Predicting and Monitoring for Cardiac Toxicity in Pediatric AML

点击翻译按钮获取中文摘要

Abstract: Children with acute myeloid leukemia (AML) receive the maximum cumulative anthracycline exposure during frontline therapy (>440 mg/m2), an exposure that causes very well-described cardiac complications. The importance of acute, short-term cardiotoxicity in pediatric AML has been described recently by our group using Children’s Oncology Group (COG) clinical trial data. Specifically, nearly 40% of patients experienced left ventricular systolic dysfunction (LVSD) warranting chemotherapy modifications, and 21% suffered LVSD consistent with moderate to life-threatening cardiomyopathy or heart failure. Over 70% of LVSD was first documented during frontline therapy and was associated with a 13% absolute decrease in overall survival (OS). Notably, the decline in OS from early cardiotoxicity is larger than the improvement from any randomized intervention reported to date in pediatric AML cooperative group trials. Despite this clinical impact, there are no clinical prediction models for early chemotherapy associated cardiac toxicity. The cardiotoxicity prediction models developed to date focus exclusively on long-term childhood cancer survivors due to the absence of a sufficiently large and well-characterized de novo AML cohort. The variability in guidelines and clinical practice is fertile ground for disparities in echo monitoring and identification of cardiac outcomes, yet the well described disparities in survival outcomes by race/ethnicity in pediatric AML have not included analyses of echo adherence or cardiac dysfunction.With this application, we propose to develop a highly unique dataset including detailed demographic, clinical, genomic, treatment, and toxicity data, combined with longitudinal indices of LV size, diastolic, and systolic function from all clinical surveillance echocardiograms for patients enrolled on two COG AML trials, AAML0531 and AAML1031. With this unique data set, we will use novel, sophisticated analytic methods to develop models that: (1) continually update predictions of early cardiotoxicity risk during frontline therapy, (2) predict LV functional trajectories leading to persistent or worsening LVSD in the early/moderate time window after AML therapy, and (3) compare cancer treatment outcomes (EFS and OS) and detection of LVSD for three cardiotoxicity monitoring schedules with different echo frequencies in pediatric AML patients. The first model should enable personalized chemotherapy modifications and earlier initiation of cardiac-directed medications, thus improving survival outcomes. The second model should inform more personalized, evidence-based recommendations for off-treatment cardiotoxicity surveillance, potentially leading to improved adherence and reducing costs of unnecessary testing. The third analyses will provide a data driven guidance for on-therapy echocardiogram monitoring for pediatric patients with AML.