ESTRO 2021 Abstract Book

S120

ESTRO 2021

Conclusion This work validates previous image-based data mining studies in identifying a cardiac region within the base of the heart as strongly associated with overall survival. Importantly, for the first-time baseline cardiac function was included with mean dose to the identified region in the multi-variable analysis. Poor ejection fraction did not negate the impact of dose to the base of the heart on survival. OC-0191 Cardiac death relates to cardiac admission and left anterior descending artery RTdose in lung cancer A. Abravan 1 , C. Faivre-Finn 1 , J. Khalifa 2 , K. Banfill 1 , A. McWilliam 1 , M. van Herk 1 1 The University of Manchester, Radiotherapy related research, Manchester, United Kingdom; 2 Institut Universitaire du Cancer de Toulouse, Radiation Oncology, Toulouse, France Purpose or Objective There is emerging evidence that cardiac dose has a negative impact on overall and cardiac-specific survival following thoracic RT. In this work, we studied the association between cardiac substructure doses and cardiac admission post-RT and their impact on death due to a cardiac cause. Materials and Methods 1977 lung cancer patients treated with curative-intent RT (55Gy in 20 fractions) between 2010-2016 with available RT planning images from a single institution were identified. Primary endpoint was cardiac admission post-RT defined as hospital admission for ischemic heart disease, arrhythmia, and heart failure and secondary end point was cardiac death (CD). Cardiac admission post-RT as well as medical history prior to RT including ischemic heart disease, arrhythmia, heart failure, stroke, hypertensive disease, and artery disease were obtained based on the relevant WHO-ICD10 codes from Hospital Episode Statistics. CD was identified based on death certificates from Public Health England recorded after day 1 of RT. For each patient, planning CT scan was deformably registered to 3 template patients with 13 cardiac substructures segmented. Then, mean, and maximum dose to each substructure were calculated and averaged over 3 templates. For each endpoint, substructure doses along with clinical factors were bootstrapped 500 times into elastic-net regression. Variables with highest magnitude and frequency were included into the Cox model. Cox regression, with time- varying covariance, was used for predicting CD to incorporate the time at which cardiac admission happens during follow-up. Results 1218 patients with no records of cardiac admission at baseline were included in the analysis. 402 patients (33%) were admitted for cardiac causes post-RT. Of 984 patients with death certificate available, 79 (8%) have died of cardiac causes. Following variable selection, a cox model for cardiac admission was proposed including mean dose to left anterior descending artery (LAD) (HR=1.02, p=0.001), age (HR=1.04, p<0.001), log tumor volume (HR=2.81, p<0.001), adjusted for gender (HR=1.15, p=0.2), hypercholesterolemia (HR=1.25, p=0.1), and mean lung dose (HR=0.99, p=0.1). Cox model for CD, with cardiac admission as time-varying covariance, shows a significant association between CD and cardiac admission post-RT (HR=14.97, p<0.001) (Figure 1), mean LAD dose (HR=1.04, p<0.001), adjusted for age (HR=1.02, p=0.05), gender (HR=1.58, p=0.05), and artery disease (HR=1.6, p=0.2).