Early flow cytometry-based MRD predicts relapse in intermediate-risk AML from low- and middle-income countries

Abstract

Abstract Measurable Residual Disease (MRD) monitoring by Multiparametric Flow Cytometry (MPFC) is a powerful prognostic tool in Acute Myeloid Leukemia (AML), capable of detecting residual leukemic cells beyond morphological remission. While molecular methods are restricted to cases with trackable mutations, MPFC is broadly applicable but underutilized in low- and middle-income countries due to lack of standardization. This study assessed the feasibility and clinical impact of a standardized MPFC-MRD protocol within a national consortium in Brazil. Bone marrow samples from non-APL AML 18-65 years-old patients (IC-AML2015 trial; n=226) were evaluated post-induction 1 (MRD1, n=185) and post-induction 2 (MRD2, n=160) using an 8-color MPFC panel. Leukemia-Associated Immunophenotypes (LAIPs) and “Fifferent-from-Normal” (DfN) strategies were defined using 46 reference bone marrow samples from healthy and regenerating non-myeloid controls. MRD was categorized as Positive (P, ≥0.1%), Low Positive (LP, ≥0.02 and<0.1%), or Negative (N, <0.02%). Patients were stratified by ELN2017-based risk stratification (without TP53, ASXL1 and RUNX1 assessment) and clinical endpoints included relapse-free survival (RFS) and cumulative incidence of relapse (CIR), analyzed by Kaplan-Meier and Cox regression. Included patients had a median age of 49 years-old [47–51 CI] and were treated uniformly with induction 1: daunorubicin (60 mg/m², days 1–3) and cytarabine (200 mg/m², days 1–7); induction 2: daunorubicin (60 mg/m², days 1–3) and cytarabine (1 g/m², twice daily, days 1–6); consolidation was 1-2 cycles of intermediate-doses of cytarabine (1 g/m², twice daily, days 1–6) and/or autologous BM transplantation, per clinical judgment. All MRD analyses were centralized at Ribeirao Preto Medical School Flow Cytometry Lab using FACS Canto II standard procedures and a FlowJo-based analysis pipeline designed specifically for the proposed protocol. Reference samples established empty gates for LAIP-based-DfN identification, applied longitudinally across diagnostic and follow-up samples. The antibody panel included CD45, CD34, CD117 and CD33 as backbone antigens as well as differentiation and leukemic stem cell markers. A consensus tube containing a combination of markers (HLA-DR/CD13/CD7/CD56) to assess maturation discrepancies and the most frequent aberrant markers were applied to all cases regardless of the diagnostic phenotype. MPFC-MRD analysis was feasible in 93–97% of samples, with low rejection due to hemodilution or poor cellularity (3.1% MRD1; 2.4% MRD2). The consensus tube captured up to 86% of MRD+ samples. Among MRD+ cases, 73% were detected by the diagnostic LAIPs and 27% identified by DfN. MRD positivity was found in 21% of patients at MRD1 and 12.5% at MRD2. Immunophenotypic profiles varied by risk group; intermediate-risk patients often exhibited CD117+/CD34var/CD56+ and CD34+/CD33wk/CD13+ patterns. Using a 0.02% MRD threshold, MRD was significantly associated with poorer RFS (MRD1, p=0.038; MRD2, p=0.0014) and higher CIR (MRD2, p=0.0042). Stratified analysis confirmed prognostic relevance of MRD ≥0.02% (Low Positive) in intermediate-risk patients (MRD1 RFS, p=0.0064; MRD2 RFS, p=0.0034); no significant impact was observed in the favorable-risk group. Sequential monitoring showed that persistent Low Positive MRD after both induction cycles conferred the poorest outcomes, and MRD clearance by MRD2 was linked to improved RFS. Multivariate Cox models confirmed MRD as an independent prognostic factor for RFS in intermediate-risk patients (MRD1 HR=8.64; MRD2 HR=7.63, p<0.05), but not in favorable-risk individuals. Age ≥40 years was also an adverse factor. These findings underscore the clinical relevance of deep early MRD clearance in improving outcomes. MPFC-based MRD assessment in AML patients using a single centralized protocol was feasible, reproducible, and clinically informative in the IC-AML study. Standardized treatment and MRD workflows minimized variability and enabled a robust analysis. A lower MRD threshold (≥0.02%) was able to predict relapse. These results support the integration of harmonized MPFC-MRD protocols into routine AML management, particularly in settings where molecular testing is limited.