MRD Negativity Does Not Homogenize Prognosis in Pediatric AML: A TARGET-AML Study

Abstract Background The 2025 ELN-DAVID consensus emphasizes genetic-subgroup-specific MRD interpretation in AML. Whether genetic subtypes retain prognostic information after early MFC-MRD negativity in pediatric AML remains unclear. Methods Among 885 children with AML in the TARGET-AML database, the primary analysis cohort comprised 295 patients achieving MFC-MRD < 0.05% at both end of Course 1 and Course 2 after excluding KMT2A-r (CBF-AML 123, non-CBF/non-KMT2A 172). Multivariable Cox and Fine-Gray models were used for EFS, OS, and CIR; Fine-Gray results are reported as subdistribution hazard ratios (sHR). MRD × genetic subtype interaction was assessed in the extended MRD-evaluable cohort (Course 1: n = 679; Course 2: n = 594), because MRD status has only one level in the double-negative cohort. Results CBF-AML was associated with significantly better OS (HR 0.50, 95% CI 0.28–0.89, P = 0.019) but not EFS (HR 0.88, P = 0.546) or CIR (sHR 0.96, 95% CI 0.63–1.46, P = 0.860). Event composition analysis showed similar proportions of patients with relapse (30.1% vs 30.8%), but among patients who relapsed, the proportion of post-relapse death was lower in CBF-AML (37.8% vs 69.8%; Fisher P = 0.005; crude OR 3.74, 95% CI 1.44–10.15, for non-CBF/non-KMT2A vs CBF-AML), suggesting that the OS association may be primarily related to a lower proportion of post-relapse death rather than reduced relapse risk. The OS association remained essentially unchanged after excluding CR1 HSCT recipients (HR 0.50, P = 0.019) and after additional adjustment for FLT3-ITD and NPM1 (HR 0.48, P = 0.017). MRD status significantly modified the association between genetic subtype and EFS at the end of both Course 1 (LRT P = 0.030) and Course 2 (LRT P = 0.0005); MRD positivity was strongly associated with worse EFS in non-CBF/non-KMT2A (Course 1 HR 2.94; Course 2 HR 4.60) and KMT2A-r (Course 1 HR 2.54; Course 2 HR 2.11), whereas no statistically significant association was observed in CBF-AML (Course 1 HR 1.28; Course 2 HR 0.90). Conclusions Among children with AML achieving early MFC-MRD negativity, CBF-AML was associated with significantly better OS, but this association was not accompanied by improvements in EFS or CIR, possibly reflecting a lower proportion of post-relapse death rather than reduced relapse risk. MRD status modified the prognostic impact of genetic subtype, consistent with the broader principle of the ELN-DAVID 2025 framework; however, this study does not validate a specific MRD threshold. These findings suggest that MRD negativity does not define a uniform risk state in pediatric AML within this historical cohort.