SAVE (NCT05360160): the first prospective menin-inhibitor trial to report a NUP98-rearranged subgroup outcome
also called SAVE trial · revumenib + decitabine/cedazuridine + venetoclax · NCT05360160 · SNDX-5613 + DEC/C + VEN
In a 42-patient trial of a menin-blocking drug combined with two other oral drugs, nine patients had leukaemia driven by a NUP98 gene fusion, and they did respond. But their remissions lasted about half as long as those of the other patients.
The same thing, for a scientist
SAVE (NCT05360160) is a phase 1/2 study of an all-oral triplet — revumenib plus decitabine/cedazuridine plus venetoclax — in relapsed/refractory AML aged ≥12 y. It enrolled 42 patients, of whom **21% (n≈9) carried a NUP98 rearrangement**, alongside 40% KMT2A-rearranged and 38% NPM1-mutant [E1]. The composite complete remission (CRc) rate was 71% and the CR/CRh rate 60% overall, with flow MRD negativity in 80% of responders. The genotype-resolved durability is the load-bearing number for this beat: **median duration of CR/CRh was 10.5 months overall, not reached in KMT2Ar, 10.7 months in NPM1mt, and 5.9 months in NUP98r** [E1]. This is, as of 2026-09-19, the **only prospective clinical trial result in which a NUP98-altered subgroup of a menin-inhibitor study has been reported at all** — the dedicated NUP98-admitting cohorts of AUGMENT-101 and cAMeLot-1 have never reported (see `absence-nup98-response-rate-in-menin-inhibitor-trials`). Two cautions govern how far it can be pushed: the NUP98r subgroup **response rate** is not in the abstract and the article body is embargoed, so only durability is quotable; and this is a triplet, so a NUP98r response cannot be attributed to menin inhibition alone — decitabine and venetoclax are both independently active in AML. The paper carries a published erratum [E2] whose content could not be retrieved.
What the number actually says
The verbatim sentence is: "The median duration of CR/CRh for all patients was 10.5 months, not reached in KMT2Ar, 10.7 months in NPM1mt, and 5.9 months in NUP98r." 1
Three things follow and a fourth does not.
- NUP98-rearranged patients respond. A median duration of remission is only definable among responders, so at least some of the ~9 NUP98r patients achieved CR/CRh. That is the first prospective-trial evidence that the preclinical menin dependency of NUP98-fusion AML 34 has a clinical correlate.
- Their remissions are the shortest of the three genotypes, roughly half the NPM1mt figure and far short of the not-yet-reached KMT2Ar figure.
- The ordering is consistent with the preclinical phenotype. Revumenib in NUP98::NSD1 primary AML "elicits differentiation and suppresses proliferation but not apoptotic death" 4 — a cytostatic/differentiating rather than cytotoxic effect, which is exactly the shape of a response that is real but does not last.
- What does not follow: that the NUP98r response rate is lower. The rate is not reported in the abstract, and duration and rate are different quantities. An entry that converted 5.9 months into "NUP98 patients respond less often" would be inventing a number.
Why a triplet result cannot be read as a menin-inhibitor result
SAVE gave revumenib with decitabine/cedazuridine (days 1–5) and venetoclax (days 1–14) 1. Both partners have single-agent and doublet activity in relapsed AML independent of menin biology. Nothing in the reported design separates the contributions, and 52% of patients had prior venetoclax exposure 1, which complicates the comparison in the other direction.
So the honest reading is: NUP98-rearranged patients treated with a revumenib-containing triplet can enter remission, and those remissions are shorter than for the other two menin-susceptible genotypes on the same regimen. The within-trial comparison across genotypes is the part that is protected from the confound, because all three genotypes received the same triplet.
Resistance signal in the same trial
"Emergent mutations in the menin-binding site occurred in 13%" 1. The trial therefore also documents on-target escape at the MEN1 interface in a population that includes NUP98r patients, though the abstract does not break the 13% down by genotype. This is the same escape route already catalogued for KMT2A-rearranged disease in this graph (resistance-men1-interface-mutations), and its appearance here is a direct argument that the NUP98 and KMT2A families share a target rather than merely a drug.
Safety, for the paediatric relevance
Enrolment was from age 12 1. Grade ≥3 events included febrile neutropenia 36%, lung infection 21%, thrombocytopenia 21%; differentiation syndrome occurred in 10% (5% grade 3) and resolved with glucocorticoids 1. NUP98 rearrangements are disproportionately paediatric, so the lower age bound of 12 leaves the youngest NUP98r patients outside this dataset entirely.
| nct | NCT05360160 |
|---|---|
| phase | Phase 1/2 |
| status | reported (primary publication 2026) |
| n | 42 |
| n nup98 rearranged | 9 |
| PCT nup98 rearranged | 21 |
| PCT kmt2a rearranged | 40 |
| PCT npm1 mutant | 38 |
| median age years | 40 |
| age range years | 12-82 |
| crc rate PCT | 71 |
| cr crh rate PCT | 60 |
| mrd negativity PCT of responders | 80 |
| median duration cr crh months all | 10.5 |
| median duration cr crh months npm1mt | 10.7 |
| median duration cr crh months nup98r | 5.9 |
| note on null | median_duration_cr_crh_months_kmt2a_r is null because the published value is 'not reached', which is not a number; it is not a missing value. |
| rp2d | revumenib 160 mg twice daily with a strong CYP3A4 inhibitor |
| menin binding site mutations PCT | 13 |
| differentiation syndrome PCT | 10 |
| readout | CRc rate; CR/CRh rate; duration of CR/CRh by genotype |
| nup98 subgroup response rate | not reported in the abstract; article body not accessible |
arms
[ "revumenib + decitabine/cedazuridine + venetoclax (all-oral triplet)" ]
- A 5.9-month median remission is short enough that menin inhibition in NUP98-rearranged disease is plausibly a bridge to transplant rather than a therapy, which changes the trial design, the endpoint and the age range that matters.
- Emergent menin-binding-site mutations at 13% in this trial mean the resistance route already mapped for KMT2A-rearranged disease is live here too; a NUP98-directed menin programme inherits that liability on day one rather than escaping it.
- Because SAVE is a triplet, a positive NUP98 signal may be carried by venetoclax and decitabine. Designing a NUP98-specific menin monotherapy trial on the strength of this result risks discovering that the active partner was never the menin inhibitor.
- Differentiation syndrome occurred in 10% and NUP98-fusion AML is a differentiation-block disease with high HOXA output; a paediatric NUP98 population may carry a differentiation-syndrome risk at least as high, and the youngest patients are untested.
- What is the CR/CRh RATE in the NUP98-rearranged subgroup of SAVE? Only duration is in the abstract; the rate is the number that would settle whether NUP98 responds as often as KMT2A-r, and it is sitting in an embargoed article body.
- Is the 5.9-month median driven by on-target MEN1-interface mutation, by the cytostatic nature of the response, or by NUP98r co-mutations (FLT3-ITD, WT1)? The trial reports 13% menin-binding-site mutations but not their genotype distribution.
- Does the NUP98r durability disadvantage persist when patients are bridged to transplant? SAVE reports remission duration, not transplant censoring, and for a paediatric NUP98-rearranged population transplant is the usual intent.
- What did the erratum correct?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In the SAVE phase 1/2 trial of revumenib plus decitabine/cedazuridine plus venetoclax, 21% of 42 relapsed/refractory AML patients carried a NUP98 rearrangement, and the median duration of CR/CRh was 5.9 months in NUP98r versus 10.7 months in NPM1mt and not reached in KMT2Ar.
All-Oral Combination of Revumenib, Decitabine, and Venetoclax for Relapsed or Refractory AML (SAVE) · Issa GC, et al. · Journal of Clinical Oncology 44(22):2110-2120 2026What this evidence cannot show
Single-arm, 42 patients total, so the NUP98r subgroup is about 9 patients and the 5.9-month median rests on a handful of events with no confidence interval given in the abstract. The NUP98r-specific CR/CRh RATE is not in the abstract and the article body was not accessible (PMC embargo), so only the durability figure is quotable. Three drugs were given together, so no NUP98r response can be attributed to menin inhibition alone. Median age 40 y with a lower bound of 12 y, which under-represents the paediatric population where NUP98 fusions are commonest. Subgroup durability was almost certainly not a pre-specified powered comparison.
Evidence that argues against this (2)
- The same trial reports emergent menin-binding-site mutations in 13% of patients, so the shorter NUP98r remission duration may reflect on-target resistance shared with other genotypes rather than anything specific to NUP98 biology. The abstract does not give the genotype breakdown of those mutations, so the competing explanations cannot be separated from the published data. Issa GC et al. 2026
- A published erratum exists for this article. Its content could not be retrieved in this session, so it is unknown whether any of the quoted numbers were corrected. Any downstream use of the 5.9-month figure should re-check the erratum first. Issa GC et al. 2026
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
A published erratum to the SAVE report exists and its content was not retrievable in this session.
Erratum: All-Oral Combination of Revumenib, Decitabine, and Venetoclax for Relapsed or Refractory AML (SAVE) · Issa GC, et al. · Journal of Clinical Oncology 44(24):2361 2026What this evidence cannot show
The existence and identifiers of the erratum were verified from PubMed; the text was not. It is recorded here so that a reader does not quote the primary numbers without checking it.
Evidence that argues against this (1)
- An erratum in JCO is frequently an author-list or affiliation correction rather than a data correction, so its existence is not by itself evidence that the efficacy numbers changed. Treating it as impeaching the result would overstate it. Issa GC et al. 2026
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Menin-MLL1 inhibition is effective in NUP98-fusion leukaemia models in vitro and prolongs survival of mice engrafted with NUP98-NSD1 and NUP98-JARID1A patient-derived xenografts.
The menin-MLL1 interaction is a molecular dependency in NUP98-rearranged AML · Heikamp EB, et al. · Blood 139(6):894-906 2022What this evidence cannot show
In vitro work used mouse leukaemia cell lines driven by NUP98-HOXA9 and NUP98-JARID1A; the tool compound was VTP50469, not the clinical agent revumenib. Survival prolongation in PDX is not cure, and the abstract does not report the magnitude.
Evidence that argues against this (1)
- A menin-independent dependency of the same disease has since been reported: KAT6A and KAT7 acetyltransferase complexes are described as molecular dependencies and therapeutic targets in NUP98-rearranged AML, which means menin-MLL1 is one of at least two targetable nodes and not necessarily the dominant one. Michmerhuizen NL et al. 2025
- E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Revumenib specifically blocks leukaemic engraftment in NUP98::NSD1 FLT3-ITD patient-derived AML, but elicits differentiation and suppression of proliferation rather than apoptotic death.
The MLL-Menin Interaction is a Therapeutic Vulnerability in NUP98-rearranged AML · Rasouli M, et al. · HemaSphere 7(8):e935 2023What this evidence cannot show
The cytostatic-not-cytotoxic phrasing is quoted from the abstract; the underlying apoptosis assays were not read in this session. The PDX carried FLT3-ITD, a co-lesion that is itself targetable and that changes the interpretation of any survival benefit.
Evidence that argues against this (1)
- If the effect is differentiating rather than cytotoxic, a shorter remission is the predicted outcome and menin inhibitor monotherapy would be expected to fail in NUP98-rearranged disease. That prediction is consistent with SAVE's 5.9-month figure but has never been tested directly, because no monotherapy NUP98 cohort has reported. Miao H et al. 2024
revumenib arm of the all-oral SAVE triplet
venetoclax days 1-14 of the SAVE triplet
- treated by →NUP98::NSD1, t(5;11)(q35;p15.5)
NUP98r subgroup, 21% of 42, median CR/CRh 5.9 months
- treated by →NUP98::KDM5A (NUP98-JARID1A), t(11;12)(p15;p13)
NUP98r subgroup is genotype-pooled and not resolved by partner
emergent menin-binding-site mutations in 13% of patients
the target whose inhibition the trial evaluates
- ← related toCounted absence: no menin-inhibitor trial has ever reported a response RATE in NUP98-rearranged patients
SAVE is the only exception to the absence: duration reported, rate not
the clinical readout of the borrowed-cofactor prediction
sets the comparator a 5.9-month remission has to beat
nine NUP98-rearranged patients with banked molecular characterisation
Decitabine/cedazuridine (ASTX727) is the hypomethylating component of the SAVE regimen, NCT05360160, minimum eligible age 12 years.