ESTRO 38 Abstract book

S517 ESTRO 38

Table 1 Results from univariate logistic regression analysis for clinical and dosimetric parameters. Clinical and

Odds ratio (95%CI) 0.32(0.14- 0.72) 1.11(1.00- 1.24) 1.06(0.53- 2.11) 1.24(0.61- 2.55) 1.28(0.60- 2.73)

P- value

dosimetric parameters

Coefficient Standard error

Age (years) (> 65 vs ≤ 65)

-1.141

0.413

0.006

PTVp-L (per cm increase ) 0.107

0.055

0.049

No. of pack- years (>18 vs ≤ 18 ) Alcohol (yes vs no)

0.055

0.354

0.877

0.218

0.365

0.550

Gender (female

vs

0.249

0,386

0.519

male)

Table 2 Predictive parameters from LASSO logistic regression analysis of the final model. Predictor Coefficient Odds ratio Age( > 65 vs ≤ 65) -0.528 0.59 PTVp-L (per cm increase ) 0.026 1.03 Lower thoracic ESCC -0.633 0.53 constant -1.254 Figure 1

Conclusion A multivariate NTCP model including age, PTVp-L, and tumor location predicts Grade ≥ 2 AET after radio(chemo)therapy for ESCC. The results of this study suggest that the risk of Grade ≥ 2 AET decreased in elderly or lower thoracic esophageal cancer patients while increased in patients with longer PTVp-L. PO-0955 Radiomics signature as predictors of survival and local control after pancreatic carcinoma SBRT L. Cozzi 1 , T. Comito 2 , A. Fogliata 2 , C. Franzese 2 , D. Franceschini 2 , C. Bonifacio 3 , A. Tozzi 2 , L. Di Brina 2 , E. Clerici 2 , S. Tomatis 2 , G. Reggiori 2 , F. Lobefalo 2 , A. Stravato 2 , P. Mancosu 2 , A. Zerbi 4 , M. Sollini 5 , K. Margarita 5 , A. Chiti 6 , M. Scorsetti 1 1 Humanitas Research Hospital and Humanitas University, Radiation Oncology Dept and Biomedical Science Dept, Milan-Rozzano, Italy ; 2 Humanitas Research Hospital, Radiation Oncology Dept, Milan-Rozzano, Italy ; 3 Humanitas Research Hospital, Radiology Dept, Milan- Rozzano, Italy ; 4 Humanitas Research Hospital, Pancreatic Surgery Dept, Milan-Rozzano, Italy ; 5 Humanitas Research Hospital, Nuclear Medicine Dept, Milan-Rozzano, Italy ; 6 Humanitas Research Hospital and Humanitas University, Nuclear Medicine Dept and Biomedical Science Dept, Milan-Rozzano, Italy Purpose or Objective Pancreatic adenocarcinoma has generally unfavorable prognosis. Combined chemio-radiotherapy is a frequently adopted treatment, especially for patients unfit to surgery, leading to possible overall survival at 2 years of about 30%. To better personalize the treatment, the ability of a radiomics signature to predict clinical outcome after stereotactic body radiation therapy has been explored. Material and Methods One-hundred patients have been selected from the institutional database, presenting histologically proven, inoperable locally advanced pancreatic cancer and treated in our department with SBRT with a total dose of 45 Gy in 6 fractions. All patients underwent planning CT with no contrast agent, where the CTV was delineated and accurately cleaned from confounding regions. Radiomics texture features were extracted for these CTVs using the LifeX software tools. A total of 41 features were computed. The statistical analysis was conducted with dedicated routines on R scripts. The cohort of 100 patients was separate in two groups of 60 and 40 patients, used for the training and validation processes, respectively. In the training phase, with the univariate Cox regression, the prognostic value of clinical and radiomic features was investigated to the predict

Figure 2