S647
Clinical – Head & neck
ESTRO 2026
next step, the aim is to further predict the occurrence of radio therapy-related toxicities from the longitudinal immune status of the patients by the aforementioned machine learning models. First data on that analysis will be presented. Conclusion: This study highlights the value of immune profiling in HNSCC for the identification of prognostic biomarkers. The discovered immune signature demonstrates potential for predicting DFS in patients undergoing R(C)T and may contribute to future approaches in personalized therapy. Further validation in larger independent cohorts is necessary for clinical translation. Keywords: immune monitoring, biomarkers, machine- learning long-term outcomes following selective clinical target volume optimization in neck level III for nasopharyngeal carcinoma with IMRT Ting Qiu 1 , rong li Wu 2 1 Radiation Oncology, The Affiliated Cancer Hospital of Nanjing Medical University, nanjing, China. 2 Radiation Oncology, Jiangsu Cancer Hospital, nanjing, China Purpose/Objective: This study aims to assess long-term efficacy and radiotherapy-related toxicity of an institutional protocol that optimizes the dose and delineation scope of level III target volume for nasopharyngeal carcinoma (NPC) patients based on cervical lymph Digital Poster 4084
Radiotherapy, Ordensklinikum Linz Barmherzige Schwestern, Linz, Austria. 12 Medical Faculty, Johannes Kepler University, Linz, Austria. 13 Department of Radiotherapy and Radiation Oncology, Kliniken Maria Hilf, Moenchengladbach, Germany. 14 Department of Radiation Oncology, Chemnitz Hospital, Chemnitz, Germany. 15 Department of Radiation Oncology, Universitätsklinikum Erlangen, Erlangen, Germany Purpose/Objective: Immunological biomarkers are gaining increasing significance in the assessment and management of solid tumors. However, in the treatment of head and neck squamous cell carcinoma (HNSCC), particularly those biomarkers derived from peripheral blood play a minor role to date. To address this limitation, the prospective DIREKHT01 study (NCT02528955) included a translational research program aimed at longitudinally monitoring patients’ immune profiles. The objective was to identify predictive immune signatures for prognosis and therapy optimization in HNSCC. Material/Methods: The immune status of 70 HNSCC patients (including 21 with oral cavity, 48 with oropharyngeal tumors (41 HPV+ tumors) and 1 with laryngeal tumor) was assessed before and after concurrent chemoradiotherapy (R(C)T), as well as during follow- up. A flow cytometry-based assay was used to analyze 45 immune parameters per time point from whole blood samples. The data served as input for a machine learning model designed to predict disease-free survival (DFS). The model development included data- driven variable selection, followed by modelling with Logistic regression, support vector machines and XGBoost. The model's validity was verified through multiple rounds of nested cross-validation. Results: By evaluating the immune status before and after R(C)T, a signature of various immune parameters predictive for the DFS were identified. Key contributors to this signature were HLA-DR positive T cells and monocytes after RT, pre-treatment frequencies of basophils and natural killer cells, and the frequency of HLA-DR-positive monocytes prior to R(C)T. The resulting model achieved a Matthews correlation coefficient (MCC) of 0.681 and was equally informed by pre- and post-treatment immune parameters, highlighting the importance of monitoring immune
node characteristics. Material/Methods:
A retrospective analysis was conducted on newly diagnosed, non-metastatic NPC patients treated with intensity-modulated radiotherapy (IMRT) between 2012 and 2020. Patients treated with our individualized target volume protocol (modified group) were compared to those receiving standard radiotherapy (comparison group). The modified protocol stratifies risk based on cervical lymph node characteristics to guide dose levels (50.4 –70 Gy): level III high-risk area receives 60 Gy, low-risk area 50.4 Gy, equivocal lymph nodes 60 Gy, positive lymph nodes without extracapsular extension (ENE) 64–70 Gy, and those with ENE 66–70 Gy. Additionally, low-risk lymph nodes anterior to the cervical vascular sheath and the carotid artery were selectively spared. Baseline characteristics of the two groups were balanced via propensity score matching (PSM), followed by survival analysis and evaluation of toxicities. Results: A total of 1107 patients were included (469 modified
dynamics. Both adaptive and innate immune parameters contributed to the signature.
Incorporating clinical variables did not improve model performance (MCC = 0.678), but yielded an alternative model of comparable accuracy that combined post- treatment immune markers with clinical characteristics. Among the predictive clinical factors, concurrent CT was a key determinant of DFS.In the
Made with FlippingBook - Share PDF online