ESTRO 2020 Abstract Book

S116 ESTRO 2020

performance status, N stage, histology, and other-causes of death as a competing risk, for patients without PCD the mean dose to the heart-region (>10Gy) was associated with a higher risk of CD (adjusted HR=2.8; p<0.001, Fig. 1B). In patients with PCD, higher mean dose to the lung-region (as continuous variable) was associated with a higher risk of CD (adjusted HR=1.02; p=0.02, Fig. 1C).

Proffered Papers: Proffered papers 13: Lung

OC-0226 Pre-existing cardiac disease and cardiopulmonary RT dose associate with cardiac death in lung cancer A. Abravan 1 , C. Faivre-Finn 1 , K. Banfill 1 , A. McWilliam 1 , M. Van Herk 1 1 The University of Manchester/ The Christie NHS Foundation Trust, Division of Cancer Sciences/ Radiotherapy Related Research, Manchester, United Kingdom Purpose or Objective Cardiopulmonary toxicity is an adverse event following thoracic RT. It has generally been studied with overall survival as endpoint and without correcting for pre- existing cardiac disease (PCD). In this work we identified anatomical regions where the received RT dose correlates with cardiac death (CD), taking PCD into account. Material and Methods 1917 lung cancer patients treated with curative-intent RT (55Gy in 20 fractions) between 2010 and 2016 from a single institution were studied. Primary end point was CD occurring after day 1 of RT and scored based on relevant WHO-ICD10 codes on medical certificate of cause of death from Public Health England data. PCD were collected from Hospital Episode Statistics data prior to RT. To identify the most sensitive anatomical region matched cohort image- based data-mining was used. Two groups were defined with propensity score matching (1:2, caliper=0.2): patients who died of CD and patients who didn’t – matched on age, tumor volume, PCD and presence of infectious disease/pneumonitis prior to RT. To identify regions where dose correlates with CD, all dose matrices were aligned to a randomly chosen reference patient, mean dose distributions calculated for the matched groups and regions of significant difference detected. In the analyses, the whole cohort was used and those patients with cause of death recorded as ‘cancer’ only were treated as censored. Cumulative incidence estimates of CD, with other-causes of death as a competing risk, were compared by 2-sided p-value using Fine and Gray competing risk regression. Results From a total of 1653 deaths, 1243 patients had cause of death, PCD, and images available. Of these 377 died of known non-cardiac causes, and 175 of cardiac causes. PCD was present in 419 patients (34%). There was an increased risk of CD in patients with PCD (HR=4.24; p<0.001), with 2- year cumulative incidence estimates of 27% compared to 8% (Fig. 1A). Data mining identified a ‘heart-region’ close to the right atrium and a ‘lung-region’ peripheral in the right lung where dose was significantly higher for CD at 1 year (Fig. 2). After adjusting for sex, age, tumor volume,

Conclusion Patients with PCD are at higher risk of CD. In patients without PCD , dose to the right atrium correlates with risk of CD. In patients with PCD , peripheral right lung dose seems to have the largest impact on CD in line with literature on pulmonary hypertension. The absence of an effect of heart dose in this group may be due to the existing high risk of cardiac events, making it difficult to