Early cardiotoxicity and survival in pediatric Acute Myeloid Leukemia: The mediating roles of relapse and salvage chemotherapy intensity

Abstract

Abstract Background Treatment for pediatric acute myeloid leukemia (AML) involves intensive anthracycline (ATC)-based chemotherapy, with cardiotoxicity being a common complication. While early cardiotoxicity has been associated with worse overall survival (OS) in clinical trial cohorts, evidence from large real-world cohorts remains limited. Furthermore, the role of relapse as a mediator of this relationship and the influence of early cardiotoxicity on the choice of salvage therapy at relapse remain unclear. We aimed to (1) estimate the direct and indirect (i.e., through differential occurrence of relapse) effects of early cardiotoxicity on OS, and (2) explore two potential pathways underlying the indirect effect, namely differences in relapse risk and selection of reinduction regimens at relapse. Methods We used data from the REAL–AML Cohort, a retrospective cohort of pediatric AML patients (<19 years) treated at 17 U.S. institutions from 2011 onwards merged with the Pediatric Health Information System (PHIS) database. Patients with de novo AML were included with follow-up through five years. Early cardiotoxicity within one year of initiating frontline chemotherapy was considered as a time-varying exposure, defined as ejection fraction <55% or shortening fraction <28%, consistent with ATC modification thresholds on the previous COG trials. We used time-varying Cox proportional hazards models to examine pairwise relationships between early cardiotoxicity, relapse, and OS. Time-varying causal mediation analysis using g-computation was applied to estimate direct and indirect effects (through relapse) of early cardiotoxicity on OS, adjusting for both time-fixed and time-varying confounders. Among patients with relapsed AML, we assessed whether early cardiotoxicity influenced the selection of reinduction regimen (i.e., intensive ATC-based, intensive non-ATC, or non-intensive), the impact of reinduction regimen intensity on post-relapse survival, and the potential effect modification by history of early cardiotoxicity. As an extension, we conducted stratified analyses by era (pre-2017, post-2018) defined by the marked increase in use of dexrazoxane for primary prevention of ATC-associated cardiotoxicity in the pediatric AML patient population (average rates of 21.3% pre-2017 vs. 71.5% post-2018). Results Among 908 patients over a median follow-up of 51.8 months (IQR: 25.3–60.0), 247 (27.2%) patients developed early cardiotoxicity, 285 (31.4%) experienced relapse, and OS was 71.9%. Early cardiotoxicity was associated with increased hazard of relapse (HR = 1.31, 95% CI: 1.00–1.71) and worse OS (HR = 1.82, 95% CI: 1.37–2.43). The total effect of early cardiotoxicity corresponded with a 6.8% absolute increase in 5-year mortality (95% CI: 0.0–13.6), with approximately 40% of the effect being mediated through differential relapse risk. Upon relapse, patients with a history of early cardiotoxicity were more likely to receive intensive non-ATC reinduction regimens than those without prior cardiotoxicity (56.9% vs. 38.7%). Furthermore, post-relapse survival was significantly lower for patients treated with intensive non-ATC regimens versus ATC-based regimens (HR = 1.66, 95% CI: 1.11-2.49). The effect of regimen intensity on post-relapse survival was modified by history of early cardiotoxicity, with a greater survival benefit from intensive ATC-based reinduction regimens observed among patients without prior early cardiotoxicity. Stratified analysis by dexrazoxane era revealed similar patterns in the mediating role of relapse and the differential effects of reinduction intensity on post-relapse survival. However, the influence of early cardiotoxicity in the selection of non-ATC salvage regimens was attenuated in the post-2018 era when dexrazoxane cardioprotection was more common. Conclusion Early cardiotoxicity was associated with worse OS in a large real-world pediatric AML cohort. Our results demonstrate that nearly half of the absolute effect is mediated through an increased risk of relapse, likely secondary to ATC modifications. Early cardiotoxicity may further affect OS by limiting the use of intensive ATC-based reinduction regimens at relapse leading to reduced post-relapse survival. The new era of near universal use of dexrazoxane cardioprotection in pediatric AML therapy may present opportunities for intensification of reinduction regimens with ATC at relapse.