Oncofusion

start / graph / epi-nup98-nsd1-flt3-itd-confound

Epidemiologystrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Qn-nup98-family

NUP98::NSD1 without FLT3-ITD loses its prognostic significance — the confound that undermines the fusion's own reputation

also called NUP98-NSD1 FLT3-ITD · dual-positive AML · NUP98::NSD1 confound

NUP98-NSD1 leukaemia has a terrible reputation, but three quarters or more of these patients also carry a separate FLT3 mutation. When you analyse the ones without it, the fusion stops predicting a bad outcome — so the reputation may belong to the other mutation.

The same thing, for a scientist

NUP98::NSD1 co-occurs with **FLT3-ITD** at extraordinary rates: **91%** in the founding prognostic report [E1], **82%** across 1,421 patients on five COG/CCG and SWOG trials [E2], **74%** in the largest paediatric landscape cohort with *WT1* also mutated in 42% and "almost half" of FLT3-ITD-positive NSD1 patients also carrying *WT1* [E3], and **76% FLT3-ITD with 50% WT1** in adults [E5]. The dual-positive group has CR 27% versus 69% (p < 0.001) and 3-year OS 31% versus 48% (p = 0.011) [E2]. **And then the confound:** in the same 1,421-patient multivariable analysis, *"NUP98/NSD1 without FLT3/ITD lost its prognostic significance"* [E2], a finding independently reproduced — *"the presence of only NUP98::NSD1 fusion had no significant impact on EFS or OS"* [E4]. This matters for the beat directly. The menin-dependency literature for NUP98 fusions is dominated by NUP98::NSD1 models, most of which carry FLT3-ITD; the one non-responding PDX in the founding menin study carried FLT3-ITD among its co-mutations; the two published PDX series **disagree** about whether FLT3-ITD blocks response; and the only combination-therapy paper in the family pairs a menin inhibitor with a FLT3 inhibitor [E6]. A large fraction of what is attributed to the fusion may belong to the co-mutation.

Go deeper
The co-occurrence numbers, across sources
cohortFLT3-ITD in NUP98::NSD1WT1source
founding prognostic report91%45%1
1,421 patients, 5 COG/CCG + SWOG trials82%—2
paediatric landscape, 108 NSD1 cases74%42%3
adult NUP98-rearranged, n=9576%50%5

For comparison, in the same landscape cohort other NUP98 partners show FLT3-ITD in 18% and WT1 in 16% 5. So the co-occurrence is specific to NSD1, not a property of NUP98 rearrangement generally — which makes it a partner-specific confound, and means analyses that pool NUP98 partners import it unevenly.

The triple-positive configuration — NUP98::NSD1 + FLT3-ITD + WT1 — is common enough to be the modal genotype of this fusion in several cohorts 35.

What the confound does to the prognostic claim

Two independent multivariable analyses reach the same conclusion 24:

  • "NUP98/NSD1 without FLT3/ITD lost its prognostic significance" in a 1,421-patient analysis 2.
  • "The presence of only NUP98::NSD1 fusion had no significant impact on EFS or OS"; the dual-positive group gave EFS HR 3.2 (p = 0.001) but no significant OS effect (HR 1.3, p = 0.618) 4.

This does not mean NUP98::NSD1 is harmless. It means the observed prognostic effect in unstratified cohorts is carried largely or entirely by the co-mutation, and the fusion-only group is small enough that a real but modest fusion effect could be missed. Both readings are live and neither analysis was powered to distinguish them.

The opposite conclusion is also on record and from the largest cohort: NUP98 fusions constitute "a highly refractory class of AML, which justifies reclassification... regardless of fusion partner, as a high-risk subtype" 3. That is a risk-stratification recommendation under uncertainty rather than a causal claim, and this entry does not treat the two as contradictory.

A further complication from a fourth analysis: patients with ITD-positive/NUP98::NSD1 continued to have poor outcomes with intensified therapy including sorafenib, and FLT3-ITD allelic ratio had no independent impact once co-mutations were modelled 7. So the confound is not simply 'it's the FLT3-ITD and FLT3 inhibitors will fix it'.

Why this matters specifically for the menin question

Four connected observations:

  1. The founding menin-dependency study's non-responding PDX carried WT1, ASXL1, IDH1, BCORL1 and FLT3-ITD; a responding PDX was WT1-mutant without that constellation 6.
  2. A second group reported the opposite: menin-inhibitor monotherapy was sufficient in NUP98-rearranged AML with co-occurring FLT3-ITD and WT1 mutations, though adding a FLT3 inhibitor improved it 8.
  3. The only combination paper in the family pairs menin inhibition with kinase inhibition (CDK6 or FLT3), reporting limited single-agent sensitivity 9.
  4. The published single-patient NUP98 complete remission escaped through RAS-pathway co-mutation (pt-story-nup98-nsd1-icovamenib-complete-remission).

Taken together these say that co-mutation, not fusion partner, may be the dominant determinant of menin-inhibitor response in NUP98-rearranged disease — the hypothesis hyp-nup98-menin-response-is-comutation-gated. Nobody has tested it prospectively, and no trial stratifies NUP98 patients by FLT3-ITD or WT1 status.

The practical consequence is uncomfortable: because ~80% of NUP98::NSD1 patients are FLT3-ITD-positive, a NUP98 cohort recruited without stratification will be dominated by a co-mutation whose effect on menin response two laboratories cannot agree on.

Facts and figures
metricco-mutation frequency and its effect on prognostic attribution
populationpaediatric and adult NUP98::NSD1 acute myeloid leukaemia
dual positive cr PCT27
dual positive cr comparator PCT69
dual positive os 3yr PCT31
dual positive os 3yr comparator PCT48
key findingNUP98/NSD1 without FLT3/ITD lost its prognostic significance in a 1,421-patient multivariable analysis, reproduced independently
why it bounds the menin claimThe NUP98 menin-dependency literature is dominated by NSD1 models, most carrying FLT3-ITD; the one non-responding PDX carried it; two laboratories disagree on whether it blocks response; the only combination paper pairs menin with FLT3 inhibition.
sorafenib caveatITD-positive/NUP98::NSD1 patients continued to have poor outcomes with intensified therapy including sorafenib; FLT3-ITD allelic ratio had no independent impact once co-mutations were modelled
flt3 itd cooccurrence PCT by cohort
Bertrums 202374.0Bertrums 2023: 74Ducourneau 2026 (adult)76.0Ducourneau 2026 (adult): 76Ostronoff 201482.0Ostronoff 2014: 82Hollink 201191.0Hollink 2011: 91
wt1 cooccurrence PCT by cohort
Bertrums 202342.0Bertrums 2023: 42Hollink 201145.0Hollink 2011: 45Ducourneau 2026 (adult)50.0Ducourneau 2026 (adult): 50
flt3 itd cooccurrence PCT by cohort — the numbers behind the chart
{
  "Hollink 2011": 91,
  "Ostronoff 2014": 82,
  "Bertrums 2023": 74,
  "Ducourneau 2026 (adult)": 76
}
wt1 cooccurrence PCT by cohort — the numbers behind the chart
{
  "Hollink 2011": 45,
  "Bertrums 2023": 42,
  "Ducourneau 2026 (adult)": 50
}
cooccurrence in non nsd1 nup98 partners
{
  "FLT3-ITD_pct": 18,
  "WT1_pct": 16
}
no trial stratifies by
[
  "FLT3-ITD status within NUP98-rearranged patients",
  "WT1 status",
  "MEIS1 expression"
]
How this could bite us back
  • If the poor prognosis of NUP98::NSD1 belongs largely to FLT3-ITD, then the urgency argument for a NUP98-directed menin inhibitor is partly borrowed from a co-mutation that already has approved targeted drugs.
  • A NUP98 cohort recruited without FLT3-ITD stratification is dominated by a co-mutation whose effect on menin response is disputed, so a negative trial would be uninterpretable and a positive one would not tell you who to treat.
  • Adding a FLT3 inhibitor to a menin inhibitor, as the preclinical combination work suggests, doubles the toxicity and the drug-interaction burden in children - and revumenib already requires CYP3A4 dose adjustment.
Still open
  1. Is there a fusion-only NUP98::NSD1 effect that current cohorts are simply too small to see? Only about one in five NUP98::NSD1 patients lacks FLT3-ITD, so an adequately powered fusion-only analysis would need pooling across every cooperative group.
  2. Does FLT3-ITD block menin-inhibitor response or not? Two laboratories report opposite results in NUP98-NSD1 PDX with the same co-lesion, and no head-to-head comparison exists.
  3. Why do modern FLT3 inhibitors appear to rescue adult NUP98-rearranged outcome while sorafenib did not rescue the paediatric ITD-positive/NSD1 group? Potency, selection, or age?
  4. Should a NUP98 menin-inhibitor trial stratify by FLT3-ITD at randomisation? Given 74-91% co-occurrence in the commonest partner, an unstratified cohort is effectively a study of dual-positive disease.
Every source behind this entry (9)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    NUP98/NSD1 characterises a poor prognostic group with a distinct HOX gene expression pattern, co-occurring with FLT3-ITD in 91% and WT1 mutation in 45%, with median WBC 147 x10^9/L, FAB M4/M5 in 63%, cytogenetically normal in 78%, and 4-year EFS below 10% in both paediatric and adult patients.

    PMID 21813447DOI
    What this evidence cannot show

    The 91% FLT3-ITD co-occurrence is from a small NUP98/NSD1 group within a larger cohort, and the EFS figure is not adjusted for the co-mutation - which is precisely the analysis later papers ran and which changed the conclusion.

    Evidence that argues against this (1)
    • A larger multivariable analysis found that NUP98/NSD1 without FLT3/ITD lost its prognostic significance entirely, so the poor prognosis established here may be attributable to the co-mutation rather than the fusion. Ostronoff F et al. 2014
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Across 1,421 patients on five COG/CCG and SWOG trials, NUP98/NSD1 was present in 7% of cytogenetically normal AML and 15% of FLT3-ITD AML, 82% of NUP98/NSD1 patients were also FLT3-ITD positive, dual-positive patients had CR 27% versus 69% (p<0.001) and 3-year OS 31% versus 48% (p=0.011), and NUP98/NSD1 without FLT3/ITD lost its prognostic significance in multivariable analysis.

    PMID 25145343DOI
    What this evidence cannot show

    Because 82% of NUP98/NSD1 patients carry FLT3-ITD, the fusion-only group is roughly one in five of an already small subgroup, so 'lost its prognostic significance' is substantially a statement about power. A real but modest fusion-only effect would not be detectable in that group.

    Evidence that argues against this (1)
    • The largest paediatric NUP98 cohort nonetheless concludes that NUP98 fusions should be reclassified as high-risk regardless of partner, on the grounds that the class as a whole is refractory. That recommendation does not depend on separating fusion from co-mutation and is arguably the right policy under the uncertainty this entry describes. Bertrums EJM et al. 2023
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In the paediatric NUP98 landscape cohort, NUP98::NSD1 co-occurred with FLT3-ITD in 74% and WT1 mutation in 42%, with almost half of FLT3-ITD-positive NSD1 patients also carrying a WT1 mutation, and trisomy 8 in 18.8% versus 5.3%.

    PMID 36815378DOI
    What this evidence cannot show

    Retrospective cohort assembled over roughly two decades with varying mutation-detection methods, so co-mutation frequencies across eras are not strictly comparable. 74% versus the 91% of the founding report may reflect detection differences rather than population differences.

    Evidence that argues against this (1)
    • The same cohort reports FLT3-ITD in only 18% and WT1 in 16% of non-NSD1 NUP98 partners, which means that pooling NUP98 partners in any analysis imports the confound unevenly and that findings from NSD1-dominated cohorts should not be generalised to the family. Ducourneau B et al. 2026
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    An independent analysis found that the presence of only the NUP98::NSD1 fusion had no significant impact on EFS or OS, while dual NUP98::NSD1/FLT3-ITD positivity gave EFS HR 3.2 (p=0.001) with no significant OS effect (HR 1.3, p=0.618).

    PMID 39182032DOI
    What this evidence cannot show

    This paper reports a NUP98-rearrangement frequency far above every cooperative-group series and its denominator is inconsistently stated in the abstract, so its cohort definition should be treated with caution even though its multivariable conclusion agrees with the larger COG/SWOG analysis. Details were obtained in a delegated research pass and the primary tables were not read.

    Evidence that argues against this (1)
    • That the dual-positive group shows a strong EFS effect but no significant OS effect is internally odd for a group with reported induction failure, and suggests either salvage and transplant rescue these patients or the OS analysis is underpowered. Either way the paper's own numbers do not cleanly support a simple 'it's the FLT3-ITD' story. Ostronoff F et al. 2014
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In an adult cohort of 95 NUP98-rearranged AML patients, NUP98::NSD1 accounted for 54% and carried FLT3-ITD in 76% and WT1 mutation in 50%, against 18% and 16% in other NUP98 partners; median OS was 15.2 months overall, but with FLT3 inhibitors median OS was not reached with 5-year OS 53.3%, and transplant and FLT3 inhibition both carried protective hazard ratios.

    PMID 42120929DOI
    What this evidence cannot show

    Adults, so not directly transferable to the paediatric population this project concerns. The FLT3-inhibitor subgroup is not randomised and patients who receive a FLT3 inhibitor are those who survive to receive one, which is a strong selection effect on a not-reached median. Details from a delegated research pass; primary figures not read.

    Evidence that argues against this (1)
    • A separate paediatric analysis reports that ITD-positive/NUP98::NSD1 patients continued to have poor outcomes with intensified therapy INCLUDING sorafenib, which directly opposes the inference that FLT3 inhibition rescues this genotype. Tarlock K et al. 2024
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In the founding menin-dependency study, a NUP98-NSD1 PDX carrying WT1, ASXL1, IDH1, BCORL1 and FLT3-ITD did not respond to menin inhibition while other NUP98-rearranged PDX models did, giving 2 of 3 responders.

    The menin-MLL1 interaction is a molecular dependency in NUP98-rearranged AML · Heikamp EB, et al. · Blood 139(6):894-906 2022
    PMID 34582559DOI
    What this evidence cannot show

    One non-responding model out of three, carrying five co-mutations at once, so no individual co-mutation can be assigned responsibility. Read via a delegated full-text pass in this session.

    Evidence that argues against this (1)
    • A different group reported that menin-inhibitor monotherapy was sufficient to suppress NUP98-rearranged AML with co-occurring FLT3-ITD and WT1 mutations, which is the opposite conclusion about the same two co-lesions. The two reports have never been reconciled. Rasouli M et al. 2023
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Patients with ITD-positive/NUP98::NSD1 AML continued to have poor outcomes with intensified therapy including sorafenib, with 5-year EFS 22.2% for the poor-risk co-mutation group versus 64% for ITD-positive patients with a favourable co-mutation; FLT3-ITD allelic ratio had no independent impact once co-mutations were modelled.

    PMID 38295280DOI
    What this evidence cannot show

    The poor-risk group pools NUP98::NSD1 with WT1 and UBTF alterations, so the NUP98::NSD1-specific sorafenib result is not isolated. n = 464. Details from a delegated research pass; primary tables not read.

    Evidence that argues against this (1)
    • Sorafenib is a multi-kinase inhibitor with modest FLT3-ITD potency compared with later agents such as gilteritinib and quizartinib, so failure with sorafenib does not establish that FLT3 inhibition in general cannot help this genotype. Ducourneau B et al. 2026
  8. E8moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Menin inhibitor monotherapy was sufficient to suppress NUP98-rearranged AML with co-occurring FLT3-ITD and WT1 mutations, and co-administration with a FLT3 inhibitor enhanced the antileukaemic effect.

    PMID 37520776DOI
    What this evidence cannot show

    Small numbers of patient-derived samples and xenograft models; the same paper characterises the effect as differentiation and suppression of proliferation rather than apoptotic death, so 'sufficient to suppress' is a statement about growth, not eradication.

    Evidence that argues against this (1)
    • The founding study's non-responding PDX carried FLT3-ITD among its co-mutations, giving the opposite result for the same co-lesion in a different laboratory's model. Neither group has attempted a head-to-head comparison and the discrepancy is unresolved. Heikamp EB et al. 2022
  9. E9moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Menin inhibitor MI-3454 gives limited sensitivity as a single agent in NUP98-translocated leukaemia and requires combination with a CDK6 inhibitor (palbociclib) or a FLT3 inhibitor (gilteritinib) for effective treatment.

    PMID 38890447DOI
    What this evidence cannot show

    Preclinical models; 'limited sensitivity to single agents' is a characterisation whose quantitative basis was not read in this session. A combination benefit in models does not establish that monotherapy fails in patients.

    Evidence that argues against this (1)
    • The founding menin study reports single-agent VTP50469 IC50 values of 84 nM for NUP98-KDM5A, which is not 'limited sensitivity'. The disagreement may be partner-specific - the combination work focuses on NSD1-type disease where FLT3-ITD co-mutation is near-universal - which would make this a co-mutation effect rather than a statement about menin inhibition in the family. Heikamp EB et al. 2022
Connected to