S598
Clinical – Head & neck
ESTRO 2026
benefit for postoperative management. Radiotherapy remains a key modality for local control, particularly in patients with advanced disease or adverse pathological features. Keywords: salivary gland, chemoradiotherapy
Digital Poster 2490
Real-world outcomes of salivary gland carcinoma patients treated with radiotherapy at a single institution Gyöngyi Kelemen, Zsófia Máthé, Melinda Csenki, Zoltán Varga, Em ő ke Borzási, Viktor Paczona, Ferenc Borzák, Emese Fodor, Árpád Puskás, Anikó Maráz, Judit Oláh, Katalin Hideghéty Department of Oncotherapy, Albert Szent-Györgyi Clinical Center, Szeged, Hungary Purpose/Objective: Primary salivary gland carcinomas (SGCs) are rare, heterogeneous malignancies representing 4–5% of head and neck cancers. Their rarity, variable histology and biological behavior make optimal management challenging. This retrospective, real-world study aimed to evaluate clinicopathological characteristics and survival outcomes of SGC patients treated at University of Szeged, focusing on the role of radiotherapy. Material/Methods: Clinical and pathological data of patients with histologically confirmed SGC treated between 2008 and 2024 were retrospectively reviewed. Parameters included TNM stage, histological subtype, vascular invasion, and radiotherapy dose. Treatment response at 3 months, overall survival (OS), and progression- free survival (PFS) were assessed. Kaplan-Meier estimation with log-rank test was used for the statistical analysis. Results: Thirty-three patients were analyzed. Median age was 60.7 years (range 26.9–85). The most frequent subtype was adenoid cystic carcinoma (48.5%). Radiotherapy was delivered in 25 cases, 75.8% of patients, with a mean total dose of 63.9 Gy (range 37.5–72.1). Seven patients received concomitant cisplatin chemotherapy.After a median follow-up of 3.16 years, the mean overall survival (OS) was 11.9 years. subtype did not influence survival.Mean OS was 11.5 years in irradiated versus 7.2 years in observed patients (p=0.6). Despite more advanced disease and close margins in the radiotherapy group, survival tended to be longer, suggesting a possible therapeutic benefit. Distant metastases occurred in 11 (33.3%) and local recurrence in 3 patients (9.1%). Mean PFS was 10.1 years (9.1 years with radiotherapy vs 7.3 years without; p=0.33). If chemoradiotherapy was performed, the PFS was significantly better (p<0.05). Conclusion: This real-world analysis highlights the heterogeneity and therapeutic complexity of salivary gland carcinomas. Although radiotherapy did not significantly improve survival, similar or slightly better outcomes in higher-risk cases indicate potential
Digital Poster 2590
Deep learning based automatic whole-body muscle composition analysis for prognostic assessment in Head and Neck Cancer Yi Rong, Libing Zhu, Aman Anand, Nathan Y. Yu, Quan Chen Radiation Oncology, Mayo Clinic, Phoenix, USA Purpose/Objective: Sarcopenia is a well-established predictor of overall survival in head and neck squamous cell carcinoma (HNSCC), typically assessed by skeletal muscle index (SMI) at the L3 vertebral level. However, the L3 region is often excluded in routine head and neck imaging. This study aimed to (1) develop and validate a deep learning-based auto-segmentation model for whole- body skeletal muscle analysis, and (2) evaluate the prognostic utility of SMI and skeletal muscle density (SMD) at alternative vertebral levels for overall survival A two-dimensional nnU-Net model was trained using 617 multi-institutional CT scans and validated on 98 cases. Model accuracy was evaluated using Dice similarity coefficient (DSC), mean surface distance (MSD), and surface DSC (SDSC). External validation was performed using 215 public HNSCC cases from MD Anderson Cancer Center (2003–2013). SMI and SMD (mean Hounsfield Units) were computed at 22 vertebral levels. Correlation between AI- and manually derived areas was assessed using Pearson’s coefficient. Prognostic value for overall survival (OS) was determined using Cox regression. Results: The model achieved excellent segmentation performance with mean DSC of 0.97–0.98 across muscle, subcutaneous tissue, and bone, MSD of 0.04– 0.05 cm, and SDSC of 0.99 ± 0.01. AI- and manually segmented muscle areas were highly correlated (r = 0.95, p < 0.0001). In the external validation, SMI and SMD at thoracic (T11) and cervical (C2) levels were independently associated with OS (T11: HR = 2.66, 95% prediction in HNSCC. Material/Methods: CI 1.15–6.14; C2: HR = 2.22, 95% CI 1.19–4.22). No significant difference was observed between AI- derived and manual survival stratifications at L3.
Made with FlippingBook - Share PDF online