No IDH inhibitor has ever been reported in KMT2A-rearranged leukaemia: 4 literature records, all reviews, and 1 registry record, an adult master protocol - and the first version of this query returned a false zero
also called IDH inhibitor absence in KMT2A-r · enasidenib and ivosidenib in KMT2A-rearranged leukaemia
Two IDH-blocking drugs are approved, and the IDH2 mutation they relate to is commoner in KMT2A-rearranged leukaemia than in other kinds. Yet on the day this was checked, nobody had published or registered a study of either drug in that group - and the first search that said 'zero' was broken.
The same thing, for a scientist
**A bounded, dated absence.** On 2026-09-21, crossing a six-name IDH-inhibitor list with four KMT2A-rearrangement terms in PubMed returns **4 records, every one of them a broad review** - *The Care and Cure of the Leukemias in 2026*, a differentiation-syndrome review, an AML management review and a targeted/epigenetic-therapy review - and **none is a trial, cohort or case report of an IDH inhibitor in KMT2A-rearranged disease** [E1]. ClinicalTrials.gov crossed the same way returns **1 record**: the **Beat AML master protocol NCT03013998**, an adult (minimum age 18) biomarker-assignment study of 3,000 estimated patients in which an enasidenib-plus-azacitidine substudy is assigned on **IDH2 mutation** and KMT2A appears as a separate assignment biomarker - so it is not an IDH-inhibitor study in KMT2A-rearranged patients [E2]. **The instrument failure is the more useful half of this entry.** The first form of the query put `t(4;11)` inside the parenthesised OR group and **returned 0** - a clean, well-formed, hypothesis-confirming zero. Removing the translocation parentheses, and nothing else, returned **4**. **This graph already holds the rule** (*PubMed's query parser breaks on translocation parentheses; `der(n)` is safe alone and lethal next to any other parenthesis*) and the beat hit it anyway, because a zero is exactly the answer an absence check hopes for. **Positive controls through the identical shape**, as rule 67 requires: the inhibitor clause crossed with NPM1 returns **12**; the inhibitor clause alone returns **644**; the KMT2A clause alone returns **3,160**. Two large sets, a small intersection, and the intersection is reviews.
The queries, so the absence is bounded and not asserted
All run 2026-09-21 against PubMed esearch with querytranslation inspected, and against the ClinicalTrials.gov API v2.
| query | records | what they are |
|---|---|---|
(enasidenib OR ivosidenib OR olutasidenib OR AG-221 OR AG-120 OR FT-2102) AND (KMT2A-rearranged OR MLL-rearranged OR KMT2Ar OR 11q23) | 4 | 41761287, 41751911, 41749912, 41003912 - all reviews, titles read |
the same with t(4;11) added inside the OR group | 0 | artefact - PubMed parser breaks on translocation parentheses |
(... same inhibitors ...) AND NPM1 | 12 | rule-67 positive control through the identical shape |
| inhibitor clause alone | 644 | operand control |
| KMT2A clause alone | 3,160 | operand control |
(enasidenib OR ivosidenib OR olutasidenib) AND (KMT2A OR MLL OR 11q23) AND (trial OR case OR cohort) | 4 | 41751911, 41749912, 40848224, 36107670 - reviews plus an extramedullary-AML annotation paper |
ClinicalTrials.gov (enasidenib OR ivosidenib OR olutasidenib) AND (KMT2A OR MLL) | 1 | NCT03013998, Beat AML, adult master protocol |
ClinicalTrials.gov (enasidenib OR ivosidenib OR olutasidenib) | 127 | registry operand control |
ClinicalTrials.gov KMT2A | 77 | registry operand control |
Name list (rule 72, and it is the instrument here): enasidenib, AG-221, CC-90007, Idhifa; ivosidenib, AG-120, Tibsovo; olutasidenib, FT-2102; vorasidenib. Singular, plural and nominalisation of IDH inhibitor were all run in an earlier form. Not searched: IDH305, BAY-1436032 and other discontinued development compounds; a reader wanting a tighter bound should add them.
Rule 80 bound. All literature counts are over titles and abstracts. A KMT2A-rearranged patient treated with enasidenib inside a larger cohort paper, or named in a supplementary table, is invisible to every count above. The honest claim is: no paper and no registered protocol is titled or abstracted as an IDH-inhibitor study in KMT2A-rearranged leukaemia, and the supporting tables of the field were not searched.
Why the absence is probably reporting rather than exposure
Both drugs are licensed by IDH genotype, with no KMT2A-related exclusion. IDH2 alteration occurs in about 16% of KMT2A-altered AML, so in the FoundationOne series alone roughly 83 patients carried both lesions. Some of those people have been given enasidenib. What does not exist is anyone reporting how they did.
That makes this a cheap gap to close, and the entry names the route: a retrospective subgroup analysis of existing enasidenib cohorts by KMT2A status. No new patient, no new protocol, no consent beyond what already exists. It is the highest-value, lowest-cost item this beat can hand forward.
And it has a deflating possibility that should be stated in advance, so that a null is not later dressed up: the two lesions may simply not interact, in which case an IDH2-mutant KMT2A-rearranged patient responds to enasidenib like any other IDH2-mutant patient and the subgroup is uninteresting. Ivosidenib's own pivotal study found no preexisting co-occurring single gene mutation predicting response or resistance in 258 patients, which is the prior.
The instrument failure, recorded because it is a live trap
This project's brief carries the rule already: PubMed's query parser breaks on translocation parentheses. The beat read that rule, wrote a query containing t(4;11) inside an OR group, got 0, and was one step from recording a counted absence that was an artefact.
What saved it was rule 67's operand check, not memory of the parentheses rule: the inhibitor clause returns 644 and the KMT2A clause returns 3,160, and two sets that size having an empty intersection is a claim needing its own evidence. Re-running without the translocation term returned 4.
The generalisable form, which is not about PubMed. A query defect and a true absence produce the same output, and in an absence check the defect's output is the one you are hoping for. The operand control is therefore not optional overhead - it is the only thing distinguishing the two, and it costs one extra call.
| script | none - esearch and ClinicalTrials.gov API v2 queries run by hand 2026-09-21, querytranslation inspected for every PubMed query |
|---|---|
| status | run |
| result summary | Literature 4 records, all reviews. Registry 1 record, an adult master protocol assigning on genotype. No trial, cohort or case report of an IDH inhibitor in KMT2A-rearranged leukaemia at any age. |
| date of absence | 2026-09-21 |
| instrument failure recorded | The identical query with t(4;11) inside the parenthesised OR group returned 0. PubMed's parser breaks on translocation parentheses. The zero was an artefact, caught by the rule-67 operand control rather than by remembering the rule. |
| cheapest way to close it | Retrospective subgroup analysis of existing enasidenib cohorts by KMT2A status. No new patient, no new protocol. |
| absence class | measured-not-reported |
| absence class basis | REGISTRY CHECK RUN 2026-09-21 (king), ClinicalTrials.gov API v2. 26 IDH-inhibitor leukaemia trials are completed, terminated or active-not-recruiting, and 6 have POSTED RESULTS (NCT02719574, NCT01915498, NCT03173248, NCT02577406, NCT02677922, NCT05282459). Exactly one registered study returns on an IDH-inhibitor intervention crossed with KMT2A, 'MLL rearranged' or 'MLL-rearranged': NCT03013998, the biomarker-based AML master protocol - which independently reproduces this entry's own count of one registry record, by a different route. Baseline cytogenetics are recorded at enrolment in all of these, so KMT2A status EXISTS in the trial databases; posted results are aggregate and not stratified by cytogenetic subgroup, so it has not been reported. Closing this needs a sponsor to publish a subgroup, not a new trial. |
| absence class assigned | 2026-09-21 (cycle 47, king) |
inputs
[ "PubMed esearch, 6 IDH-inhibitor names x 4 KMT2A terms, plus rule-67 positive control and two operand controls", "ClinicalTrials.gov API v2 free-text query, 3 IDH-inhibitor names x (KMT2A OR MLL), plus two operand controls" ]
name list searched literature
[ "enasidenib", "ivosidenib", "olutasidenib", "vorasidenib", "AG-221", "AG-120", "FT-2102" ]
name list searched registry
[ "enasidenib", "ivosidenib", "olutasidenib" ]
name list NOT searched
[ "IDH305", "BAY-1436032", "CC-90007 (as a code)", "Idhifa", "Tibsovo", "Rezlidhia" ]
controls
{
"rule_67_positive_same_shape": "inhibitor clause AND NPM1 = 12",
"operand_inhibitors_alone": 644,
"operand_kmt2a_alone": 3160,
"registry_operand_inhibitors": 127,
"registry_operand_kmt2a": 77
}archive checked
{
"archive": "ClinicalTrials.gov API v2",
"date": "2026-09-21",
"idh_inhibitor_leukaemia_trials": 26,
"with_posted_results": 6,
"studies_crossing_idh_inhibitor_with_kmt2a_or_mll": 1,
"the_one": "NCT03013998",
"note": "Aggregate posted results cannot be subgrouped by the reader; this is a reporting gap, not a measurement gap."
}- This absence must not be read as a reason to start such a trial. The rationale rests on one retrospective series with no outcome data, an internally inconsistent p-value and a 43%-duplication comparison arm, and the paralogue control is flat.
- Nor should it be read as 'the combination is untried'. Both drugs are licensed on IDH genotype with no KMT2A exclusion, so patients with both lesions have been treated; what is missing is the report. Presenting an absence of reporting as an absence of experience would misrepresent the safety position to anyone acting on it.
- How many patients in the pooled enasidenib development programme were KMT2A-rearranged, and what were their response rates? The data exist and nobody has reported them.
- Does any paediatric IDH-inhibitor study collect KMT2A status? IDH mutations are rare in children, so the answer may be that the question never arises.
- Would adding IDH305, BAY-1436032 and other discontinued compounds to the name list change the count from 4? The bound here is explicitly not complete on that axis.
- How many other counted absences in this graph were written with a translocation term inside a parenthesised OR group? A grep over the recorded query strings would answer it and has not been run.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
Crossing six IDH-inhibitor names with four KMT2A-rearrangement terms in PubMed on 2026-09-21 returns 4 records, all of which are broad reviews and none of which reports an IDH inhibitor given in KMT2A-rearranged leukaemia; positive controls through the identical query shape return 12 (inhibitors x NPM1), 644 (inhibitors alone) and 3,160 (KMT2A terms alone).
PubMed esearch counted-absence query set for IDH inhibitors in KMT2A-rearranged leukaemia, with rule-67 positive and operand controls · Oncofusion knowledge-graph project, Q81-methylome-axis · NCBI E-utilities esearch, run 2026-09-21 with querytranslation inspected for every query 2026What this evidence cannot show
Bounded by titles and abstracts (rule 80): a KMT2A-rearranged patient treated with an IDH inhibitor inside a larger cohort paper or a supplementary table is invisible to this count. Bounded by the name list (rule 72): IDH305, BAY-1436032 and other discontinued compounds were NOT searched. The four returned titles were read; their full texts were not. An earlier form of this query including t(4;11) inside the parenthesised OR group returned 0 and that zero was an artefact of PubMed's parser breaking on translocation parentheses - it is not the figure reported here.
Evidence that argues against this (1)
- Both IDH inhibitors are licensed on IDH genotype with no KMT2A exclusion, and IDH2 alteration occurs in about 16% of KMT2A-altered AML, so patients carrying both lesions have certainly been treated - this is an absence of subgroup reporting, not of exposure, and should never be quoted as 'the combination has never been given'. Batayneh O et al. 2026
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.trial_registry
A ClinicalTrials.gov query for enasidenib, ivosidenib or olutasidenib crossed with KMT2A or MLL returns exactly one study on 2026-09-21 - the Beat AML master protocol NCT03013998, a phase 2/3 biomarker-assignment study with a minimum age of 18 and an estimated enrolment of 3,000, in which an AG-221 (enasidenib) plus azacitidine substudy is assigned on IDH2 mutation while KMT2A is a separate assignment biomarker - against registry operand controls of 127 studies for the inhibitors alone and 77 for KMT2A alone.
A Master Protocol for Biomarker-Based Treatment of AML (The Beat AML Trial) · Beat AML, LLC · ClinicalTrials.gov 2026What this evidence cannot show
Retrieved from the raw ClinicalTrials.gov API v2 on 2026-09-21 with eligibility, sponsor, phase and intervention list; no whyStopped field is set and the record is RECRUITING. The registry free-text query matches whatever text the sponsor supplied, so a protocol that enrols KMT2A-rearranged patients without writing the word KMT2A anywhere would be missed. The substudy-to-biomarker mapping is inferred from the intervention labels (BAML-16-001-S3 = AG-221 + azacitidine) and from the protocol's stated biomarker-assignment design; the individual substudy documents were not retrieved. Only three IDH-inhibitor names were used on the registry side, fewer than the six used on the literature side.
Evidence that argues against this (1)
- A master protocol with 3,000 estimated patients assigning on a biomarker panel that includes both KMT2A and IDH2 will inevitably contain patients with both, so even this single record is closer to unreported exposure than to a genuine gap in clinical activity. Beat AML et al. 2026
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The rationale that would motivate such a study exists but is thin: IDH2 alteration is enriched in KMT2A-altered AML at 16.0% versus 10.4% (recomputed odds ratio 1.630, p = 4.1e-4) while IDH1 in the same table is flat at 6.7% versus 7.6% (p = 0.53), and 43.1% of the KMT2A-altered arm are duplications rather than fusions.
Characterizing KMT2A Rearrangement in Acute Myeloid Leukemia: A Comprehensive Genomic Study. · Batayneh O, Moein M, Naji NS, et al. · Cancers (Basel) 2026What this evidence cannot show
One retrospective commercial-panel series, adult AML, median age 62, no outcome data, no lesion-type stratification of the co-mutation table, and an internal p-value inconsistency between the abstract (<0.0001) and Table 2 (0.0004). Read in full text from PMC12784994 on 2026-09-21.
Evidence that argues against this (1)
- Ivosidenib's pivotal study found that no preexisting co-occurring single gene mutation predicted response or resistance in 258 treated patients, which is a direct prior against expecting a KMT2A-defined subgroup effect for this drug class at all. DiNardo CD et al. 2018
NOT RE-BOUNDED: 489 full-text records and 131 even when tightened, none screened; the burden is stated and no zero is asserted
SURVIVES: twelve IDH-inhibitor names x KMT2A/MLL/cell lines return 8 records, all reviews or general AML
Bounded and dated 2026-09-21: 4 literature records, all reviews; 1 registry record, an adult master protocol assigning on IDH2 genotype.
Same bound, and for ivosidenib the absence is reinforced by a measured null: IDH1 is not enriched in KMT2A-altered AML.
Tests the therapeutic consequence of the enrichment and finds none reported at any age.
The subgroup with no IDH-inhibitor data, despite an approved drug and a matching biomarker.