ESTRO 2024 - Abstract Book

S1304

Clinical - Head & neck

ESTRO 2024

3. Adam MA, Thomas S, Hyslop T, Scheri RP, Roman SA, Sosa JA. Exploring the Relationship Between Patient Age and Cancer-Specific Survival in Papillary Thyroid Cancer: Rethinking Current Staging Systems. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. Dec 20 2016;34(36):4415-4420. doi:10.1200/jco.2016.68.9372 4. Gao J, Chen C, Tao YL, Wang XH, Chang H, Li XH, et al. New suggestion for clinical downstaging of nasopharyngeal carcinoma in the era of intensity-modulated radiotherapy. Chin J Radiat Oncol. 2017;26(6):614-620. doi:10.3760/cma.j.issn.1004-4221.2017.06.002

1535

Digital Poster

Development of individual prognostic model to predict disease related outcomes in oral cavity cancer

Divyani Chowdhury, Sanjoy Chatterjee, Indranil Mallick

Tata Medical Center, Radiation Oncology, Kolkata, India

Purpose/Objective:

Oral cavity cancers contribute to a significant disease burden across all ages and populations from a global perspective with increasing incidence in South Asian countries. Prognostication of oral cavity cancers is usually based on the AJCC/TNM staging system that is based on tumor size, depth and nodal metastases. However it is well known that other clinical and pathological factors (i.e. differentiation, perineural invasion, lymph vascular invasion, surgical margins etc.) also affect the prognosis. Our goal was to develop an individualized risk prediction model for patients treated with surgery and adjuvant therapy and to assess model performance based on discrimination and calibration indices.

Material/Methods:

Clinical and pathological features and outcome information of 937 patients treated at a single institution was used for modelling. The sample size was formally calculated based on published guidelines on outcome prediction models for time-to-event data. To prevent testimation bias, a fully pre-specified model using twelve parameters - age, primary site, maximum tumour dimension, depth of invasion, differentiation, margin status, lympho vascular invasion, perineural invasion, bone involvement, total number of positive nodes, extra nodal extension and number of nodes with extracapsular extension was used for model development. The primary prediction metric was 2 year disease-free survival. The Cox Proportional Hazards assumption was not met for several predictor variables. Thus the accelerated failure time model was used after testing assumptions. Modelling was performed using the rms package of the R statistical software. Restricted cubic splines were used to allow non-linearity in continuous variables. Instead of arbitrary test and validation cohorts within the same institution, rigorous internal validation was performed with bootstrapping (300 bootstraps) and optimism-corrected model evaluation metrics were generated.