Oncofusion

start / graph / concept-what-mrd-negative-actually-means

Explainerstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q34-mrd-monitoring

What MRD-negative actually means: a quantitative range, a laboratory judgement, and three incompatible definitions

also called MRD negativity · quantitative range · limit of detection MRD · positive below quantitative range · MRD threshold

MRD-negative does not mean no leukaemia cells; it means none were found by a test with a stated floor, and different tests have different floors. Two laboratories can look at the same marrow and one calls it negative while the other calls it positive, because the words describe where the test stops rather than where the disease stops.

The same thing, for a scientist

Three distinct things get reported as *MRD-negative*. **(1) Below the quantitative range** - the target is amplified but cannot be reliably quantified. **(2) Not detected** - no signal at all. **(3) Below a protocol threshold** - a number chosen for allocation, typically 10^-4 or, in Interfant-06, 5x10^-4. Only the third is a decision rule; the first two are analytical states, and the EuroMRD framework exists because 30 laboratories needed to agree which is which so that the same data give the same risk-group assignment everywhere [E1]. The gap between the three is not academic. In Interfant-06 every patient's assay reached a quantitative range of at least 5x10^-4 and that value became the high/intermediate boundary [E2] - the clinical cut-point is the analytical floor, so *intermediate* means *somewhere between the floor and nothing* rather than a measured level. In paediatric B-ALL, moving from flow to sequencing at the same nominal 0.01% threshold reclassified 38.7% more patients as positive, and those patients did worse than the flow-negative ones [E3]; sequencing also identified 19.9% of standard-risk patients with **no detectable MRD at any level**, whose 5-year EFS was 98.1% and OS 100% [E3]. That last group is the clearest demonstration that *not detected* and *below threshold* are different clinical states carrying different prognoses.

Go deeper
The three states, and which one a report is giving you
Reported asWhat it means analyticallyWhat it licenses clinically
Not detectedno amplification or no events above backgroundthe strongest negative available; in paediatric B-ALL the no-detectable-MRD group had 5-year EFS 98.1% 3
Positive, below quantitative rangetarget present, level not reliably quantifiablenot a negative. EuroMRD defines this state precisely so that it is not silently rounded down 1
Below the protocol thresholda level was measured and it is under the allocation cut-pointa treatment allocation, not a statement about disease

The most common error is collapsing rows 2 and 3 into row 1. An entry, a trial report or a clinical note saying MRD-negative without a method, a quantitative range and a threshold is not interpretable, and this graph should treat such a statement as incomplete rather than as a finding.

A worked example of why: AUGMENT-101 reports that 15 of 22 responders (68.2%) had no detectable residual disease 4. The abstract does not state the assay, the sensitivity or the tissue. That number is quoted widely; it is not comparable with an Interfant MRD negativity rate of 19.4% at end of induction 2, which was measured by patient-specific PCR to a stated quantitative range in a frontline setting.

The quantitative range is a property of the patient's assay, not of the method

For a patient-specific assay - ASO-PCR against an Ig/TCR junction or qPCR against a genomic fusion breakpoint - sensitivity is established per patient, by testing that patient's primers against dilutions of their own diagnostic DNA. Interfant-06 states that patient-specific primers were designed from the junctional region and tested for sensitivity and specificity at the national reference laboratories, and that all patients' targets reached a quantitative range of at least 5x10^-4 2. That is a floor across the cohort; individual assays did better and the distribution is not published.

Two consequences. First, a method does not have a sensitivity; an assay does. Quoting PCR MRD is sensitive to 10^-5 describes the best case. Second, because the protocol's clinical threshold (5x10^-4) equals the guaranteed analytical floor, the Interfant-06 intermediate category is defined from below by the assay and from above by a committee, and a child moving between those categories may be moving because of assay performance.

Droplet digital PCR changes the mechanics without changing the logic: it quantifies absolutely and needs no standard curve, but the existing qPCR guidelines had to be adapted to cover droplet generation and intra- and inter-assay variation, and three new analytical quality controls with pass/fail limits were required before it could be used clinically 5. In 82 remission samples from 6 patients, ddPCR and qPCR agreed on 91% of results 5.

Negativity is not the same thing in flow, molecular and sequencing MRD

The three platforms fail to detect for different reasons, so their negatives are not the same negative.

  • Flow. A negative means no population with an abnormal phenotype was resolved above the acquired-event floor. Classical flow is generally held not to reach a solid 10^-4 6. It also cannot see a clone that has changed phenotype, so its negative is conditional on the panel - the point made at length in concept-mrd-flow-cytometry-and-the-lineage-switch-blind-spot.
  • Patient-specific molecular. A negative means the patient's own target was not amplified. It is conditional on that target still existing in the residual clone, which in this disease is exactly what cannot be assumed for Ig/TCR targets 7.
  • Sequencing. A negative means no read matched the index clone above the background of the repertoire. It is the only one of the three that can report no detectable disease at any level as a distinct state, which is where its prognostic advantage came from 3.

In the largest infant comparison, flow and fusion-transcript PCR were concordantly negative in 383 of 942 samples (40.7%) and disagreed in 118 (12.5%), with the molecular assay uniquely positive five times as often as flow 8. Twelve and a half per cent of marrows in a single cohort had a negative by one definition and a positive by another.

Facts and figures
qpcr vs ddpcr agreement PCT91
three states
[
  "not detected",
  "positive below quantitative range",
  "below protocol threshold"
]
interfant06
{
  "quantitative_range_floor": "5e-4 in all patients",
  "thresholds": {
    "intermediate": "<5e-4",
    "high": ">=5e-4"
  },
  "eoi_negative_pct": 19.4,
  "source_pmid": "33405950"
}
platform reclassification
{
  "flow_to_sequencing_extra_positive_pct": 38.7,
  "threshold": "0.01%",
  "n_paired_samples": 619,
  "no_detectable_mrd_subgroup_pct": 19.9,
  "no_detectable_mrd_efs_5y_pct": 98.1,
  "source_pmid": "29284596"
}
platform discordance infants
{
  "samples": 942,
  "concordant_pct": 87.4,
  "discordant_pct": 12.5,
  "source_pmid": "34970734"
}
reporting minimum
[
  "method",
  "analyte",
  "quantitative range or limit of detection",
  "threshold used",
  "tissue",
  "timepoint relative to protocol phase"
]
worked example of incomplete reporting
{
  "trial": "AUGMENT-101 phase II",
  "statement": "15 of 22 (68.2%) had no detectable residual disease",
  "missing": [
    "assay",
    "sensitivity",
    "tissue",
    "denominator definition"
  ],
  "source_pmid": "39121437"
}
How this could bite us back
  • Reporting a negative without its floor makes therapy reduction look safe when it has not been tested. The 98.1% event-free survival of the no-detectable-MRD group in paediatric B-ALL is the kind of number that invites de-escalation, and it was defined retrospectively on a selected cohort by an assay not in routine use.
  • In Interfant-06 the clinical intermediate/high boundary equals the guaranteed analytical floor of the assay. A child can cross the line that decides their consolidation style because their patient-specific primers performed slightly better or worse, and nothing in the report distinguishes that from biology.
  • Twelve and a half per cent of infant marrows in one cohort were negative by one platform and positive by another. Whichever platform a centre owns determines which children are intensified, and there is no reference method to appeal to.
  • A drug's response rate quoted as MRD-negativity without an assay is unfalsifiable. If menin inhibition partly works by differentiation, a phenotype-based negative and a molecular negative are measuring different things, and the trial literature currently does not say which one it used.
Still open
  1. What is the distribution of achieved quantitative range across Interfant-06 patients? Only the floor is published, and whether the intermediate category is mostly assay-limited or mostly genuinely low-level disease depends entirely on that distribution.
  2. How many patients in the infant flow-versus-molecular comparison would have changed risk group, as opposed to how many samples disagreed? Sample-level concordance is the reported statistic and patient-level reclassification is the clinically relevant one.
  3. What assay produced the AUGMENT-101 no-detectable-residual-disease figure, and would the same marrows be negative by a patient-specific genomic breakpoint assay? The number anchors the clinical case for menin inhibition and its method is not in the primary report's abstract.
  4. Does the no-detectable-at-any-level state exist in infant KMT2A-rearranged ALL, and does it carry the same excellent prognosis it does in standard-risk paediatric B-ALL? A patient-specific qPCR cannot report that state; a sequencing assay could.
Every source behind this entry (8)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The European Study Group on MRD detection in ALL, 30 MRD-PCR laboratories worldwide, developed guidelines for interpreting real-time quantitative PCR MRD data whose application ensures identical interpretation of MRD data between different laboratories of the same MRD-based clinical protocol and facilitates comparison of MRD data across treatment protocols.

    PMID 17287850DOI
    What this evidence cannot show

    A consensus interpretive standard, not a measurement. It specifies how to classify a curve, including the positive-below-quantitative-range state; it does not make different platforms comparable, and it predates ddPCR and sequencing-based MRD.

    Evidence that argues against this (1)
    • The framework required substantial adaptation for droplet digital PCR, where there are no standard curves and droplet number, control-gene performance and patient-specific marker performance each need their own quality control with pass/fail limits - so the guideline's interpretive rules are tied to one platform's mechanics. Chen D et al. 2021
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In Interfant-06, patient-specific primers were designed from the junctional region of the identified genomic rearrangement and tested for sensitivity and specificity at the national reference laboratories, real-time quantitative PCR data were analysed according to EuroMRD guidelines, all patients had both MRD-PCR targets with a quantitative range of at least 5x10^-4, MRD results were classified as negative, intermediate (below 5x10^-4) and high (at or above 5x10^-4), and 42 of 217 patients (19.4%) were MRD-negative at end of induction.

    PMID 33405950DOI
    What this evidence cannot show

    Read in full text. The at-least-5x10^-4 figure is a floor across the cohort; the distribution of individual assay sensitivities is not reported, so how many assays were much better than the floor is unknown. Because the clinical threshold equals the guaranteed floor, the intermediate category cannot be separated from assay-limited quantification in this dataset.

    Evidence that argues against this (1)
    • The same study reports high concordance between the two target types (94% of 223 samples) and no significant discrepancy at end of induction, so within one well-run trial the definitional ambiguity did not translate into much disagreement between assays. Stutterheim J et al. 2021
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In 619 paired paediatric B-ALL marrow samples, high-throughput sequencing of IGH and TRG identified 55 (38.7%) more patients as MRD-positive than flow cytometry at a 0.01% threshold, those discrepant patients had worse outcomes than flow-MRD-negative patients, and the increased analytic sensitivity of sequencing identified 19.9% of standard-risk patients with no MRD at any detectable level whose 5-year event-free survival was 98.1% and overall survival 100%.

    PMID 29284596DOI
    What this evidence cannot show

    Paediatric B-ALL across two COG trials, not infants and with no KMT2A-rearranged subgroup; AALL0232 enrolled only patients with day-29 flow MRD below 0.1%, so the cohort is selected on the comparator's result. The no-detectable-MRD group is retrospectively defined and has not been used prospectively to reduce therapy.

    Evidence that argues against this (1)
    • The same study found similar 5-year EFS and OS by both methods at the 0.01% threshold, so on the primary prognostic comparison the two definitions of negative performed alike and the difference is confined to reclassification at the margin. Wood B et al. 2018
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In the phase II portion of AUGMENT-101, among efficacy-evaluable patients with relapsed or refractory KMT2A-rearranged acute leukaemia treated with revumenib, the complete remission plus complete remission with partial haematologic recovery rate was 22.8%, the overall response rate 63.2%, and 15 of 22 patients (68.2%) had no detectable residual disease.

    PMID 39121437DOI
    What this evidence cannot show

    The abstract does not state which MRD assay was used, at what analytical sensitivity, in which tissue, or how the 22-patient denominator was defined relative to the 57 efficacy-evaluable and 94 treated patients. A no-detectable-residual-disease rate reported without a method is not comparable with any of the frontline MRD figures in this beat, and is used here only as an example of the reporting problem.

    Evidence that argues against this (1)
    • The phase I report of the same drug states that clearance of residual disease was demonstrated using sensitive clinical assays, plural and unnamed, alongside hallmarks of differentiation into normal haematopoietic cells - so the founding evidence has the same reporting gap and adds a mechanism (differentiation) that could affect a phenotype-based assay. Issa GC et al. 2023
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Six qPCR MRD assays for Ig/TCR gene rearrangements plus an albumin control assay were migrated to droplet digital PCR and used on 82 remission samples from 6 patients with ALL; three analytical quality controls with pass/fail limits were developed because existing real-time qPCR guidelines needed adaptation to cover droplet generation and intra- and inter-assay variation in the absence of standard curves; ddPCR had comparable sensitivity and quantitative range to the established qPCRs, with 91% overall agreement of MRD results and R^2 = 0.97 in 32 quantifiable samples.

    PMID 39506256DOI
    What this evidence cannot show

    Six patients and six assays; a methodological validation, not a clinical outcome study. The 91% agreement is across 82 samples from those six patients, so samples are not independent. No infant or KMT2A-rearranged patients are identified.

    Evidence that argues against this (1)
    • Nine per cent of results disagreed between two platforms measuring the same targets on the same samples, which is a floor on platform-attributable discordance even when the target, the patient and the laboratory are held constant. Chen D et al. 2021
  6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    MRD techniques need to be sensitive to 10^-4 or better, broadly applicable, accurate, reliable, fast and affordable; classical flow cytometry does not reach a solid 10^-4 and classical allele-specific oligonucleotide PCR is time-consuming and labour intensive, which is why high-throughput sequencing and next-generation multidimensional flow cytometry are being explored.

    PMID 25999452DOI
    What this evidence cannot show

    A review by the developers of both the EuroFlow and EuroMRD standards, describing the criteria they set. The classical-flow statement is a general claim about older panels and is explicitly contrasted with the next-generation flow the same authors developed.

    Evidence that argues against this (1)
    • In the largest infant comparison, flow was uniquely positive in 2.0% of 942 samples and most discordance sat at the sensitivity limits, so the practical sensitivity deficit of flow in this specific disease is narrower than a general 10^-4 statement implies. Popov A et al. 2023
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In Interfant-06 the KMT2A genomic breakpoint fusion sequence was preferred over immunoglobulin and T-cell receptor rearrangements as the MRD target because earlier studies obtained more reliable data with it than with the occasionally absent, oligoclonal or less stable IG/TR rearrangements, and false positivity of IG/TR PCR targets in regenerating bone marrow has been described.

    PMID 33405950DOI
    What this evidence cannot show

    Read in full text. The characterisation of IG/TR targets is the authors' summary of prior work, cited rather than measured here. False positivity in regenerating marrow is asserted with citations, not quantified in this paper.

    Evidence that argues against this (1)
    • IG/TR-based MRD remained prognostic in KMT2A-germline infants on the same protocol, where it is the only molecular option, so target instability does not make an IG/TR negative uninformative as a class. Stutterheim J et al. 2022
  8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Across 942 bone marrow samples from 123 infants with KMT2A-rearranged ALL, 383 samples (40.7%) were concordantly MRD-negative by multicolour flow cytometry and fusion-gene-transcript PCR, 441 (46.8%) were positive by both, 99 (10.5%) were positive only by fusion transcript and 19 (2.0%) only by flow, for an overall concordance of 87.4%, with most discordance occurring where residual leukaemia was present at levels close to the sensitivity limits.

    PMID 34970734DOI
    What this evidence cannot show

    One group's cohort on the MLL-Baby protocol. Sample-level concordance does not translate directly into patient-level risk-group disagreement, and the paper does not report how many patients would have been differently classified.

    Evidence that argues against this (1)
    • The authors' own conclusion is that the two methods are highly comparable and both of high clinical value, and that the discordance is concentrated near the sensitivity limits rather than representing systematic disagreement about disease presence. Popov A et al. 2023
Connected to