Real-world post-relapse survival outcomes in children with Acute Myeloid Leukemia by race and ethnicity

Abstract Introduction: Non-Hispanic Black (NHB) and Hispanic children with acute myeloid leukemia (AML) have historically had worse overall survival (OS) compared to their non-Hispanic White (NHW) peers. While our prior work demonstrated comparable risk of relapse by race/ethnicity, it remains unknown whether there are differential outcomes post-relapse which could drive OS disparities. In a large, multi-institutional real-world cohort of children with AML, we evaluated by race/ethnicity (1) post-relapse OS, and (2) potential mechanistic drivers of post-relapse outcomes including frontline treatment history and clinical features at relapse. Methods: We conducted a retrospective cohort analysis of patients in the REAL-AML cohort, a multi-institutional real-world cohort including pediatric (<19 years old) patients treated for de novo AML at 17 US institutions from 2011-2024. Patients with acute promyelocytic leukemia were excluded. This analysis was restricted to patients who experienced relapse following an initial complete remission (CR1) and who were identified as NHB, Hispanic, or NHW in the medical record. Multi-racial and other racial/ethnic groups were excluded from analyses due to small sample sizes. Kaplan Meier survival curves were generated for post-relapse OS (time elapsed between date of first relapse and death from any cause). Cox models compared the hazard of death by race/ethnicity adjusting for a priori determined factors of age at relapse, sex, and cytomolecular risk profile at relapse. We performed pairwise comparisons between NHB and Hispanic patients against NHW patients (reference group) using chi-square and Fisher's exact tests for frontline therapy characteristics (history of bacteremia/sepsis, CTCAE Grade ≥2 cardiotoxicity, dexrazoxane receipt, hematopoietic stem cell transplant [HSCT] receipt in CR1, completion of frontline therapy) and clinical characteristics at relapse (early vs. late relapse, cytomolecular risk profile, body mass index [BMI], acuity, Grade ≥2 cardiotoxicity measured by echocardiogram within 7 days of relapse). Early relapse was defined as <1 year from diagnosis. Cytomolecular risk profile was operationalized as only favorable markers vs. both favorable and unfavorable vs. neutral vs. only unfavorable. Acuity was defined as requiring ICU-level of care within initial 72 hours of relapse presentation. Results: A total of 268 patients (16.0% NHB, 29.9% Hispanic, median follow up 29.2 months) were included in the analytic cohort. Overall, the 3-year post-relapse OS was 41.0% (NHB 34.9%, Hispanic 40.0%, NHW 43.4%). In adjusted multivariable analyses, NHB patients experienced increased hazard of death (HR 1.64, 95% CI: 1.06-2.56) while Hispanic patients had similar outcomes (HR 1.14, 95% CI: 0.79-1.63) compared to NHW patients. This effect was restricted to outcomes for patients experiencing early relapse (NHB 3-year post-early relapse OS 8.7% vs. Hispanic 25.6% vs. NHW 25.3%). The incidence of early relapse was comparable across race/ethnicity (NHB 53.5%, Hispanic 53.8%, NHW 54.5%). We did not observe any statistically significant racial/ethnic differences in frontline bacteremia/sepsis, cardiotoxicity, dexrazoxane receipt, HSCT in CR1, rates of frontline therapy completion, or acuity at time of relapse. At relapse, NHB patients were more likely to have only favorable cytomolecular markers (23.8% vs. 9.8% in NHW patients, p=0.03). Compared to NHW patients (29.8%), both Hispanic (47.1%, p=0.03) and NHB (56.2%, p=0.01) patients were more likely to be overweight/obese. Seven percent of NHB patients had Grade ≥2 left ventricular systolic dysfunction on their most recent echo at relapse; this dysfunction was not observed in any of the NHW or Hispanic patients.Conclusion: NHB children with AML experience greater than 50% increased hazard of death post-relapse compared to NHW children, driven by poor outcomes after early relapse. This survival disparity occurs despite comparable rates of several major frontline therapy characteristics and overall rates of frontline therapy completion, suggesting that downstream intermediates are driving OS differences. Our data indicate potential differences in cardiovascular risk at the time of relapse, including elevated BMI and cardiac dysfunction. An expanded analysis will examine whether these differences mediate survival through re-induction regimen selection, relapse treatment response, HSCT receipt, and treatment-related mortality.