Simulation CT Features and Radiation Cardiotoxicity in Non-Small Cell Lung Cancer

Abstract

Radiation cardiotoxicity is a significant clinical dilemma in non-small cell lung cancer (NSCLC) radiation therapy (RT). Baseline cardiovascular (CV) status may influence the risk of cardiotoxicity, and may be ascertainable from the appearance of the heart on simulation computed tomography (CT). We examined the association of CT features with incidental heart dose and risk of cardiac events in NSCLC. Patients treated with curative-intent RT between 2015 and 2020 at a regional center were identified. Clinical notes were interrogated for baseline patient and CV health details, and follow-up CV events. Cardiac events were verified by a cardiologist. A deep learning-based auto-segmentation tool was applied, allowing extraction of a pre-specified list of volume parameters in a programming environment. CAC was graded as none, mild, moderate and severe in patients with a non-contrast scan. The craniocaudal relationship of the PTV and heart (Feng atlas) were annotated. A total of 478 patients were included, with a median age of 70 and Charlson Index of 5. The median mean heart dose was 6.3 Gy (IQR 2.7-11.4). The median lung V20 was 20.0% (IQR 14.8-27.1). Cardiovascular risk factors were common, with most patients having 2 (39%) or 3 (31%). A history of previous cardiac events was common, including myocardial infarction (14%), arrhythmia (11%) or heart failure (9%). A total of 6.9% and 7.1% patients developed a new atrial arrhythmia (AA) or heart failure (HF) after completing RT. The volume metrics with the highest AUC for AA and HF events were the left atrium (LA) (AUC 0.67, p = 0.0002) and left ventricle:right ventricle (LV:RV) ratio (AUC 0.66, p = 0.0021). Kaplan-Meier analysis for cardiac events dichotomizing at the optimal cut-point for maximum sensitivity and specificity demonstrated significantly different rates for both AA (LA 109cc, HR 3.35, 95% CI 1.64-6.83, p = 0.0009) and HF (LV:RV ratio 1.61, HR 2.37, 95% CI 1.19-4.74, p = 0.0143). Only 2 patients with non-contrast scans developed a myocardial infarction, both had mild CAC. The incidence of pooled cardiac events was not significantly different between patients with no (n = 2/21, 9.5%), mild (n = 10/38, 26.3%), moderate (n = 8/53, 15.1%) and severe (n = 7/24, 29.2%) CAC (p = 0.3916). Where the inferior border of the PTV was above the superior border of the heart, mean heart dose was significantly lower than compared with overlap of levels (1.9 Gy v 9.7 Gy, p<0.0001), and this was true for 3DCRT (n = 139, p<0.001), IMRT (n = 94, p<0.001) and VMAT (n = 145, p<0.001) patients. LA volume and LV:RV volume ratio are predictive for the development of AA and HF respectively. CAC grade did not differentiate patients by risk of cardiac events. Where the craniocaudal level of the PTV doesn't overlap with the level of the heart, the cardiac dose is likely to be very low. Several simulation CT features are associated with cardiac events following treatment for NSCLC and prospective evidence of cardiac risk could enable medical optimization prior to RT.