Oncofusion

start / graph / hyp-menin-inhibitors-belong-frontline-in-infants

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q6b-therapeutics2

Hypothesis: reserving menin inhibitors for relapse in infants is actively harmful, because pretreated disease is already resistant

also called frontline menin hypothesis · pre-existing resistance hypothesis

Menin inhibitors are currently given only after other treatments fail. But there is evidence that leukemia which has already been through heavy chemotherapy stops responding to menin inhibitors even before it has ever seen one. If that is right, saving the drug for later means giving it when it can no longer work.

The same thing, for a scientist

In patient-derived xenografts, highly pretreated KMT2A::AFF1 ALL samples were much less sensitive to menin inhibition than earlier samples from the same patients, despite no prior menin-inhibitor exposure, and with sustained on-target drug activity against canonical targets. Untreated infant ALL samples in the same study were sensitive. We hypothesise that chemotherapy exposure itself selects for a menin-independent transcriptional state, and therefore that the standard development path — prove safety in relapse, then move forward — systematically tests menin inhibitors in the population least able to respond, and that the observed 22.8% CR+CRh in relapsed disease materially understates frontline activity in infants. This is directly testable and, unusually, the clinical experiment is already partly running in TINI 2.

Go deeper
The observation this rests on1 deeper

Serial patient-derived xenografts from two patients with multiply relapsed KMT2A::AFF1 ALL showed that samples from later in the disease course were much less sensitive to menin inhibition than samples from earlier in the same patient, and none of these patients had ever received a menin inhibitor 1.

Two details make this more than an anecdote:

  • The drug was still working. Transcriptomic analysis documented sustained on-target efficacy against the canonical menin-inhibitor targets in the resistant cells. This is not a pharmacokinetic or target-engagement failure; the cells simply stopped needing the thing the drug removes.
  • The untreated comparators were sensitive. In the same study, three AML samples with high-risk fusion partners (MLLT10, AFDN) and two infant ALL samples were sensitive to menin inhibition.

So the pattern is: untreated or lightly treated KMT2A-rearranged disease responds; multiply pretreated disease does not; and the transition happened without menin-inhibitor exposure.

Why infants are the sharpest test case

Three features of infant KMT2A-rearranged ALL point the same way.

  1. Outcomes are worst and unchanged. Interfant-06 6-year EFS was 46.1% overall and 20.9% in the high-risk stratum, and did not improve on Interfant-99. Two randomised intensifications failed. There is no dose-intensity lever left.
  2. The genome is nearly silent. Infant KMT2A-rearranged leukemia has one of the lowest somatic mutation burdens in oncology, so the disease at diagnosis is close to fusion-driven and little else — the most menin-dependent state it will ever be in.
  3. Every relapse is pretreated by definition. Because infant ALL is treated on intensive multi-agent protocols from the start, a relapsed infant is exactly the multiply-pretreated phenotype the xenograft data flag as resistant.

Put together: the population most likely to respond to a menin inhibitor is an infant at diagnosis, and the population the drug is currently licensed for is the same infant eighteen months later, after the state that makes them responsive has been selected away.

What the field is actually doing, and the gap

Over 3,000 adults are randomised across four frontline phase 3 menin trials (NCT07007312 n=1300, NCT07223814 n=875, NCT06652438 n=448, NCT07211958 n=468) 2. Zero infants are randomised to a menin inhibitor anywhere.

The partial exception is TINI 2 (NCT05848687), a recruiting phase 1/2 total-therapy protocol for infants that includes ziftomenib alongside blinatumomab, bortezomib and vorinostat 3. It is n=90, non-randomised and multi-agent, so it cannot isolate the menin contribution — but it will generate the first frontline infant exposure data, including tolerability in babies, which is the rate-limiting unknown.

Meanwhile NCT05761171, the COG study of revumenib plus chemotherapy in relapsed/refractory KMT2A-rearranged infant leukemia (n=78), has closed to accrual. Its readout is the direct test of whether relapsed infants respond, and under this hypothesis it should underperform the adult AUGMENT-101 numbers rather than match them.

Facts and figures
statementChemotherapy exposure selects for a menin-independent transcriptional state in KMT2A-rearranged leukemia. Consequently, menin inhibitors will show substantially higher and more durable activity given frontline to untreated infants than the relapsed/refractory data predict, and the conventional relapse-first development path systematically underestimates the drug class in this population.
origindata-pattern
prior probability0.4
predictions
[
  "COG NCT05761171 (revumenib + chemotherapy in RELAPSED/REFRACTORY KMT2A-rearranged infant leukemia) will report a CR+CRh rate at or below the adult AUGMENT-101 figure of 22.8%, not above it.",
  "Ex vivo menin-inhibitor sensitivity of paired diagnosis and relapse samples from the same infant will be systematically lower at relapse, with the drop uncorrelated with MEN1 mutation status.",
  "Frontline exposure in TINI 2 will show higher MRD clearance attributable to the ziftomenib-containing phase than relapsed-setting response rates would predict.",
  "A menin-dependency transcriptional signature scored on diagnostic versus relapse infant samples will shift away from menin dependency at relapse.",
  "Untreated infant samples will be more menin-sensitive ex vivo than untreated adult KMT2A-rearranged AML samples, because of the lower co-mutation burden."
]
experiments
[
  "exp-orthogonality-time-to-resistance-panel"
]
analyses
[
  "ana-menin-resistance-cell-state-signature"
]
falsifiers
[
  "Paired diagnosis and relapse infant samples show equal ex vivo menin-inhibitor sensitivity.",
  "COG NCT05761171 reports a CR+CRh rate in relapsed infants substantially ABOVE the adult AUGMENT-101 figure, showing that pretreatment does not blunt the drug in this population.",
  "The reduced sensitivity in pretreated xenografts is shown to be driven by an identifiable genetic lesion (for example TP53 or a specific RAS allele) rather than by a transcriptional state, which would make it co-mutation biology rather than a chemotherapy-selected state.",
  "Frontline menin exposure in infants produces response depths no better than the relapsed setting."
]
How this could bite us back
  • If this is wrong, the recommendation it implies — giving a differentiating agent with a 16% grade 3+ differentiation-syndrome rate and QTc prolongation to a newly diagnosed infant who might be cured by chemotherapy plus blinatumomab — exposes children to serious toxicity for nothing.
  • Frontline use also means long exposure, which is exactly the setting in which the unresolved endocrine and carcinogenicity questions about menin inhibition become real, and the paediatric trial that would have characterised long exposure (NCT07498465) was withdrawn.
  • The hypothesis rests on serial samples from two patients. Acting on an n of 2 in a population of infants is precisely the reasoning failure this project is meant to prevent, and the hypothesis should be labelled speculative until paired diagnosis-relapse data exist.
  • Moving a drug frontline removes it as a salvage option; if resistance is selected by the menin inhibitor itself rather than by chemotherapy, frontline use could leave relapsed infants with fewer options rather than more.
Still open
  1. Do paired diagnosis-relapse infant samples exist in sufficient number to test this directly? The Interfant biobank and the COG infant studies should have them, and this is a single ex vivo sensitivity experiment.
  2. Is the selecting pressure chemotherapy specifically, or simply time and subclonal evolution? Distinguishing them changes whether the conclusion is 'give menin inhibitors earlier' or merely 'relapsed disease is worse', which is trivially true and not actionable.
  3. If frontline menin inhibition in infants is right, how would a randomised trial ever be run? Interfant-21 is enrolling now with blinatumomab and will not add an arm; the next protocol is years away, and by then the adult phase 3 results will have made equipoise hard to sustain.
  4. Is the AUGMENT-101 63.2% overall response rate in a heavily pretreated adult population evidence against this hypothesis? It is the strongest counter-argument and it deserves a direct answer rather than being explained away.
Every source behind this entry (3)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In patient-derived xenografts, highly pretreated KMT2A::AFF1 ALL samples were much less sensitive to menin inhibition than samples obtained earlier in the same patients' disease course despite no prior menin-inhibitor exposure, with transcriptomic analysis documenting sustained on-target efficacy against canonical menin-inhibitor targets in the resistant cells; in the same study, three AML samples with high-risk fusion partners (MLLT10, AFDN) and two infant ALL samples were sensitive to menin inhibition.

    Upfront menin-inhibitor resistance in multiply pretreated leukemias · See PubMed record PMID 41027490 for the full author list (not transcribed here) · Experimental Hematology 2025
    PMID 41027490DOI
    What this evidence cannot show

    Serial samples from only TWO patients. This is the entire empirical basis for the central claim and it is thin. Xenograft drug sensitivity is an imperfect proxy for clinical response, and 'multiply pretreated' confounds chemotherapy exposure with disease aggressiveness, time and subclonal evolution — the selecting agent may not be chemotherapy at all. The authors themselves state that the drivers of upfront resistance need broader clarification.

    Evidence that argues against this (1)
    • AUGMENT-101 enrolled a heavily pretreated relapsed/refractory population and still produced a 63.2% overall response rate with 68% of assessed responders MRD-negative, which is hard to reconcile with a strong claim that pretreated disease is broadly menin-resistant. Issa GC et al. 2025
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.trial_registry

    Four recruiting phase 3 menin-inhibitor trials in newly diagnosed adult AML have a combined planned enrolment of 3,091 patients (NCT07007312 n=1300, NCT07223814 n=875, NCT06652438 n=448, NCT07211958 n=468), while no registered trial randomises a menin inhibitor in frontline infant ALL.

    ClinicalTrials.gov API v2 search across menin inhibitors, retrieved 2026-09-18 · US National Library of Medicine, ClinicalTrials.gov · ClinicalTrials.gov 2026
    NCT07007312
    What this evidence cannot show

    A registry snapshot; the absence of an infant randomisation is an absence of evidence in the registry, which may miss non-US/EU academic protocols. Planned enrolment figures are targets.

    Evidence that argues against this (1)
    • Two corrections to this absence claim, both from the raw ClinicalTrials.gov v2 API on 2026-09-19. First, the adult framing is not quite right: one of the four phase 3 studies, NCT07211958 (REVEAL-ND NPM1, Syndax, estimated n=468), has a MINIMUM AGE OF 12 YEARS, so the 3,091 figure is not exclusively adult enrolment. Second, and more important, the absence is of a RANDOMISATION, not of frontline infant exposure: NCT05848687 (TINI 2, maximum age 1 year, recruiting since 2023-11-03, estimated n=90) already administers the menin inhibitor ziftomenib to infants as part of frontline total therapy. Stating that no trial gives a menin inhibitor frontline to infants would be false; stating that none randomises one is true and is the claim that should be carried. Stanford University 2023
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.trial_registry

    TINI 2 (NCT05848687) is a recruiting phase 1/2 total-therapy protocol for infants with acute lymphoblastic leukemia, enrolment 90 across 24 sites, whose interventions include the menin inhibitor ziftomenib alongside blinatumomab, bortezomib, vorinostat, mitoxantrone, dexamethasone, PEG-asparaginase, mercaptopurine and methotrexate.

    TINI 2: Total Therapy for Infants With Acute Lymphoblastic Leukemia II · Gruber TA (sponsor-investigator) · ClinicalTrials.gov 2026
    NCT05848687
    What this evidence cannot show

    Registry record only; no results. Non-randomised and multi-agent, so it cannot attribute effect to ziftomenib. The registry does not make clear which patients receive ziftomenib or at what point in therapy, and that should be read from the protocol.

    Evidence that argues against this (1)
    • TINI 2 cannot test this hypothesis, because its sole primary outcome is a surrogate that is only weakly predictive in this disease and ziftomenib is not separately randomised among nine interventions. On MRD as the endpoint: in 249 infants with KMT2A-rearranged ALL on Interfant-06, DNA-PCR MRD at end of induction predicted outcome, but infants who were MRD-NEGATIVE at end of induction still had a 6-year disease-free survival of only 60.2% (95% CI 43.2-73.6) — roughly two in five MRD-negative infants relapsed. Worse for a single-arm design, the effect of consolidation type INTERACTS with MRD level: among MRD-negative infants, lymphoid-style consolidation gave 6-year DFS 78.2% versus 45.0% for myeloid-style, while among high-MRD infants the ordering reversed (45.9% versus 23.2%). An MRD readout from a non-randomised nine-drug backbone therefore cannot attribute benefit to ziftomenib, and would not establish cure even if it improved. Stutterheim J et al. 2021
Connected to