Validation of the 18F-FDG PET Image Biomarker Model Predicting Late Xerostomia after Head and Neck Cancer Radiotherapy

Purpose/Objective(s) PET image biomarkers can improve the prediction of radiation-induced late xerostomia in head and neck cancer (HNC) patients [1,2]. A biomarker, 90th percentile (P90) standardized uptake value was identified as a promising predictor for xerostomia, where HNC with lower metabolic activity pre-radiation parotid gland (PG) were more at risk developing after radiation [2]. Addition P90 improved model performance reference based on baseline (Xer-base) PG mean dose. The aim current study to validate pre-treatment PET-P90 model an external validation cohort. Materials/Methods treated primary (chemo-)radiation curative intent. 18F-FDG PET/CT images acquired 2 weeks before radiotherapy. main endpoint patient-rated moderate-to-severe 12 months radiotherapy which assessed via EORTC QLQ-H&N35; questionnaire. dose contralateral extracted from distribution, respectively. We validated that Xer-base only, well PET-P90 model, had additional variable. closed testing procedure used test whether original coefficients (see [2]) these models needed be updated or revised Finally, (AUC, R2) evaluated compared model. Calibration tested Hosmer-Lemeshow (HL test). Results cohort consisted 127 40% developed xerostomia. indicated all did not require any update revision Compared (Table 1. AUCref=0.67, R2ref=0.13), showed better predictive (AUCP90=0.71, R2P90=0.18). both good test: pref=0.09, pP90=0.09). Conclusion performed this cohort, outperformed This contributes hypothesis is factor