Abstract Book

S56

ESTRO 37

determine which toxicity is most relevant for survival and identify clinical and treatment related factors for both. Material and Methods This retrospective cohort study included 216 esophageal cancer patients referred for definitive (chemo)radiotherapy between January 2007 to December 2013. All patients were subjected to a follow up program consisting of regular follow up visits including esophageal endoscopy and CT scans. Hospital charts of all patients were reviewed and late toxicities were assessed in accordance with the CTCAE 4.0. Clinical factors as described in the table and dose volume parameters for whole heart, substructures of the heart and lungs were retrospectively assessed.

patient cohort however, pulmonary toxicity was more important for OS for which the lung_V45 was an independent predictor. These results indicate that reducing the cardiac dose at the expense of the dose to the lungs is probably not the right treatment planning strategy. PV-0104 Dosimetric Analysis of Hepatic Toxicity after Stereotactic Body Radiation Therapy R. Luo 1 , T.S. Su 2 , P. Liang 3 , T. Cheng 3 , Y. Zhou 4 , Y. Huang 4 1 University Medical Center Freiburg, Radiation Oncology, Freiburg, Germany 2 Affiliated Tumor Hospital of Guangxi Medical University, Department of Radiation Oncology, Nanning, China 3 Rui Kang Hospital- Guangxi Traditional Chinese Medical University, Cyberknife Center, Nanning, China 4 Rui Kang Hospital- Guangxi Traditional Chinese Medical University, Department of Radiation Oncology, Nanning, China Purpose or Objective To build and validate multivariate models of normal tissue complication probability (NTCP) and their corresponding nomograms for radiation-induced hepatic toxicity (RIHT) prediction after SBRT. Material and Methods Data were obtained from 85 HCC patients who participated in Phase II clinical trial of SBRT. Total dose 39-50 Gy was administered in 3-5 fractions on consecutive days. Eligibility criteria: (1) primary HCC; (2) pre-treatment Child-Pugh (pre-CP) class A or B; (3) 1 to 3 nodular HCC lesions with a total diameter < 10 cm; Exclusion criteria: (1) portal vein thrombosis and extrahepatic metastases; (2) recurrence after liver resection or transplantation. After SBRT, a progression of at least 1 or 2 points in CP score was classified as RIHT (≥1) or (≥2). Two dosimetric datasets (V x the percentage (%) of normal liver volume receiving x Gy or more; VS x , the absolute normal liver volume (mL) spared from at least x Gy) were collected and used to build NTCP models separately. NTCP models for RIHT (≥1) or (≥2) based on clinical and dosimetric parameters were developed using logistic regression with cross-validation. Nomograms for each NTCP model were formulated. In the additional analysis, Youden's index in conjunction with the receiver operating characteristic (ROC) analysis was applied to obtain cut-off points for each independent dosimetric risk factor. External data is extracted from an independent cohort (101 pre-CPA patients) to validate our NTCP model. The validation study applied RIHT (≥2) as the only endpoint. Results RIHT grades (=0, =1, ≥2) was found to be the prognostic factors for overall survival. Twenty (23.5%) patients and 12 (14.2%) patients experienced RIHT (≥1) and RIHT (≥2), respectively. V 15 , VS 10 and pre-CP were found to be the independent risk predictors for both RIHT (≥1) and RIHT (≥2). V 15 ≤ 33.1% and VS 10 ≥416.2 mL to be the cut-off points for RIHT (≥ 1) risk stratification, and V 15 ≤ 21.5% and VS 10 ≥ 621.8 mL for RIHT (≥2) were derived from ROC analyses. Four NTCP models (Table 1) based on their own risk factors and four corresponding nomograms were generated. Our NTCP models showed good model performance (AUC, 0.83-0.89). V 15 were extracted from the external cohort and our NTCP model and its corresponding nomogram (based on V 15 plus pre-CP) showed a good prediction performance (AUC = 0.78) for RIHT (≥2) in this validation cohort.

Results With a median follow up of 27 months, 97 out of 216 patients developed locoregional recurrences. The median disease-free survival (DFS) was 64 months (95% CI 58, 7- 69, 3 months). Radiation pneumonitis (RP) was scored in 60 patients. Most lung DVH parameters correlated with this clinical endpoint. In a multivariable logistic regression analysis, the mean lung dose and the lymph node status remained significant prognostic factors for the development of RP (p<0.05). Cardiac complications were diverse; the most frequently occurring complication was pericardial effusion (PE) (69 patients). The dose to the whole heart, the pericardium and the right ventricle all correlated highly with this endpoint. In the multivariable logistic regression analysis, the V35 of the right ventricle and the N-status remained significant prognostic factors for PE (p<0.05). A Cox regression analysis was performed to inv estigate the impact of cardiopulmonary toxicity on OS. In univariable analysis, most lung dose parameters, but hardly any cardiac dose parameter correlated with OS. In the multivariable analysis, a higher V45 of the lungs (Lung_V45) and a higher UICC stage significantly correlated with worse OS. None of the cardiac dose volume parameters remained significant. Conclusion Cardiac dose volume parameters predicted the risk on cardiac toxicity endpoints and pulmonary dose volume parameters did so on radiation pneumonitis. In this