Dosimetric Predictors of Cardiac Events After Concurrent Chemoradiotherapy for Locally Advanced Esophageal Cancer

Abstract

To investigate the predictive value of cardiac substructure dosimetric parameters for cardiotoxicity and overall survival (OS) in patients with esophageal cancer receiving concurrent chemoradiotherapy (CCRT).We retrospectively reviewed the records of 213 patients with thoracic esophageal cancer receiving CCRT from 2010 to 2016. The primary endpoint of this study was grade ≥ 3 cardiac events. Dose-volume histogram (DVH) parameters for whole heart (WH), left atrium (LA), left ventricle (LV), right atrium (RA), right ventricle (RV), left anterior descending (LAD), left circumflex (LCX), right coronary artery (RCA), and pericardium were extracted. Baseline cardiac status was assessed by noting whether patients had pre-existing cardiac disease. For those without pre-existing cardiac disease, the World Health Organization/International Society of Hypertension (WHO/ISH) risk score was calculated to assess patients' baseline cardiac risk. Competing risk analysis and Cox regressions analysis were performed.Median follow-up was 27 months. Forty-nine patients (23.0%) had pre-existing cardiac disease. Thirty-eight patients (17.8%) developed 42 grade ≥ 3 cardiac events, at a median of 12.5 months to first event. The most frequently observed cardiac events were acute coronary syndrome (ACS, n = 14), congestive heart failure (CHF, n = 10), and arrhythmia (n = 9). 89.5% of the first grade ≥ 3 events (n = 34) occurred within the first three years after CCRT. The best multivariable model based on lowest AIC (348.6) included the following covariates: pre-existing cardiac disease, age, and mean LCX dose. At 36, 60, and 100 months, the AUCs for this best model were 0.789, 0.762, and 0.849, respectively. For those without pre-existing cardiac disease (n = 164), variables incorporated in the optimal multivariable model (AIC = 155.4) were WHO/ISH 10-year risk and LV V20. At 36, 60, and 100 months, the AUCs for this optimal model were 0.682, 0.666, and 0.834, respectively. 10.3% developed one or more grade ≥ 3 ACS/CHF events. Under multivariate analysis, pre-existing cardiac disease and mean LCX dose were included in the optimal ACS/CHF events predicting model (AIC = 198.8). Pericardium V15 and LA V55 were significantly associated with pericardial and arrhythmic events, respectively. Grade ≥ 3 cardiac events and LA V45 were independent predictors for OS.Whether for all cardiac events or the specific ACS/CHF, pericardial, or arrhythmic events, the predictive performance of corresponding substructure doses outperformed the WH doses. Grade ≥ 3 cardiac events and higher LA dose were associated with worse OS.