83 Late cardiac toxicity in survivors of esophageal cancer for myocardial necrosis and predicts the development of heart failure and overall survival as well [23]. Echocardiography was performed according to the guidelines of the European Association of Cardiovascular Imaging [24]. This protocol included assessments of right and left systolic and diastolic function parameters, strain imaging, valve disorders and signs of pulmonary hypertension. An ECG triggered CT-scan was performed on a dual source CT-scanner without contrast enhancement in order to quantify the number of coronary calcifications. This was calculated and expressed as the Coronary Artery Calcium (CAC) score based on the Agatston method [25]. A cardiac MRI scan was performed during breath hold and ECG monitoring on a 1.5 Tesla MRI scanner (Magnetom Avanto-fit, Siemens Healthineers, The Hague, Netherlands). T1 images were acquired with and without contrast enhancement in order to assess patterns of myocardial fibrosis and to enable T1 mapping to quantify myocardial abnormalities. Cine and delayed enhancement images (4 chamber, 2 chamber and short axis) were acquired for functional evaluation and measurements[26,27]. Results of the imaging techniques were assessed while being blinded for treatment group and medical history. In order to identify possible relationships between dose distribution parameters and diagnostic tests, detailed information on cardiac radiation dose distributions was collected. The radiotherapy planning CT scan, 3D treatment plan and delineated structures were transferred to the Mirada Medical treatment planning system (version 1.2.0). Additional contouring of substructures, and the left ventricular myocardial segments of the heart was subsequently performed according to previously published guidelines [28,29]. These retrospective data were exported to our research database. As this trial was designed as a pilot study, it was not powered for statistically significant (p<0.05) differences between the two groups. We consider differences up to a p value below 0.20 relevant for further analyses and worthwhile presenting. Binary endpoints were analysed using a logistic regression analysis, while for continuous endpoints a linear regression analysis was performed. To compensate for potential imbalances between the groups we tested and corrected for confounding variables. Mean values were used in presenting the data. 6
RkJQdWJsaXNoZXIy MTk4NDMw