ESTRO 2020 Abstract Book

S537 ESTRO 2020

tomography (PET) CT imaging study for each patient was used for segmentation. All FDG-avid intrathoracic tumors were delineated on the CT scan per RTOG contouring guidelines. Normal lung was contoured through thresholding in a commercial treatment planning system. The tumors were subtracted from the normal lung and radiomics were extracted with an in-house developed software for the resultant region of interest (ROI). 108 first-, second-, and third-order texture features within the ROI for each patient were analyzed for association with the primary endpoint. A univariate logistic regression model was used to estimate odds ratio (OR), the p-value, and receiver operating characteristic curves (ROC) for each imaging feature. Results Six out of 108 texture features showed significant association with CIP (p-values ranging from 0.02 to 0.03) and all exhibited large effect (OR < 0.3 or > 3.5). Despite the low number of events (only 9 out of 49) the ROC curves yielded values ranging from 0.789 to 0.794.

diagnostic threshold was tested using Receiver operating characteristics (ROC) curve. The chi square test was carried out to test the different risk factor groups against likelihood of grade ≥2 pneumonitis. Results Median age of the patients were 75 years with 52.6% female. Median follow up was 56 months. Sixty-five percent patients received 55Gy in 5-fraction regimen. 43.5% patients had ECOG performance status (PS) of 2 and 16.2% had PS of 3. Median Charlson comorbidity index was 6 (range 2-11). Median SUV max of tumour was 6.5. Only 31.8% had confirmed histological diagnosis. Median tumour size was 20mm (range 4mm-63mm). Median planning target volume (PTV) was 30.3 cc. Median values of R100, R50 and D2cm were 1.1, 5.6, 32.8Gy. Median value of mean lung dose(MLD), V20 and V12.5 were 3.9Gy, 5 %and 9.3%. 85(6.7%) patients developed symptomatic RP (grade ≥2) with only 5(0.4%) developing grade 3 RP. 5% of patients developed rib fracture but only 28% of these were symptomatic. On univariate analysis lower lobe tumour location, larger tumour size, PTV, MLD, V20 and V12.5 were significantly associated with grade ≥2 RP. On multivariate analysis only higher MLD was associated with grade ≥2 pneumonitis (Table 1). ROC curve analysis showed optimal diagnostic threshold for tumour size, PTV, MLD, V20 and V12.5 (Figure 1); are 22.5mm (AUC- 0.565), 27.15cc (AUC-0.58), 3.7Gy (AUC- 0.633), 4.6% (AUC-0.597), 9.5% (AUC-0.616).The incidence of ≥grade 2 RP were significantly higher for values higher than the ROC threshold.

Conclusion This preliminary study suggests that texture features on pre-immunotherapy CT imaging can be linked to CIP for patients with advanced NSCLC treated with Nivolumab. Future directions of this research include expansion of this study across the full database, correlation of the radiomics features with blood biomarkers, and the inclusion of tumor burden as an additional covariate in the analyses. PO-1007 Predictors of radiation pneumonitis in early stage lung cancer treated with SABR. A. Saha 1 , N. Hatton 1 , M. Beasley 1 , K. Franks 1 , P. Jain 1 , M. Teo 1 , K. Clarke 1 , P. Dickinson 1 , P. Murray 1 , J. Lilley 2 1 St. James Institute of Oncology- Leeds Teaching Hospitals- NHS trust, Clinical Oncology, Leeds, United Kingdom ; 2 St. James Institute of Oncology- Leeds Teaching Hospitals- NHS trust, Medical Physics, Leeds, United Kingdom Purpose or Objective Predictors of radiation pneumonitis (RP) in Stereotactic Ablative Radiotherapy (SABR), the standard treatment for medically inoperable peripheral early stage lung cancer, are poorly defined. In this study, we aimed to identify clinical and dosimetric parameters which can predict symptomatic RP in early stage lung cancer patients treated with SABR. Material and Methods 1266 patients treated with SABR in our institute (May 2009 -August 2018) were included. Electronic medical records were reviewed for baseline characteristics, treatment details and toxicity. Dosimetric data was extracted from XIO and Monaco software. Patients were treated according to the UK SABR consortium guidelines. RP was graded retrospectively using CTCAE version 4, based on available clinical and imaging information. Univariate and multivariate binary logistic regression were performed to determine predictive factors for grade ≥2 RP, using IBM SPSS statistics version 21 software. Goodness of fit was assessed using the Hosmer and Lemeshow test. Optimal

Conclusion SABR treatment resulted in very low rate of grade 3 pneumonitis. Lower lobe tumour location, larger tumour size, PTV, mean lung dose, V20 and V12.5 were found to be significant predictors of symptomatic radiation pneumonitis. PO-1008 Predictive dosimetric parameters of lymphopenia in stage III NSCLC patients with chemoradiation I.J. Cho 1 , Y. Kim 1 , S. Ahn 1 , W. Chung 1 , T. Nam 1 , M.S. Yoon 1 , J. Jeong 1 , J. Song 1 , S. Cho 1