected AUC of 0.847.
Conclusion: Pre-PLC, Bone V5, and Lung V40 are critical determinants of severe RIL in breast cancer patients. Specifically, Bone V5 > 35.47% represents a significant high risk for severe lymphopenia. To mitigate the risk of treatment-induced immunosuppression, radiotherapy planning should prioritize the optimization of Bone V5 below this threshold while maintaining lung dose constraints.