ESTRO 2025 - Abstract Book

S3949

Radiobiology - Normal tissue radiobiology

ESTRO 2025

3977

Digital Poster External validation and updating of prediction model for Xerostomia in a local quality registry cohort Emmy Dalqvist 1 , Tiziana Rancati 2 , Anna Embring 3,4 , Gabriella Alexandersson Von Döbeln 4,5 , Claes Mercke 3,6 , Ingmar Lax 3,6 , Signe Friesland 3,6 , Eva Onjukka 1,3 1 Department of Nuclear Medicine and Medical Physics, Karolinska University Hospital, Stockholm, Sweden. 2 Data Science Unit, Fondazione IRCCS Instituto Nazionale dei Tumori, Milan, Italy. 3 Department of Oncology Pathology, Karolinska Institutet, Stockholm, Sweden. 4 Department of Radiotherapy, Karolinska University Hospital, Stockholm, Sweden. 5 Department of Clinical Science, Intervention and Technology, Karolinska Institutet, Stockholm, Sweden. 6 Department of Head, Neck, Lung and Skin Tumors, Karolinska University Hospital, Stockholm, Sweden Purpose/Objective: This study aimed to locally validate an NTCP model for xerostomia, enabling its future use in individualized NTCP based treatment planning. The validation cohort was selected from a local quality registry. Material/Methods: The model for xerostomia of grade ≥ 2 (6 months post radiotherapy) from the updated Dutch National Indication Protocol for Proton Therapy [1] was selected for validation. This TRIPOD Type 4 model was developed using a large cohort (750 patients for development and 395 patients for validation). It includes the mean dose to both the parotid- and submandibular glands, in addition to the baseline score for xerostomia. In the training cohort, the prevalence of the endpoint was 46%, assessed via the EORTC QLQ-H&N35 questionnaire. For our local validation cohort, data were extracted from the registry for patients treated between 2013 and 2022. The cohort was limited to patients with a relevant tumour location in relation to the salivary glands and a follow-up within two years. Patients with missing data were excluded, leaving a final cohort of 545 patients with an average follow-up of 12 months (range 6-24). Clinicians evaluated grading according to the RTOG/EORTC Radiation Morbidity Scales, and the prevalence of xerostomia grade ≥ 2 was 30%. As baseline data were unavailable in the validation cohort, a baseline score of zero was assumed for all patients. Results: The calibration curve demonstrated that the model predicted the dose-response relationship well, see Figure 1 (left). Notably, the local validation cohort showed a slightly stronger dose-response, with a slope of 1.12. However, the calibration intercept (calibration-in-the-large) revealed an overestimation of xerostomia, with an intercept of - 0.12. To address this, a closed testing procedure was employed for external calibration [2], indicating the need to recalibrate the model. After recalibration, the updated calibration curve showed an improved intercept and the slope was adjusted slightly, see Figure 1 (right). The log-likelihood value improved from -329 to -322 during the recalibration process. Both models are shown in Figure 2.