Oncofusion

start / graph / concept-mrd-timepoints-and-decisions-infant-all

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

Which MRD timepoint drives which decision in infant ALL, and how good the evidence for each is

also called end of induction MRD · EOI MRD · end of consolidation MRD · EOC MRD · TP2 · TP4 · TP5 · MRD timepoints infant ALL · MRD-directed therapy

Leftover leukaemia is measured at fixed points in treatment, and each point answers a different question: after the first five weeks it says which kind of follow-on chemotherapy to give, and after the next block it says who is doing so badly that only a transplant or an experimental drug is left. The evidence behind each of those decisions is strong for predicting outcome and much weaker for choosing the treatment.

The same thing, for a scientist

Interfant-06 sampled marrow at diagnosis, end of induction (EOI, TP2), end of consolidation (EOC, TP4) and after MARMA (TP5), with MRD mandatory only for medium-risk patients at TP5 to determine transplant eligibility in first remission [E1]. Prognostic strength rises across the sequence. **EOI:** 42 of 217 (19.4%) MRD-negative; 6-year DFS 60.2% negative, 45.0% intermediate (<5x10^-4), 33.8% high (>=5x10^-4), P = .0039 [E1]. **EOC:** 72 of 178 (40.4%) negative; 68.2%, 40.1% and 11.9% respectively, P < .0001 [E1]. **TP5:** 106 of 168 (63.1%) negative; 62.7%, 44.2% and 11.1%, P < .0001 [E1]. Two structural findings matter more than the numbers. First, **EOI MRD is a treatment-selection variable, not just a prognostic one**: its prognostic effect was confined to the lymphoid-consolidation arm (hazard ratio for high EOI MRD 5.36 there, versus no gradient at all in the myeloid arm), meaning MRD-negative infants did better with lymphoid-style consolidation (78.2% versus 45.0%) and high-MRD infants better with myeloid-style (45.9% versus 23.2%) [E1]. Second, **becoming negative late is almost as good as being negative early**: 40 patients MRD-positive at EOI who cleared by EOC had 6-year DFS 65.7% against 72.0% for those negative at both [E1]. Interfant-99 had already shown three MRD-defined groups with all high-risk patients (MRD >=10^-4 at TP3) relapsing [E2], and COG AALL0631 reproduced the prognostic separation by flow at three timepoints in KMT2A-rearranged but not KMT2A-germline infants [E3].

Go deeper
The decision table, with the strength of evidence for each
TimepointDecision it drivesEvidenceHow strong
End of induction (TP2)which consolidation style: lymphoid IB versus myeloid ADE/MAEinteraction between EOI MRD and randomised arm, P = .01 1Moderate. A subgroup interaction inside a randomised trial whose primary comparison was null, on non-mandatory MRD data (249 of 439 eligible). The authors present it as a hypothesis for the next protocol
End of consolidation (TP4)identifies a group with dismal outcome (6-year DFS 11.9% if high) for experimental therapy or transplantprognostic, P < .0001, in both consolidation arms 1Strong as prognosis. No evidence at all that acting on it improves outcome
After MARMA (TP5)transplant eligibility in first remission for medium-risk patients - the one mandatory measurementprotocol rule 1Protocol convention. The transplant decision rests on the prognostic gradient, not on a randomised comparison of transplant by MRD status
EOI plus EOC combineddistinguishes new negatives (65.7%) from persistently positive (three-fold increased risk of treatment failure)Cox model on both timepoints 1Strong as prognosis

The recurring gap is the same one in every MRD literature: the measurement predicts, and whether intervening on it helps has been tested once, indirectly, as a subgroup interaction.

The EOI interaction, and why it is the most interesting result in this beat

Interfant-06 randomised medium- and high-risk patients between lymphoid and myeloid consolidation and found no difference overall. Split by EOI MRD, the arms behave oppositely 1:

  • MRD-negative at EOI (n=42): 6-year DFS 78.2% with lymphoid IB (n=20) versus 45.0% with myeloid ADE/MAE (n=22).
  • MRD-high at EOI (n=93): 45.9% with myeloid (n=43) versus 23.2% with lymphoid (n=50).

The authors' reading is that ALL-type induction selects patients: those who clear it are lymphoid-responsive and should continue lymphoid; those who do not are behaving myeloid and should switch 1. They support it with immunophenotype - 81.1% of high-MRD patients expressed at least one myeloid marker against 50% of MRD-negative patients, P = .0186, with a logistic-regression odds ratio of 2.54 for high EOI MRD given any positive myeloid marker 1.

This connects MRD directly to this graph's lineage layer. The claim is that response to lymphoid therapy is a functional lineage assay, and that a subset of infant KMT2A-rearranged ALL is myeloid enough that myeloid therapy suits it better - which is the treatment-side counterpart of the lineage-plasticity evidence in mechanism-lineage-switch-kmt2a. Two cautions: the arms are small (n = 20 to 50 per cell), and censoring the 67 patients who received transplant in first remission did not change the result 1, which strengthens it, but the interaction itself sits inside a null trial.

Timepoints are protocol phases and do not transfer

TP2 in Interfant-06 is the end of a five-week induction of dexamethasone, vincristine, cytarabine, daunorubicin and asparaginase following a seven-day prednisone prophase 4. COG AALL0631's TP1-TP3 are different phases of a different regimen 3. CCCG-ALL-2015 measures at day 19 and day 46 5. MLL-Baby samples at its own protocol points 6. An MRD value is a response to a specific quantity of specific drugs, so a threshold validated on one backbone carries no automatic meaning on another.

The cut-points differ too, and they are not interchangeable. Interfant-06 uses negative / intermediate (<5x10^-4) / high (>=5x10^-4) on a molecular assay 1; Interfant-99 used 10^-4 2; COG AALL0631 uses 0.01% (10^-4) by flow 3; CCCG-ALL-2015 used 1% for induction failure 5. A child classified high risk on one protocol can be intermediate on another with the identical marrow. concept-risk-group-cut-points-are-administrative-choices-infant-all covers the general form of this; the specific point here is that the timepoint and the threshold are a pair, and quoting either alone is uninterpretable.

Facts and figures
mandatory measurementmedium-risk patients at TP5 only, for HSCT eligibility in CR1
evidence gapno randomised trial has tested whether intervening on an MRD result improves outcome in infant ALL; the only treatment-selection evidence is a subgroup interaction inside a null randomisation
interfant06 timepoints
after MARMA168.0after MARMA: 168end of consolidation178.0end of consolidation: 178end of induction217.0end of induction: 217
eoi by consolidation arm dfs 6y PCT
interaction_p0.010interaction_p: 0.01hr_high_mrd_lymphoid_arm5.360hr_high_mrd_lymphoid_arm: 5.36high_lymphoid23.2high_lymphoid: 23.2negative_myeloid45.0negative_myeloid: 45high_myeloid45.9high_myeloid: 45.9negative_lymphoid78.2negative_lymphoid: 78.2
new negatives
n40.0n: 40dfs_6y_pct65.7dfs_6y_pct: 65.7comparator_negative_both_…72.0comparator_negative_both_pct: 72
interfant06 timepoints — the numbers behind the chart
[
  {
    "code": "TP2",
    "name": "end of induction",
    "n_tested": 217,
    "negative_n": 42,
    "negative_pct": 19.4,
    "dfs_6y_pct": {
      "negative": 60.2,
      "intermediate": 45,
      "high": 33.8
    },
    "p": 0.0039
  },
  {
    "code": "TP4",
    "name": "end of consolidation",
    "n_tested": 178,
    "negative_n": 72,
    "negative_pct": 40.4,
    "dfs_6y_pct": {
      "negative": 68.2,
      "intermediate": 40.1,
      "high": 11.9
    },
    "p": 0.0001
  },
  {
    "code": "TP5",
    "name": "after MARMA",
    "n_tested": 168,
    "negative_n": 106,
    "negative_pct": 63.1,
    "dfs_6y_pct": {
      "negative": 62.7,
      "intermediate": 44.2,
      "high": 11.1
    },
    "p": 0.0001
  }
]
interfant06 cutpoints
{
  "intermediate": "<5e-4",
  "high": ">=5e-4"
}
eoi by consolidation arm dfs 6y PCT — the numbers behind the chart
{
  "negative_lymphoid": 78.2,
  "negative_myeloid": 45,
  "high_lymphoid": 23.2,
  "high_myeloid": 45.9,
  "interaction_p": 0.01,
  "hr_high_mrd_lymphoid_arm": 5.36
}
new negatives — the numbers behind the chart
{
  "n": 40,
  "dfs_6y_pct": 65.7,
  "comparator_negative_both_pct": 72
}
other protocols
{
  "COG AALL0631": "flow MRD at TP1-TP3, 0.01% threshold",
  "CCCG-ALL-2015": "day 19 and day 46/49; induction failure at >=5% blasts or MRD >=1%",
  "Interfant-99": "TP2 and TP3, 1e-4 threshold",
  "MLL-Baby": "flow plus fusion-transcript PCR at protocol-specific points"
}
How this could bite us back
  • The end-of-induction result is used to intensify - and in the myeloid-arm direction that means anthracycline, etoposide and mitoxantrone added to an infant who has already had daunorubicin in induction. The evidence for that switch is one interaction term in a negative trial.
  • A high MRD at end of consolidation predicts a 6-year disease-free survival of 11.9%, and there is no evidence that anything currently available changes it. Measuring it reliably identifies a group for whom the honest offer is an experimental protocol, which is a very different conversation from a risk-adapted intensification.
  • MRD was not mandatory in the trial that generated this evidence. Building the next protocol's mandatory stratification on non-mandatory data means the patients who defined the thresholds are the ones whose centres chose to measure.
  • Thresholds and timepoints are protocol-specific and are routinely quoted without either. A child transferred between protocols, or treated off-protocol on a published threshold, can be allocated to a risk group that the number was never validated for.
  • If MRD is measured on a fusion-specific DNA target, the assay will remain positive through a lineage switch or under a differentiating agent. Intensifying lymphoid therapy against a persistently positive junction in a clone that has already switched is the specific way this decision rule can cause harm.
Still open
  1. Does the end-of-induction MRD by consolidation-arm interaction replicate? It is the only treatment-selection evidence in infant MRD and it comes from a subgroup of a null trial. Interfant-21 is randomising blinatumomab, not this interaction, so it may never be tested directly.
  2. Is transplant in first remission actually better than continued chemotherapy for infants with high MRD after MARMA? The mandatory TP5 measurement exists to make that decision and the decision itself has never been randomised by MRD status.
  3. New negatives at end of consolidation do almost as well as early negatives. Does that mean the kinetics of clearance carry no information beyond the final state, or that the intervening therapy was doing the work? The two readings imply opposite things for shortening therapy.
  4. Every protocol uses a different timepoint and threshold pair. Would a single pooled reanalysis of infant KMT2A-rearranged cohorts on a common scale of log MRD versus time produce a continuous risk model that outperforms all of the categorical schemes?
Every source behind this entry (6)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In 249 infants with KMT2A-rearranged ALL on Interfant-06, MRD measured by DNA-based PCR at end of induction, end of consolidation and after MARMA predicted disease-free survival (EOI: 60.2% negative, 45.0% intermediate, 33.8% high, P = .0039, with 42 of 217 negative; EOC: 68.2%, 40.1%, 11.9%, P < .0001, with 72 of 178 negative; TP5: 62.7%, 44.2%, 11.1%, P < .0001, with 106 of 168 negative); the prognostic effect of EOI MRD was present only in patients given lymphoid-style consolidation (hazard ratio 5.36 for high MRD) with an interaction P = .01, so MRD-negative patients did better with lymphoid-style (78.2% versus 45.0%) and high-MRD patients better with myeloid-style consolidation (45.9% versus 23.2%); 40 patients who converted from positive at EOI to negative at EOC had 6-year DFS 65.7% against 72.0% for those negative at both; and high EOI MRD was associated with myeloid-marker expression (81.1% versus 50%, P = .0186; odds ratio 2.54 for any positive myeloid marker).

    PMID 33405950DOI
    What this evidence cannot show

    Read in full text. MRD was not mandatory except for medium-risk patients at TP5, so 249 of 439 eligible patients contributed and the sample is a convenience subset (the authors report no significant difference in known prognostic features between included and excluded patients). The consolidation interaction is a subgroup analysis within a trial whose primary randomised comparison was null, with 20 to 50 patients per cell. Censoring the 67 patients transplanted in first remission did not change the result. The myeloid-marker association is an odds ratio with a lower confidence bound of 1.03.

    Evidence that argues against this (1)
    • The trial's primary randomised comparison found no difference between consolidation arms (6-year DFS 39.3% versus 36.8%, P = .47), so the entire treatment-selection claim rests on a post-hoc interaction within a negative trial and could be chance. Pieters R et al. 2019
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In 99 infants treated on Interfant-99 with MRD analysed by real-time quantitative PCR of rearranged immunoglobulin, T-cell receptor and MLL genes, higher MRD at end of induction (TP2) and end of consolidation (TP3) was significantly associated with lower disease-free survival, and combined TP2 and TP3 information defined three groups: MRD high risk (MRD at or above 10^-4 at TP3; 26% of patients) in whom all patients relapsed, MRD low risk (below 10^-4 at both timepoints; 44%) with a 13% relapse rate, and MRD medium risk (30%) with 31%.

    PMID 19212338DOI
    What this evidence cannot show

    99 patients, mixing KMT2A-germline and KMT2A-rearranged cases, which the successor study explicitly identified as a limitation. The all-relapsed finding in the MRD high-risk group is based on roughly 26 patients and is the kind of extreme subgroup result that regresses on replication - and did: the comparable Interfant-06 group had a 6-year DFS of 11.9% rather than 0%.

    Evidence that argues against this (1)
    • The same authors note that their MRD-based stratification and the Interfant-06 diagnostic stratification recognised different subgroups of patients, so an MRD-defined high-risk group is not a refinement of the presenting-feature one but a different group. Van der Velden VH et al. 2009
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    On Children's Oncology Group AALL0631, MRD by multiparameter flow cytometry at three timepoints predicted 3-year event-free survival in 117 infants with KMT2A-rearranged ALL (25% versus 49% at TP1, 21% versus 47% at TP2, 22% versus 51% at TP3 for MRD-positive versus negative at 0.01%) but not in 58 infants with KMT2A-germline ALL at any timepoint.

    PMID 37259259DOI
    What this evidence cannot show

    The KMT2A-germline arm is small with few events, so its null is underpowered. The three timepoints are AALL0631 protocol phases and do not correspond to Interfant TP2/TP4/TP5. The 0.01% flow threshold differs from Interfant's 5x10^-4 molecular threshold, so the risk groups are not the same groups.

    Evidence that argues against this (1)
    • On Interfant-06, MRD was prognostic in KMT2A-germline infants too (6-year DFS 61.4% for high end-of-induction MRD versus 87.9% undetectable, P = 0.02), so the KMT2A-germline null in AALL0631 does not replicate. Stutterheim J et al. 2022
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.trial_registry

    The registered Interfant-06 protocol specifies a 7-day prednisone prophase followed by a remission induction from day 8 to day 33 with dexamethasone, vincristine, cytarabine, daunorubicin and asparaginase, then randomised consolidation (protocol IB or ADE then MAE), MARMA, OCTADA(D) reinduction and maintenance to 104 weeks.

    International Collaborative Treatment Protocol for Infants Under One Year With Acute Lymphoblastic or Biphenotypic Leukemia (Interfant-06) · Dutch Childhood Oncology Group (lead sponsor), with 18 named national and international collaborating groups · ClinicalTrials.gov 2006
    NCT00550992
    What this evidence cannot show

    Fetched from the raw ClinicalTrials.gov API on 2026-09-20. A registry summary of what the protocol prescribes, without doses, and its recorded status is UNKNOWN.

    Evidence that argues against this (1)
    • The registry's estimated enrolment of 445 differs from the 651 actually enrolled and its statement of the high-risk age criterion differs in direction from the published report, so registry text is not a reliable substitute for the protocol document. Pieters R et al. 2019
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    CCCG-ALL-2015 assessed infants for induction failure at the day-46 bone marrow using at least 5% blasts on morphology or MRD at or above 1%, and in multivariable analysis of the whole infant cohort MRD negativity on day 19, intermediate-risk group and KMT2A-germline status remained significantly associated with superior event-free survival, while within KMT2A-rearranged patients the factors associated with inferior event-free survival were age under 3 months and MRD positivity at day 19 and day 49.

    PMID 38634141DOI
    What this evidence cannot show

    Read in full text. 134 infants of whom 71 were KMT2A-rearranged, with the authors stating the cohort under-represents high-risk infants; multivariable analysis within 71 patients is underpowered. Timepoints (day 19, day 46/49) and the 1% induction-failure threshold are specific to this protocol.

    Evidence that argues against this (1)
    • The same cohort found no significant difference in EFS or OS between dexamethasone good and poor responders, contradicting the prednisone-response criterion that Interfant protocols use for stratification - so this cohort disagrees with Interfant on early-response measures generally, not only on WBC. Leung AWK et al. 2024
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In 942 bone marrow samples from 123 infants with KMT2A-rearranged ALL on the MLL-Baby protocol, flow-cytometric MRD identified rapid response at early timepoints while PCR for fusion gene transcripts was a reliable relapse predictor at later treatment stages, and the most precise risk definition was obtained by combining the two methods.

    PMID 34970734DOI
    What this evidence cannot show

    A single group's cohort on a moderate-intensity protocol with its own timepoints. The claim that the combination gives the most precise risk definition is a within-cohort observation and has not been validated prospectively or on another backbone.

    Evidence that argues against this (1)
    • Interfant-06 used a single molecular platform and obtained clear prognostic separation at all three timepoints, so the value of combining platforms may be specific to a protocol whose molecular target is the fusion transcript rather than the genomic breakpoint. Stutterheim J et al. 2021
Connected to