Treatment for relapsed pediatric acute myeloid leukemia: variability and comparative effectiveness of current approaches to re-induction.

Post-relapse survival of pediatric acute myeloid leukemia (AML) remains poor and there is a paucity of data to inform relapse regimen selection. We sought to describe the treatment landscape for first AML relapse and compare effectiveness of prevalent salvage regimens. To do this, we conducted a retrospective cohort study using the REAL-AML dataset including pediatric patients with first relapse of AML treated at 13 US pediatric institutions from 2011-2023. Primary analyses compared overall survival (OS) by regimen backbone and by receipt of gemtuzumab ozogamicin (GO), without an a priori hypothesis about a specific regimen efficacy. Secondary outcomes included end-of-re-induction clinical remission (CR) and measurable residual disease-negative CR (MRD-CR). 247 patients met inclusion criteria (median post-relapse follow-up 28 months). Two-year OS was 46% (95% CI 39-52%). Thirty-nine unique re-induction regimens were used; 22% received non-intensive regimens. Among 173 patients eligible for comparative analyses, there were no statistically significant differences in OS or end-re-induction responses by backbone chemotherapy. Point estimates suggested benefit with addition of anthracycline to fludarabine/cytarabine±G-CSF for OS (aHR of death vs FLA(G) = 0.38, 95%CI 0.11-1.38), and for CR and MRD-negative CR (aPR = 1.3, for both). Incorporation of GO increased likelihood of CR (aPR 1.3, 95%CI 1.0-1.5) and MRD-negative CR (aPR 1.5, 95% CI 1.1-2.1) but did not improve OS (aHR 1.1, 95%CI 0.6-2.2). These contemporary, real-world data provide updated benchmarks for pediatric first-relapse AML outcomes. Use of GO in relapse may improve end-induction response but survival benefit remains unproven. Prospective evaluation and further work to personalize therapy are needed.