The Predictive Value of Changes in Basal Myocardial F-18 Fluorodeoxyglucose Uptake for Cardiotoxicity in Locally Advanced Esophageal Cancer Patients Receiving Definitive Radiotherapy

Abstract

To investigate the predictive value of changes in myocardial 18F-FDG uptake for major adverse cardiac events (MACEs) in locally advanced esophageal cancer patients receiving definitive radiotherapy. Between August 2012 and January 2018, 400 patients with stage II-III esophageal cancer receiving definitive radiotherapy at two institutions were divided into the training (n = 240) and external validation cohorts (n = 160). All patients underwent FDG-PET imaging within 1 week before treatment and 3 months after treatment. Myocardium delineation was performed by Carimas software (version 2.10) based on the AHA 17-segment model. When contouring the left ventricle, the myocardium was automatically divided into basal (segments 1-6), middle (segments 7-12), and apical (segments 13-16) regions, and the mean dose and FDG uptake parameters of each region were obtained by Carimas. Our primary endpoint was MACEs. Patient clinicopathologic factors, dosimetric parameters for the whole heart and cardiac substructures, and myocardial changes within the three regions on 18F-FDG PET were utilized to seek the best predictive models for cardiotoxicity. To avoid multicollinearity between dose-volume histogram (DVH) parameters, we selected the variables with the lowest Akaike Information Criterion (AIC) value from the DVH parameters of the same cardiac structure for the actual modeling procedure. Competing risk analysis and Cox regressions analysis were performed. The predictive performance of the models was evaluated using the area under the receiver operating characteristic curve (AUC) and Brier score. At a median follow-up interval of 78 months, 28 patients (11.7%) developed MACEs. The basal region of the myocardium received the highest radiation dose, followed by the middle and the apex region. The basal myocardial SUVmax and SUVmean significantly increased after radiotherapy while the apical and middle myocardial SUVmax and SUVmean not significantly increased. In univariate analysis, age, pre-existing cardiac disease, changes in pre- and post-treatment basal myocardial SUVmax and SUVmean (∆SUVmax and ∆SUVmean), and dosimetric parameters for MHD, mean LCX, mean LAD, and mean LV dose were associated with an increased hazard of MACEs. Multivariate analysis showed that basal ∆SUVmean retained significance after adjusting for age, pre-existing cardiac disease, and dosimetric parameters for whole heart and cardiac substructures. The AUCs and Brier scores demonstrated favorable predictive accuracies of the model's integrating variables with significant difference in multivariate analysis when predicting MACEs in the training and validation cohorts. ∆SUVmean was an independent indicator of MACE in locally advanced esophageal cancer patients receiving definitive radiotherapy. Changes in basal myocardial FDG uptake is a promising biomaker for predicting radiation-induced cardiotoxicity.