Analysis
AI & Data
AML MRD by flow cytometry
When an AML MRD cluster contains 20 events, it may be detectable without being robustly quantifiable. A 385-analysis validation study shows why the denominator, the phenotypic fingerprint and the reporting threshold all matter.
Technical Review of a peer-reviewed original AML flow cytometry study (2025). DOI 10.3390/cancers17071155. CFCM editorial synthesis; not independent scientific peer review or clinical validation.
Standfirst: When an AML MRD cluster contains 20 events, it may be detectable without being robustly quantifiable. A 385-analysis validation study shows why the denominator, the phenotypic fingerprint and the reporting threshold all matter.
The result begins with the denominator, not the percent sign
Jum’ah and colleagues examined 385 bone-marrow MRD analyses from 126 people with acute myeloid leukemia using three ten-colour flow cytometry tubes and a shared backbone including CD34, CD117, CD13, CD45 and HLA-DR. Their design combined leukemia-associated immunophenotypes (LAIP) with differences from normal maturation (DfN). A target of approximately 1.2 million acquired events per tube supported evaluation of rare populations, but acquired events, non-aggregate events and the relevant analytical denominator are not synonyms.
Twenty events detect; fifty events support quantification
The investigators used a distinct abnormal cluster of at least 20 events as their limit-of-detection criterion and at least 50 events as the quantification criterion. The conservatively validated LOQ was 0.01%; a 0.005% value at one million events was described as theoretical, not automatically transferrable to another panel or laboratory. Their reporting process also applied phenotype and threshold rules: even a tiny visually abnormal cluster should not be relabelled as quantitative clinical MRD without satisfying the complete assay protocol.
Phenotypic context makes a cluster interpretable
MRD was detected in 32 of 126 patients (25%) and in 77 of 385 analyses (20%). The two percentages answer different questions. The original paper's Figure 3 shows abnormal blasts with reduced CD13, CD38 and CD371, and absent CD117 and HLA-DR, alongside normal myeloid precursors. This visual comparison is more informative than calling a single antigen negative. Immunophenotypic shifts occurred in 16 of 77 MRD-positive analyses (21%), supporting continued use of both diagnostic reference patterns and contemporary DfN assessment.
Why analytical validation is not a universal clinical claim
Twenty-five normal marrow specimens were used to develop the reference phenotype and study specificity. A 29-of-32 split-sample agreement result was reported for negative or relatively high-burden samples; the three discordant samples had independent supporting evidence. These observations are important, but the cohort was single centre, assay thresholds were specific to its acquisition and gating strategy, and different laboratories may encounter hemodilution, regeneration, immunophenotypic drift and interobserver variability. There is no basis here for substituting a new clinical treatment decision rule.

Daily Learning Pill
A 20-event detection rule and a 50-event quantification rule are not interchangeable; both belong to a specifically validated assay and its effective cell denominator.
Read the full original article. Continue with CFCM Resources and Troubleshooting.