NUP98 is an eligibility criterion in 22 of 57 menin-inhibitor trials including a paediatric trial for that genotype alone - and NSD1 and KDM5A appear in none of them, so trials select on the fusion and never on its partner
also called NUP98 menin trial enrolment · falsifiers 3 and 5 cashed · trials select on NUP98 not on the partner · NCT07686107
A hypothesis here says NUP98-rearranged leukaemia should respond to the same drug class as our index disease, and nobody had checked whether such patients are even being treated. They are, in 22 trials including one for children with that genotype alone, and one complete remission has been published.
The same thing, for a scientist
`hyp-nup98-partner-obligate-interface` is graded **strong** at a prior of **0.85**, and cycle 61's falsifier sweep found it is the only entry at that confidence in the top of the unrun-falsifier queue. Two of its seven falsifiers are the same clinical question, and both are answerable from public sources today. Neither had been attempted. **THE PRIOR QUESTION IS ELIGIBILITY, NOT RESPONSE.** A falsifier about a response rate cannot fire if the genotype is not enrolled, and *"no responses reported"* is not evidence of no response when nobody was treated. So enrolment was measured first. **Of 57 distinct menin-inhibitor trials, 23 name NUP98 and 22 carry NUP98 in their eligibility criteria.** Positive controls confirm the record text was loaded: KMT2A 44, NPM1 41, leukemia 50, adverse 47, over 1,446,588 characters. The enrolment is not incidental - it is genotype-directed, in children, and at least one trial exists for this genotype alone: - **NCT07686107** — *Trial of Revumenib Plus FLA Chemotherapy for Children With Relapsed or Refractory NUP98-rearranged AML*, Phase 2, minimum age **1 year**, not yet recruiting. - **NCT05886049** — Phase 1b naming *"NUP98 Alterations"* in its own title. - **NCT06229912** — Phase 2 in *"Leukemia Associated With Upregulation of HOX Genes"*, which is the mechanism rather than the genotype. - Paediatric trials of revumenib, ziftomenib and bleximenib with minimum ages of **30 days, 6 months, 1 year and 2 years**. **AND THE PARTNER GENE APPEARS IN NONE OF THEM. `NSD1` and `KDM5A` each return ZERO across all 57 trial records.** Trials select on *NUP98 rearrangement* as a class and never on which partner it carries. That is independent support for this entry's own corrected position - that the dependency does **not** run through the partner-supplied reader or writer module - arriving from clinical practice rather than from mechanism. **On outcome: one published complete remission, and no trial has posted results.** PMID **39633542**, *Haematologica* 2025 (doi 10.3324/haematol.2024.286537, Carraway HE et al.), is titled *"Complete remission of NUP98 fusion-positive acute myeloid leukemia with the covalent menin inhibitor BMF-219, icovamenib."* PubMed carries **no abstract text** for it, so this is cited at title level only. All 23 NUP98-naming trials return `hasResults: false`. **So both falsifiers are answered in direction and neither fires — but only one is fully cashed.** Falsifier [5], which required a trial enrolling NUP98 patients and showing no responses, is contradicted. Falsifier [3], which required a *response rate indistinguishable from unselected refractory AML*, is **not** settled: a single case report is not a rate, and with zero posted results there is no rate to compare. The prior of 0.85 survives; the magnitude question is genuinely open and now has a named place to look.
| script | analyses/nup98-on-menin-inhibitor-trials/run.py |
|---|---|
| results | analyses/nup98-on-menin-inhibitor-trials/results.json |
| status | run |
| run date | 2026-09-22 |
| why this ran | falsifiers [3] and [5] of hyp-nup98-partner-obligate-interface, an entry graded strong at prior 0.85 whose seven falsifiers were all unrecorded when the cycle-61 sweep ranked it top of the queue at that confidence |
| THE PARTNER IS INVISIBLE TO THE CLINIC | NSD1 and KDM5A return zero across all 57 trial records while NUP98 returns 23. Trials enrol on NUP98 rearrangement as a class and never stratify by partner. This is independent support, from clinical practice rather than from mechanism, for this graph's corrected position that the menin dependency does not run through the partner-supplied NSD1 SET or KDM5A PHD module. |
| absence class | measured-not-analysed |
registry
{
"source": "ClinicalTrials.gov API v2, query.intr over nine menin-inhibitor names",
"n_distinct_trials": 57,
"record_characters": 1446588,
"positive_controls": {
"KMT2A": 44,
"NPM1": 41,
"leukemia": 50,
"adverse": 47
},
"trials_naming_NUP98": 23,
"trials_with_NUP98_in_eligibility_criteria": 22,
"trials_with_posted_results": 0,
"NSD1_mentions_across_all_57": 0,
"KDM5A_mentions_across_all_57": 0
}the genotype selected trials
{
"NCT07686107": "Trial of Revumenib Plus FLA Chemotherapy for Children With Relapsed or Refractory NUP98-rearranged AML - Phase 2, CHILD and ADULT, minimum age 1 year, not yet recruiting. A trial for this genotype alone.",
"NCT05886049": "Phase 1b of revumenib with daunorubicin and cytarabine in newly diagnosed AML with NPM1-mutated/FLT3-wildtype, KMT2A-rearranged OR NUP98 alterations - names the genotype in its own title",
"NCT06229912": "Phase 2 of revumenib in leukaemia associated with upregulation of HOX genes - selects on the mechanism rather than on the lesion",
"paediatric_reach": "minimum ages among NUP98-eligible trials include 30 days, 6 months, 1 year, 2 years and 3 years, across revumenib, ziftomenib and bleximenib"
}the one published outcome
{
"pmid": "39633542",
"title": "Complete remission of NUP98 fusion-positive acute myeloid leukemia with the covalent menin inhibitor BMF-219, icovamenib.",
"venue": "Haematologica, 2025",
"doi": "10.3324/haematol.2024.286537",
"first_author": "Carraway HE",
"CITED_AT_TITLE_LEVEL_ONLY": "PubMed carries no abstract text for this record, so nothing beyond the title was read. Patient number, prior therapy, duration of remission and fusion partner are all unknown from this source."
}falsifier adjudication
{
"[5] a trial enrolling NUP98 patients that shows no responses": "CONTRADICTED. Enrolment is broad and genotype-directed - 22 trials list NUP98 in eligibility, one is NUP98-only - and the single published outcome is a complete remission rather than an absence of response.",
"[3] a response rate indistinguishable from unselected refractory AML": "NOT SETTLED, and this is the honest limit of the analysis. A single case report is not a rate, and all 23 NUP98-naming trials return hasResults false, so there is no rate on either side of the comparison. The falsifier does not fire, but it has not been cashed either - what changed is that it now has a named place to look.",
"effect_on_the_prior": "none. 0.85 survives unchanged. Neither falsifier fired, and the enrolment finding is about opportunity rather than about efficacy."
}- THIS IS AN ENROLMENT AND LITERATURE SCAN, NOT EFFICACY DATA, AND IT MUST NOT BE READ AS SUPPORT FOR TREATING A NUP98-REARRANGED CHILD WITH A MENIN INHIBITOR OUTSIDE A TRIAL. What is established is that 22 trials permit the genotype and that one complete remission has been published at title level. No trial has posted results, no response rate exists, and the single published remission was on a covalent inhibitor that most of these trials do not use. The correct reading is that this is an answerable question with an open route - NCT07686107 is the trial designed to answer it - not that the answer is known.
- What is the NUP98 response RATE? No trial among the 23 has posted results, so falsifier [3] cannot be cashed. AUGMENT-101 and the paediatric ziftomenib and revumenib studies are where it will appear first, and a subgroup breakdown for a rare genotype usually surfaces in a conference abstract before a paper.
- Which NUP98 partner did the Haematologica patient carry? The title does not say and the record has no abstract. If it was NUP98::NSD1 or NUP98::KDM5A it speaks directly to this graph's corrected mechanism; if it was a rarer partner it is a different datum.
- Do any of the 22 protocols stratify by NUP98 partner in their statistical plan without naming the partner gene in the registry record? The registry text is terse and a zero for NSD1 and KDM5A there does not bound the protocols themselves.
- Does eligibility translate into accrual? Several of the 23 are not yet recruiting or were withdrawn, and NUP98 rearrangement is rare. The number of NUP98 patients actually dosed on a menin inhibitor is unknown and is the denominator every response question needs.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
Across 57 distinct trials returned by ClinicalTrials.gov API v2 for nine menin-inhibitor names, 23 name NUP98 and 22 carry NUP98 within their eligibility criteria, while NSD1 and KDM5A appear in zero records. No trial has posted results. NCT07686107 is a Phase 2 trial of revumenib plus FLA chemotherapy for children with relapsed or refractory NUP98-rearranged AML, minimum age 1 year; NCT05886049 names NUP98 alterations in its title; and NUP98-eligible trials of revumenib, ziftomenib and bleximenib carry minimum ages as low as 30 days. Positive controls over the same record text returned KMT2A 44, NPM1 41, leukemia 50 and adverse 47 of 57, across 1,446,588 characters.
What this evidence cannot show
A REGISTRY SCAN IS NOT A PROTOCOL READ AND ELIGIBILITY IS NOT ACCRUAL. A trial naming NUP98 in its criteria may never have enrolled one, and several of the 23 are not yet recruiting or were withdrawn. query.intr is a text search over interventions and returns records that merely mention a drug, so 57 is an upper bound on the trial set. The zero for NSD1 and KDM5A is a zero in registry text, which is terse by design - a protocol could stratify by partner in its statistical plan without naming the partner gene in the record.
Evidence that argues against this (2)
- Breadth of eligibility is weak evidence of a genotype-specific rationale. Most of these trials enrol NUP98 alongside KMT2A and NPM1 under a shared HOX-dysregulation umbrella, which means a NUP98 patient can be treated without anyone believing the dependency is partner-independent - or believing anything specific about NUP98 at all. The one NUP98-only trial is Phase 2 and not yet recruiting, so it demonstrates intent rather than result.
- The absence of NSD1 and KDM5A from trial records is at least as easily explained by clinical practicality as by biology. Fusion-partner identification requires RNA sequencing or a targeted fusion panel that many centres do not run at relapse, and eligibility criteria are written to be operable at the bedside. A criterion nobody can test would not be written regardless of whether the partner matters.
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
At least one complete remission of NUP98 fusion-positive acute myeloid leukaemia on a menin inhibitor has been published: PMID 39633542, Haematologica 2025, doi 10.3324/haematol.2024.286537, first author Carraway HE, titled 'Complete remission of NUP98 fusion-positive acute myeloid leukemia with the covalent menin inhibitor BMF-219, icovamenib.' PubMed queries confirm a small but real clinical literature: NUP98 with menin returns 31 records, with 'menin inhibitor' 7, with revumenib 7, with ziftomenib 3, against control operands of 863 for NUP98 alone and 103 for revumenib alone.
Complete remission of NUP98 fusion-positive acute myeloid leukemia with the covalent menin inhibitor BMF-219, icovamenib. · Carraway HE, Nakitandwe J, Cacovean A, et al. · Haematologica 2025What this evidence cannot show
CITED AT TITLE LEVEL ONLY. PubMed carries no abstract text for this record, so the patient count, prior therapy, depth and duration of remission, and crucially WHICH NUP98 PARTNER the patient carried are all unknown from what was read. The full text was not retrieved. A title asserting complete remission is a real datum and is not a response rate.
Evidence that argues against this (2)
- A single published complete remission is the most publication-biased unit of evidence in oncology. Case reports of non-response are almost never written, so one CR is compatible with any underlying response rate including a very low one, and it cannot distinguish a genotype-specific effect from the response an unselected refractory AML patient might have had on the same drug.
- The remission was on icovamenib/BMF-219, a COVALENT menin inhibitor, while most of the preclinical NUP98 evidence in this graph and most of the enrolling trials use the reversible agents revumenib and ziftomenib. Transferring a covalent-inhibitor outcome to the reversible class assumes an equivalence nobody has demonstrated in this genotype.
cashes falsifiers [3] and [5]: neither fires, [5] is contradicted, [3] unsettled because no trial has posted results
the sweep that selected this entry as the only confidence-strong one in the queue