ESTRO 2021 Abstract Book

S616

ESTRO 2021

We develop a Normal Tissue Complication Probability (NTCP) model for prediction of acute toxicity after radiotherapy (RT) for hea&neck cancer (HNC) with inclusion of RADIODTECT©, a blood assay derived from the radiation-induced nucleoshuttling of the ATM protein model. Materials and Methods A retrospective study was performed on 38 HNC patients (pts) treated with postoperative RT (IMRT/VMAT/TomoRT, median prescription dose 60Gy, range 50-70Gy, @2Gy/fr). Toxicity was prospectively graded with CTCAE scale and 4 endpoints were considered: grade≥3 (G3+) mucositis (OM3), G2+ OM (OM2); G3+ dysphagia (DY3) and G2+ DY (DY2). pATM quantification was assessed in lymphocytes. We considered 2 already published logistic models, which included: (i) the Equivalent Uniform Dose to the oral cavity (n=0.05, EUDOC) and the mean dose to the parotid glands (DPG) for OM, and (ii) EUDOC, the larynx EUD (n=0.35, EUDL) and the volume of pharyngeal constrictor muscles receiving>50Gy (V50CM) for dysphagia. From these models we derived a “weighted dose score” (WDS) as a linear combination of the dosimetric descriptors using their respective logistic beta-coefficients as weights. WDS was used as comprehensive dose feature to fit a logit model (characterized WDS50 , i.e. the WDS associated with 50% toxicity probability and k , the sigmoid slope) with/without inclusion of the information from the pATM test (i.e. including a “dose modifying factor”, dmf, which is the ratio between WDS50 for pts classified as radioresistant from the test and WDS50 for radiosensitive pts). Discriminative power was evaluated as AUC after internal validation using 10000 bootstrap resamplings. Results OM2 and OM3 was scored in 18 and 8 pts, respectively. DY2 and DY3 in 17 and 12 pts. Fig.1 presents the NTCP models for all endpoints, including a table with values for WDS50, k, dmf and AUC. In all cases the pATM biomarker proved to be effective in predicting radiosensitive pts (Fig.2), with dfm ranging from 0.88 for DY2 to 0.62 for OM2, and a gain in AUC (after correction for optimism, as derived from internal validation) with respect to the dosimetric model ranging from 5% to 31%. Of note, dose-response curves are steeper after the inclusion of the results from the pATM test, highlighting that the dose-response steepness in models not including any information on the single-patient radiosensitivity is a weighted average from radioresistant and radiosensitive patients.