ESTRO 38 Abstract book

S665 ESTRO 38

curve analysis was applied to estimate the predictive accuracy of dosimetric threshold values for G3/G4 RD.

patients (p=0.03). There were no significant associations between ECE and gender, age at diagnosis and pathological tumour grade. Pre-operative CT scans of 80 patients were examined and 136 metastatic lymph nodes were matched to pathological data. Of this data set, 30 (22%) lymph nodes were positive for ECE. There was a non-statistically significant trend towards more incidence of ECE seen in matted nodes against discrete nodes (27% vs. 15%, p=0.14) and those with necrosis (27% vs. 14%, p=0.11). No association was found between ECE and size or eccentricity of LN as well as smoothness or distinctness of LN borders. Radiological features such as calcified and cystic lymph nodes were too rare to have statistical significance. Conclusion We were able to demonstrate a statistically significant association between presence of ECE and number of metastatic lymph nodes as well as p16 status. Analysis of radiological features did not reveal any associations with ECE. Further analysis will be required to see if a combination of clinical and radiological features is best at predicting presence of ECE. EP-1200 Is skin dose distribution a predictive factor for the development of severe radiation dermatitis? P. Bonomo 1 , C. Talamonti 2 , L. Marrazzo 2 , I. Desideri 1 , D. Pezzulla 1 , L. Dominici 1 , A. Rampini 3 , S. Bertocci 3 , R. De Majo 3 , C. Gasperi 4 , A.S. Curion 4 , L. Lastrucci 3 , S. Pallotta 2 , L. Livi 1 , S. Caini 5 1 Azienda Ospedaliero-Universitaria Careggi, Radiation Oncology, Florence, Italy ; 2 Azienda Ospedaliero- Universitaria Careggi, Medical Physics, Florence, Italy ; 3 Ospedale San Donato, Radiation Oncology, Arezzo, Italy ; 4 Ospedale San Donato, Medical Physics, Arezzo, Italy ; 5 Cancer Research and Prevention Institute ISPO, Cancer Risk Factors and Lifestyle Epidemiology Unit, Florence, Italy Purpose or Objective In the context of locally advanced HNSCC, no biomarkers are available to predict the development of radiation dermatitis (RD). In the absence of putative skin dose constraints, alarge variability exists in IMRT planning optimization to avoid the potential risk of undue toxicity. The aim of our work was to assess whether skin dose- volume parameters may be predictive of severe RD. Material and Methods We retrospectively reviewed consecutive patients treated at two fellow institutions between Oct. 2011 and Nov. 2017 for locally advanced HNSCC. A 2:1 frequency- matched cohort analysis was performed to identify patients with similar demographic and disease-features who received cisplatin or Cetuximab. On the native planning CT scans, 3 skin ring structures were semi- automatically delineated by subtracting 2, 3 and 5 mm below the external patient surface (figure 1: 5 mm example), extending between the upper and lower limits of PTV plus a fixed 1 cm margin. For all 3 skin ring ROI’s, the following dosimetric parameters were collected: V50, V60, maximum absolute dose expressed in terms of point value and percentage of total prescription dose. Acute toxicity was prospectively recorded according to CTCAE v. 4.1. A total dose of 66-70.5 Gy was delivered in 30-35 fractions with conventional or accelerated fractionation (dose per fraction, 2/2.12/2.35 Gy, respectively). A 3-5 PTV margin was added to CTV’s (3 mm in case of daily image guidance). HPV status was not routinely available until 2015. To assess whether any patient, disease, treatment and skin dose distribution – related characteristics correlated with the development of G3/G4 RD, categorical and continuous variables were tested with Fisher’s exact chi-square test and Mann-Whitney test, respectively. A multivariate Poisson regression analysis was performed when multiple risk factors with a p value <0.05 were identified in the univariate analysis.A ROC

Results Ninety-patients were evaluated (table 1). G3/G4 RD, oral mucositis and dysphagia developed in 37 (41.1%), 40 (44.4%) and 24 (24.4%) patients, respectively. As expected, the G3/G4 RD was numerically higher in the cetuximab group (50% vs 36.6% in the cisplatin cohort; p=0.122). In univariate analysis, PS >1, weight loss at end of RT > 10 Kgs, mean relative dose intensity of Cetuximab, 2 mm skin ring V50 and V60 were significantly correlated with G3/G4 RD. In multivariate analysis, PS >1 and marked weight loss were the only 2 factors that retained significance (p=0.028 and 0.030, respectively). The best predictive accuracy of skin ring ROI’s for G3/G4 RD was provided by 19.9 cc of 2 mm skin ring at 50 Gy and 5.8 cc of 2 mm skin ring at 60 Gy (both AUC 0.61).

Conclusion Along with known risk clinical factors occurring during RT such as a marked weight loss, the extent of a 2 mm skin ring ROI receiving 50 and 60 Gy may help predict beforehand the development of severe RD.