The platform argument's best evidence: one drug, two genotypes, one interface
also called menin inhibitor cross-indication · KMT2A and NPM1 shared dependency
The same menin-blocking pill works in two different leukemias: one caused by a broken KMT2A gene, which is what most babies have, and one caused by a mutated NPM1 gene, which is the commonest genetic change in adult AML. That is the clearest proof so far that targeting a shared machine, rather than a particular mutation, can serve several separate groups of patients at once.
The same thing, for a scientist
Revumenib's first-in-human phase 1 (NCT04065399) enrolled patients with relapsed or refractory acute leukemia carrying either a KMT2A rearrangement or an NPM1 mutation, and reported a 30% CR/CRh rate in the efficacy analysis population with QTc prolongation as the only dose-limiting toxicity. The two genotypes are mechanistically joined rather than coincidentally co-targeted: NPM1-mutant AML depends on the same menin-KMT2A chromatin complex, and menin inhibition eradicates NPM1-mutant preleukemic cells in mouse models. Commercially the arithmetic is decisive: KMT2A rearrangements occur in up to 10% of acute leukemias while NPM1 mutations occur in up to 30% of AML, so the NPM1 population is roughly twice the KMT2A population in the United States. The platform worked. It worked in the adult direction.
The clinical fact3 deeper
The first-in-human trial of revumenib enrolled relapsed or refractory acute leukemia with either a KMT2A rearrangement or an NPM1 mutation, and reported a 30% rate of complete remission or complete remission with partial haematologic recovery in the efficacy analysis population, with a low frequency of grade 3 or higher treatment-related adverse events and asymptomatic QTc prolongation as the only dose-limiting toxicity 1. Remissions occurred in leukemias refractory to multiple previous lines, residual disease was cleared on sensitive assays, and differentiation syndrome appeared as a pharmacodynamic hallmark 1.
That is a single small molecule producing remissions across two distinct driver genotypes in one trial. It is the single strongest piece of evidence in this entire graph for the PI's platform thesis.
Why it is mechanism and not coincidence
The two genotypes converge on one molecular machine. Menin is an obligate adaptor for KMT2A on chromatin; the KMT2A fusion protein retains the menin-binding motif and uses menin-LEDGF to reach its target loci (interface-menin-kmt2a-mbm1, mechanism-menin-ledgf-chromatin-tethering). NPM1-mutant AML has no KMT2A fusion at all, yet depends on the same wild-type menin-KMT2A complex to sustain the HOXA/MEIS1 program, and menin inhibition eradicates NPM1-mutant preleukemic cells in mouse models 2.
This distinction matters for the platform argument. A drug that happened to hit two targets would be a lucky polypharmacology story that does not generalise. A drug that hits one obligate interface on which several genotypes independently depend is a template: find the shared dependency, drug the interface, and every genotype routed through it becomes addressable at once. That is the generalisation the lab is actually proposing, and it is already recorded as a hypothesis in this graph (hyp-obligate-partner-interface-generalises).
The arithmetic of the two populations
| Population | Share | Implied US cases/year |
|---|---|---|
| KMT2A-rearranged acute leukemia, all ages | up to 10% of acute leukemias | ~2,900 1 |
| NPM1-mutant AML | up to 30% of AML | ~6,800 1 |
| Combined menin-addressable population | — | ~9,700 |
| KMT2A-rearranged infants | — | ~120 |
| Infant share of the addressable population | — | ~1.2% |
Case counts use SEER's 2026 projections of 22,720 AML and 6,250 ALL diagnoses (see epi-kmt2a-rearranged-acute-leukemia-all-ages). The 10% and 30% shares are the framing figures from the revumenib paper itself 1.
Read the last row before celebrating the third. The platform did exactly what a platform is supposed to do: it turned a rare genotype into a viable commercial programme by joining it to a larger one. But the joining ran from the small population towards the large one, and the infant fraction of the combined population fell to roughly one in eighty. The continuation of this argument is in platform-counterargument-adult-commercial-gravity.
What would make this generalise, and what would kill it
Generalises if: a second obligate-partner interface is found that serves several genotypes, with a drug active across all of them. The candidates already mapped in this graph are the NUP98 fusion family, which is menin-independent but has its own obligate partners (hyp-nup98-partner-obligate-interface), and the SEC/DOT1L axis.
Killed if: the KMT2A and NPM1 responses turn out to be driven by different pharmacology at different exposures, or if menin inhibition proves to be a one-off because no other obligate interface is both essential to the tumour and dispensable in normal tissue. Menin itself is a tumour suppressor in endocrine tissue, which is precisely the essential-function risk the King's charter warns about; that it is tolerated at all is fortunate rather than predictable.
| combined us cases per year | 9700 |
|---|---|
| infant kmt2a r us cases per year | 120 |
| infant share of addressable PCT | 1.2 |
| phase1 cr crh PCT | 30 |
| shared target | gene-men1 |
addressable populations
[
{
"genotype": "KMT2A-rearranged acute leukemia",
"share_of_acute_leukemia_pct": 10,
"us_cases_per_year": 2900
},
{
"genotype": "NPM1-mutant AML",
"share_of_aml_pct": 30,
"us_cases_per_year": 6800
}
]trials
[ "NCT04065399" ]
- Joining a rare paediatric genotype to a common adult one makes the asset viable and simultaneously makes the paediatric population 1.2% of it. The mechanism that rescues the programme is the same mechanism that deprioritises the children it was meant to serve.
- Menin is a tumour suppressor in endocrine tissue and MEN1 loss causes multiple endocrine neoplasia. Long-term menin inhibition in a one-year-old with seventy years ahead is an essential-function risk with no long-term human data behind it, and the adult trials that generated the safety database cannot answer it.
- If resistance mutations at the menin interface (MEN1 M327I and relatives) spread across both genotypes, one resistance mechanism removes the whole platform rather than one indication, which is the concentration risk that comes with a shared target.
- Are the KMT2A-rearranged and NPM1-mutant response rates, depths and durations actually equivalent at the same exposure, or does one genotype need more drug than the other?
- Which second obligate-partner interface could repeat this trick, and does any of them join a small paediatric population to a large adult one the way menin does?
- Does a shared-interface drug ever run the other way, from a large paediatric population subsidising a small adult one, and if not, why not?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The first-in-human phase 1 of revumenib (NCT04065399) in relapsed or refractory acute leukemia with KMT2A rearrangement or NPM1 mutation reported a 30% CR/CRh rate in the efficacy analysis population, a low frequency of grade 3+ treatment-related adverse events, QTc prolongation as the only dose-limiting toxicity, clearance of residual disease and differentiation as a pharmacodynamic hallmark; the paper states KMT2A rearrangements occur in up to 10% of acute leukemias and NPM1 mutations in up to 30% of AML.
The menin inhibitor revumenib in KMT2A-rearranged or NPM1-mutant leukaemia · Issa GC, Aldoss I, DiPersio J, et al. · Nature 2023What this evidence cannot show
A phase 1 dose-finding study in heavily pretreated patients, single-arm, with the efficacy analysis population a subset of those treated; the 30% CR/CRh is pooled across both genotypes rather than reported separately in the abstract. The 10% and 30% prevalence figures are introductory framing, not measurements from this study. Several authors are Syndax employees.
Evidence that argues against this (1)
- The registrational KMT2A-rearranged cohort (AUGMENT-101) reported a CR+CRh of 22.8%, lower than the pooled phase 1 figure, and had a median patient age of 37 with no published paediatric subgroup, so the cross-genotype result should not be read as evidence of efficacy in infants. source 2024
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
NPM1-mutant AML depends on the menin-KMT2A chromatin complex despite carrying no KMT2A rearrangement, and menin inhibition targets and eradicates NPM1-mutant preleukemic cells in a mouse model.
Therapeutic targeting of preleukemia cells in a mouse model of NPM1 mutant acute myeloid leukemia · Uckelmann HJ, Kim SM, Wong EM, et al. · Science 2020What this evidence cannot show
Mouse model of the preleukemic state; eradication of preleukemic cells is not the same as eradication of established human disease, and the translation to the clinical responses in E1 is inferential.
Evidence that argues against this (1)
- The NPM1 dependency was established in a MOUSE model of NPM1-mutant AML and in preleukaemic cells, which is a stronger claim about where the dependency starts than about whether one asset serves two populations clinically. The clinical version arrived separately and is the load-bearing evidence; the mouse work should not be cited as if it established the two-population case. Science 2020
adult KMT2A-r ALL is 82.7% AFF1 versus 48.3% in infants, so an adult trial is nearly a single-fusion trial
one asset active across KMT2A-r and NPM1-mutant leukemia
the shared obligate interface that makes the platform possible
clinical proof of concept for interface-directed generalisation
population arithmetic behind the cross-indication case
- ← analogous toTRK inhibitors: one drug, seventeen cancers, and a paediatric population that actually got dosed
fusion-family platform with infants included from phase 1
large anchor indication cross-subsidising rare ones, with a decade paediatric lag
- ← refutesThe counter-argument: the menin field is an adult field, and the infant studies were withdrawn
the platform succeeded commercially and still did not reach infants
one asset serving both KMT2A-rearranged and NPM1-mutant disease is the platform argument's best evidence
- related to →How many children this affects