Risk estimation of cardiac toxicity following craniospinal irradiation of pediatric patients

Material and Methods: From 2006-2013 801 patients with early stage NSCLC were treated with CBCT guided SBRT (median 54 Gy in 3 fractions) in 5 institutes for whom treatment plans were available. 565 patients were analyzed after exclusion of synchronous or metachronous tumors (n=80), follow-up<1y (n=63), or death from cancer (93).An average anatomy was constructed based on 109 patients of the 5 institutes using deformable image registration. Subsequently, all patients were registered to this average anatomy and the corresponding dose distribution was deformed accordingly [1]. The heart and substructures right atrium, left atrium, right ventricle, left ventricle, superior vena cava, descending aorta and left pulmonary artery were contoured on the average anatomy. For each (sub)structure dosimetric parameters DV (V: 0 cc-max), VD (D: 0 Gy-max), EUDn (n: 0.1-10) were obtained.Associations of these dosimetric parameters with death were evaluated using univariate Cox regression. Per (sub)structure the parameter with the lowest Akaike information criterion was selected and used in subsequent analyses. Correlations between all (sub)structures were assessed prior to inclusion in a multivariate Cox regression. Finally, the (sub)structure(s) that remained significant in the first multivariate analysis were included in a second multivariate analysis, also including; performance status, age, gender, biological dose, distance to bronchus, comorbidity index, lung-function, tumor diameter, T-stage, institute and pack years smoking.