Analysis
Clinical Flow
Monocytes and pediatric B-ALL MRD: a strong association, not a proven mechanism
A 97-patient diagnostic immunophenotyping study links non-classical monocytes to later MRD; six post-induction single-cell cases and ten survival events set sharp limits.
At first glance, a higher percentage of non-classical monocytes in children who later remain MRD-positive looks like a mechanistic clue: perhaps these innate cells protect residual B-ALL blasts. A 2026 study offers a more demanding lesson. It measures a real clinical association at diagnosis, but the smaller single-cell component, exploratory statistics and limited survival events mean association cannot be treated as proof that monocytes sustain leukemia.
Linlin Zhang and colleagues published their retrospective study in Frontiers in Immunology on 18 August 2026 (DOI 10.3389/fimmu.2026.1776564). Of 98 initially enrolled children, one died during induction and 97 remained in the analysis. Patients were divided by end-of-induction measurable residual disease (EOI-MRD): 61 were MRD-negative and 36 MRD-positive. The investigators also studied a separate six-patient single-cell subcohort (three MRD-positive, three MRD-negative), with diagnostic and post-induction marrow samples. Findings from those six patients cannot be assigned the precision of the 97-patient clinical comparison.

What was measured, and when?
The clinical immune-phenotyping measurements in the article compare diagnostic marrow between groups defined later by EOI-MRD status. T-cell characterization included CD3, CD4, CD8, CD28, PD-1/CD279 and CTLA-4/CD152. Monocytes were gated from CD45-positive CD33-positive events and differentiated by CD14 and CD16: classical CD14-positive CD16-negative, intermediate CD14-positive CD16-positive, and non-classical CD14-low CD16-positive cells. Acquisition used a Beckman Coulter DxFLEX, analysis Kaluza, debris and aggregate exclusion, and FMO controls for selected T-cell thresholds.
EOI-MRD itself was assessed using two eight-colour B-lineage panels including CD19, CD10, CD34, CD38, CD45 and other aberrancy markers. The authors state that at least 500,000 events were collected per sample. The authors report a minimum detectable MRD sensitivity of 0.01% for their assay. This and the stated acquisition target do not alone establish independently validated limits of detection or quantification across laboratories, and monocyte proportions are not direct MRD measurements.
The clinical association
At diagnosis, median non-classical monocyte proportions were 10.09% in the subsequently MRD-positive group and 5.00% in the MRD-negative group (p=0.018, unadjusted). Several T-cell compartments also differed. The paper reports no multiple-comparison correction for this exploratory immune-subset table; therefore an individual nominal p value must not be represented as a confirmed panel-wide finding.
In a multivariable logistic model adjusting for age, risk stratification and neutrophil count, the proportion of CD14-low CD16-positive monocytes was associated with EOI-MRD positivity (reported odds ratio 1.041, 95% CI 1.008–1.075; p=0.014). Those are model-specific associations in a retrospective cohort; they neither demonstrate causal immune suppression nor establish a prospective clinical prediction rule. Disease risk and age were also independently associated with EOI-MRD, so a useful clinical analysis must examine potential confounding rather than attribute outcomes to a single cell population.
The post-induction result is not the baseline result
The six-patient single-cell component introduces an important counterpoint. At EOI, the study reports mean non-classical monocyte proportions of 6.4% in MRD-positive versus 11.7% in MRD-negative samples, but the difference was not significant (p=0.663). It would be incorrect to merge that post-treatment observation with the 97-patient diagnostic association as if they measured the same time point, denominator or platform. The authors also found an HLA-DRB5 transcriptional difference without broad significant functional reprogramming in the same subset.
Single-cell analyses used BD Rhapsody, a healthy marrow reference atlas, expression-based lineage annotation and inferred copy-number profiles for separating malignant and normal B-lineage populations. These are valuable discovery methods, not functional perturbations demonstrating that monocytes cause therapy resistance. Both cell-type assignment and biological meaning depend on reference definitions and quality filtering.
How strong is the clinical inference?
The cohort recorded only ten event-free-survival events. Kaplan–Meier analysis associated EOI-MRD with shorter event-free survival (log-rank p=0.011), but in the exploratory multivariate Cox model EOI-MRD was not independently significant (hazard ratio 1.483, 95% CI 0.271–8.111, p=0.649). The extremely wide interval is a warning against overfitting and confident prognostic claims, not evidence that standard MRD assessment lacks value.
For clinical flow cytometry, the actionable lesson is to preserve the sample/time-point denominator, validate cell-identity gates and assess model uncertainty. Distinguish diagnostic immune-context profiling from MRD detection itself; avoid calling exploratory subset changes predictive biomarkers before prospective replication, corrected analyses and functional confirmation. In practice, this study can help design a hypothesis-driven investigation with rigorously defined monocyte gates, but cannot justify changing clinical MRD interpretation thresholds.
Daily Learning Pill: A diagnostic immune-subset association and a non-significant post-induction single-cell comparison can both be true; they answer different questions in different denominators.
Related CFCM analysis: Why AI-assisted B-ALL MRD interpretation still requires threshold-aware validation.
Source and evidence tier: Zhang L et al., Frontiers in Immunology, original peer-reviewed retrospective research, 18 August 2026. DOI 10.3389/fimmu.2026.1776564. Original CFCM explanatory figure; no published journal figure reproduced. No prospective biomarker or clinical validation claimed.