Oncofusion

start / graph / drug-ivosidenib

Drugstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q81-methylome-axis

Ivosidenib - the approved IDH1 inhibitor, recorded here so that nobody reaches for it: IDH1 is the paralogue that is NOT enriched in KMT2A-rearranged AML

also called ivosidenib · AG-120 · Tibsovo

Ivosidenib blocks mutant IDH1 and produces durable remissions in adults whose leukaemia carries that mutation. It is in this graph mainly as a warning sign: it is the wrong twin. IDH2 is the one that is more common in KMT2A-rearranged leukaemia, and ivosidenib does not touch IDH2.

The same thing, for a scientist

**Ivosidenib (CHEMBL3989958, AG-120, first approval 2018) is an oral targeted inhibitor of mutant IDH1.** In the pivotal phase 1 dose-escalation and expansion, 258 patients received ivosidenib; in the primary efficacy population of 125 with relapsed or refractory IDH1-mutant AML at 500 mg daily, **CR/CRh was 30.4% (95% CI 22.5-39.3), CR 21.6% (14.7-29.8) and overall response 41.6% (32.9-50.8)**, with median durations of 8.2, 9.3 and 6.5 months; grade >=3 treatment-related events in the 179 relapsed/refractory patients included QT prolongation (7.8%), **IDH differentiation syndrome (3.9%)**, thrombocytopenia (3.4%) and anaemia (2.2%); 7 of 34 patients (21%) in CR/CRh had no residual detectable IDH1 mutation by digital PCR; and **no preexisting co-occurring single gene mutation predicted response or resistance** [E1]. **This entry exists for a negative reason and the negative reason is the point.** In the one series that compares them, **IDH1 alteration is not enriched in KMT2A-altered AML - 6.7% of 521 versus 7.6% of 3,342, recomputed Fisher odds ratio 0.876, p = 0.53** - while IDH2 in the same table is enriched at 16.0% versus 10.4% [E2]. So a reader who compresses the finding to *"IDH is enriched in KMT2A-rearranged AML"* and reaches for an IDH inhibitor has a 50% chance of reaching for the one with no biomarker rationale at all, and ivosidenib does not inhibit mutant IDH2. **No trial, case series or case report of ivosidenib in KMT2A-rearranged leukaemia was found in the literature or the registry on 2026-09-21** [E3]. **Ivosidenib's own trial adds a directly relevant negative:** no preexisting co-occurring mutation predicted response, so nothing in the IDH1 evidence base supports selecting or excluding patients by any co-mutation, KMT2A included [E1].

Go deeper
The paralogue trap, stated as a rule rather than a fact

This graph already carries a family of identifier traps - rule 62 through rule 69, data/symbol-collisions.json, 80 rows. This is one of the same shape and it is not in the registry: IDH is not a gene, it is a two-gene family whose members behave differently in the only comparison that matters here, and the drugs are not interchangeable.

IDH1IDH2
accession, lengthO75874, 414 aaP48735, 452 aa
compartmentcytosol + peroxisome (ECO:0000269)mitochondrion (ECO:0000250 only)
oncometaboliteD-2-HGD-2-HG - the same
KMT2A-altered AML6.7% vs 7.6%, p = 0.5316.0% vs 10.4%, p = 4.1e-4
approved inhibitorivosidenibenasidenib
cross-activitynone against mutant IDH2none against mutant IDH1

And the token itself is contaminated. IDH[tiab] excluding IDH1, IDH2 and isocitrate still returns 3,579 PubMed records on 2026-09-21 - the abbreviation is used for other things entirely in other specialties. Any search, any absence bound and any automated symbol extraction that treats IDH as a gene token will be wrong in both directions. A registry row is proposed in this beat's memo; the file was not edited.

What would have to be true for ivosidenib to belong in this project

Three things, none of which is currently true, and naming them is more useful than the drug's own data:

  1. An IDH1 enrichment in a KMT2A-rearranged cohort, ideally fusion-only. The one measurement available is flat, with an interval (6.7%, Wilson 4.9-9.2) that excludes an IDH2-sized effect.
  2. A mechanism that does not go through 2-HG. Since IDH1 and IDH2 mutants produce the same oncometabolite, a reason to prefer one in KMT2A-rearranged disease would have to be compartmental or clonal, and no such reason is published.
  3. Paediatric data of any kind. The pivotal study is adults; nothing was found in this beat for KMT2A-rearranged disease at any age.

Recorded as a bounded rejection (this project's standing practice of writing down what was read and not used): ivosidenib was enumerated as a candidate for this beat and dropped, on the measured IDH1 null. What would change that: a fusion-only cohort showing IDH1 enrichment, or any case report of response in a KMT2A-rearranged patient.

Facts and figures
chembl IDCHEMBL3989958
targetgene-idh1
modalitysmall molecule (selective inhibitor of MUTANT IDH1; no activity against mutant IDH2)
statusApproved (ChEMBL max_phase 4, first approval 2018) for IDH1-mutant AML in adults. No paediatric indication identified in this beat and no regulatory registry queried.
mechanism summaryInhibits the neomorphic activity of mutant IDH1, lowering D-2-hydroxyglutarate and relieving competitive inhibition of alpha-KG-dependent dioxygenases; clinically produces differentiation-associated remissions with molecular clearance in a minority.
shown for fusionNO - and the reason is recorded rather than merely absent: IDH1 is NOT enriched in KMT2A-altered AML (6.7% vs 7.6%, p = 0.53) in the same assay that finds IDH2 enriched, so there is no biomarker rationale to test it in this subgroup.
why this entry existsTo prevent the compression 'IDH is enriched in KMT2A-rearranged AML, and IDH inhibitors are approved' from producing a recommendation for the wrong paralogue's drug. Ivosidenib has no activity against mutant IDH2.
trials
[
  "NCT02074839",
  "NCT03013998"
]
key ae
[
  "QT prolongation, 7.8% grade >=3",
  "IDH differentiation syndrome, 3.9% grade >=3",
  "thrombocytopenia 3.4%, anaemia 2.2% grade >=3"
]
resistance
[
  "Not characterised in the pivotal abstract; notably no preexisting co-occurring single gene mutation predicted response or resistance."
]
name list for absence bounding
[
  "ivosidenib",
  "AG-120",
  "Tibsovo"
]
token contamination
{
  "IDH_tiab_excluding_IDH1_IDH2_isocitrate": 3579,
  "measured": "2026-09-21, PubMed esearch, querytranslation inspected",
  "status": "counted, not title-checked - an alarm rather than a characterised collision"
}
How this could bite us back
  • The single most likely way this graph could cause harm through this beat is by an agent or reader compressing 'IDH2 is enriched' into 'IDH is enriched' and proposing ivosidenib. Ivosidenib has no activity against mutant IDH2 and IDH1 shows no enrichment; the proposal would be wrong twice over.
  • Ivosidenib carries QT prolongation at 7.8% grade >=3 and differentiation syndrome at 3.9%. Revumenib carries QTc and differentiation syndrome as well. Any speculative combination stacks both in a population with neither a rationale nor a protocol.
  • The token IDH is contaminated at a level of thousands of records in PubMed, so any future automated absence bound in this graph that treats it as a gene symbol will produce a confident wrong number in whichever direction the query happens to fail.
Still open
  1. Does the IDH1 null hold in a fusion-only KMT2A cohort? The measured null comes from an arm that is 43% duplications.
  2. Is there any paediatric IDH1-mutant acute leukaemia cohort at all with KMT2A status reported? None was found here, and IDH mutations are rare in children generally.
  3. Since ivosidenib found no co-mutation predicting response or resistance in 258 patients, is co-mutation-based patient selection for this drug class ever going to be informative, or is the genotype the whole story?
Every source behind this entry (4)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    In the pivotal phase 1 study of ivosidenib (AG-120) in IDH1-mutated AML, 258 patients were treated and in the primary efficacy population of 125 with relapsed or refractory disease at 500 mg daily the CR/CRh rate was 30.4% (95% CI 22.5-39.3), CR 21.6% and overall response 41.6%; grade >=3 treatment-related events in 179 relapsed/refractory patients included QT prolongation 7.8%, IDH differentiation syndrome 3.9%, thrombocytopenia 3.4% and anaemia 2.2%; 7 of 34 patients in CR/CRh had no residual detectable IDH1 mutation; and no preexisting co-occurring single gene mutation predicted clinical response or resistance.

    Durable Remissions with Ivosidenib in IDH1-Mutated Relapsed or Refractory AML. · DiNardo CD, Stein EM, de Botton S, et al. · N Engl J Med 2018
    PMID 29860938DOI
    What this evidence cannot show

    Adults with IDH1-mutant AML; single-arm phase 1 dose-escalation and expansion (NCT02074839), funded by Agios. Efficacy reported for a defined primary efficacy population of 125 with at least 6 months of follow-up, which is a subset of the 258 treated. No KMT2A-rearranged subgroup reported; KMT2A status is not stated as an analysed covariate. Read from abstract only in this beat. Nothing paediatric.

    Evidence that argues against this (1)
    • The finding that no preexisting co-occurring single gene mutation predicted response or resistance is itself an argument against selecting patients for this drug by any co-mutation profile, including KMT2A rearrangement - so the trial that establishes the drug also undermines the only reason this beat would consider it. DiNardo CD et al. 2018
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In 3,863 adult AML cases on one hybrid-capture assay, IDH1 alteration frequency did not differ between KMT2A-altered and KMT2A-wild-type cases - 6.7% of 521 versus 7.6% of 3,342, recomputed Fisher odds ratio 0.876, p = 0.53 - whereas IDH2 in the same table was enriched at 16.0% versus 10.4%, recomputed odds ratio 1.630, p = 4.1e-4.

    Characterizing KMT2A Rearrangement in Acute Myeloid Leukemia: A Comprehensive Genomic Study. · Batayneh O, Moein M, Naji NS, et al. · Cancers (Basel) 2026
    PMID 41514668DOI
    What this evidence cannot show

    Read in full text from PMC12784994 on 2026-09-21 including Table 2; counts back-calculated from published percentages, accurate to about one case. Adult AML, median age 62. 43.1% of the KMT2A-altered arm are duplications rather than fusions and the table is not stratified by lesion type, so this null is a null about a mixed arm. A p of 0.53 excludes an IDH2-sized effect, not a small one.

    Evidence that argues against this (1)
    • A null in a 43%-duplication arm could conceal opposing effects in the fusion and duplication subsets that cancel, and the paper publishes no subset analysis that would expose that - so 'IDH1 is not enriched in KMT2A fusions' is not strictly what was measured. Batayneh O et al. 2026
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.trial_registry

    As of 2026-09-21 no trial, case series or case report of ivosidenib in KMT2A-rearranged leukaemia has been published or registered: the six-name IDH-inhibitor literature query crossed with four KMT2A-rearrangement terms returns 4 records, all broad reviews, and the ClinicalTrials.gov cross returns only the adult Beat AML master protocol NCT03013998, whose IDH arms are assigned on IDH genotype and not on KMT2A status.

    NCT03013998
    What this evidence cannot show

    Bounded by titles and abstracts on the PubMed side (rule 80) and by the name list enasidenib, ivosidenib, olutasidenib, vorasidenib, AG-221, AG-120, FT-2102 (rule 72). Positive control through the identical query shape: the same inhibitor clause crossed with NPM1 returns 12; the inhibitor clause alone returns 644; the KMT2A clause alone returns 3,160. An earlier form of the same query including t(4;11) inside the OR group returned 0 - an artefact of PubMed's parser breaking on translocation parentheses, not a result. Beat AML minimum age 18, estimated enrolment 3,000, no whyStopped set; retrieved individually from the raw API.

    Evidence that argues against this (1)
    • Ivosidenib is licensed on IDH1 genotype and KMT2A status is not an exclusion, so IDH1-mutant KMT2A-rearranged adults will have received it; the absence is of subgroup reporting rather than of exposure, and a retrospective analysis of existing cohorts could close it without a new trial. DiNardo CD et al. 2018
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Ivosidenib is ChEMBL3989958, a small molecule with maximum phase 4 and a first approval recorded in 2018, carrying the synonyms AG-120 and Tibsovo; the token IDH alone is heavily contaminated, with IDH[tiab] excluding IDH1, IDH2 and isocitrate still returning 3,579 PubMed records on 2026-09-21.

    ChEMBL molecule record CHEMBL3989958 (IVOSIDENIB) · ChEMBL, EMBL-EBI · ChEMBL database 2026
    What this evidence cannot show

    ChEMBL fetched 2026-09-21; first_approval is one global year with no indication, jurisdiction or age attached, and no regulatory registry was queried (rule 75). The 3,579 figure is an esearch count with the querytranslation inspected; the off-topic records were NOT title-checked in this beat, so the count bounds the contamination rather than characterising it - the specific alternative meanings of IDH in other specialties are not identified here.

    Evidence that argues against this (1)
    • Rule 70 requires that a token's contamination be characterised by reading titles, not just counted; this row counts 3,579 and does not read them, so it is an alarm rather than a finished collision analysis, and the registry row proposed from it should say so. ChEMBL et al. 2026
Connected to