Hemodilution assessment for high-sensitivity flow cytometric measurable residual disease evaluation in pediatric B-cell acute lymphoblastic leukemia

Abstract Background: Accurate assessment of measurable residual disease (MRD) in B-cell acute lymphoblastic leukemia (B-ALL) is critical for risk stratification and MRD-guided therapeutic decision-making. However, hemodilution due to peripheral blood, particularly in bone marrow (BM) aspirates beyond the first pull, can result in falsely low or negative MRD readings. While validated hemodilution criteria exist for AML and multiple myeloma, such criteria for B-ALL MRD remain unestablished. Aim: To develop the flow cytometry-based criteria for hemodilution assessment in BM samples submitted for MRD evaluation in pediatric B-ALL. Methods: We prospectively analyzed 207 paired BM aspirates (first and second pull) from pediatric B-ALL patients (ages 1–17 years) undergoing post-induction MRD assessment using a high-sensitivity 11-color flow cytometry assay. Seven cellular parameters were evaluated: CD117+ precursors, mast cells, CD36+ nucleated red blood cells (NRBCs), and CD10+ granulocytes (determined as a percentage of total viable cells), and CD34+ myeloblasts, plasma cells, and hematogones (as a percentage of viable mononuclear cells). We excluded 18 outlier samples (with >3 SD) using principal component analysis, resulting in 189 paired samples for downstream analysis. ROC analysis was used to determine threshold values for distinguishing first-pull (non-hemodiluted) from second-pull (hemodiluted) samples. Multivariable logistic regression was performed to identify independent parameters. Additionally, a machine learning–based logistic regression model (trained on a random subset of 114 samples: 55 first-pull, 59 second-pull) was also employed. Findings were confirmed in a validation set of 20 in vitro–hemodiluted BM samples (diluted with 50–60% peripheral blood). Results: Among first-pull samples, 123/189 (65%) were MRD-positive (median 0.03%, range 0.0002–3.7%) compared to 107/189 (57%) MRD-positive second-pull samples (median 0.012%, range 0.0002–1.7%). 16/123 (13%) initially MRD-positive cases became MRD-negative in the second pull, including three (2.4%) with clinically significant MRD levels (>0.01%). Similarly, an additional 16/110 (14.5%) samples (with MRD >0.01%) demonstrated false sub-threshold reductions (<0.01%). ROC analysis identified optimal thresholds for detecting hemodilution: CD117+ precursors ≤5.55% (AUC 0.75, p<0.0001), CD34+ myeloblasts ≤0.98% (AUC 0.69, p<0.0001), and CD10+ granulocytes >28.52% (AUC 0.65, p<0.0001). Of seven parameters studied, multivariate logistic regression showed CD117+ precursors (p<0.001), CD34+ myeloblasts (p=0.0001), and CD10+ granulocytes (p=0.006) as independent parameters. These findings were corroborated by machine learning analysis. Using three parameters, a Hemodilution Index (HI) was derived: HI = 1/(1 + EXP(-(1.1905 − 0.27672×A1 + 0.033208×B1 − 0.31811×C1))), where A1 = %CD34+ myeloblasts, B1 = %CD10+ granulocytes, and C1 = %CD117+ precursors. HI >0.5 strongly indicated hemodilution. A simplified scoring system (1 point per parameter meeting threshold) predicted hemodilution with 74.6% sensitivity, 72.5% specificity, and an AUC of 0.76 (p<0.0001) with a total score of 2 or more. The probability of hemodilution increased proportionally with score: 10% (score=0), 30% (score=1), 60% (score=2), and 90% (score=3). Conclusion: This is the first comprehensive study to establish flow cytometry-based criteria for detecting hemodilution in MRD samples from pediatric B-ALL patients. Our data showed that CD117+ precursors, CD34+ myeloblasts, and CD10+ granulocytes are reliable and independent predictors of hemodilution. The Hemodilution Index (HI) and simplified scoring system derived using these three parameters were found to be robust tools to assess BM sample quality, ensuring the accuracy and clinical reliability of MRD results in both routine practice and MRD-guided trials.