Oncofusion

start / graph / concept-partner-comutation-confound-in-kmt2a-outcome

Explainermoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q19-partner-outcome

The comutation confound: KRAS, DNMT3A, TP53 and KMT2A-PTD track with partner and carry prognostic weight of their own

also called RAS pathway confound KMT2A partner · KRAS in KMT2A-r AML · KMT2A-PTD modifier

Certain extra mutations turn up more often with certain partners, and those mutations independently predict a worse outcome. So a partner that looks dangerous may just be a partner that tends to arrive with a dangerous mutation.

The same thing, for a scientist

Co-occurring lesions are the third confound, and unlike age and lineage they have been shown to displace the partner effect in adults. In the HARMONY adult AML database (205 KMT2A-r patients, 185 with NGS), overall survival was similar across translocations (t(9;11) P=0.756) while DNMT3A (HR 2.1, P=0.047) and KRAS (HR 2.0, P=0.005) mutations were independent adverse factors, and in de novo patients under 60 a KRAS or TP53 mutation dropped the CR rate from 86% to 50% (P<0.001) and median OS from 30 to 7 months (P<0.001). In a separate 159-patient series, KRAS mutation gave 2-year OS 24.6% versus 50.5% (P=.001) and CIR 56.3% versus 34.6% (P=.018), persisted after transplantation, and was an independent factor on multivariable analysis - while the KMT2A::MLLT4/AFDN subgroup, which showed only a trend to inferior OS, was also the subgroup with the highest KRAS frequency. In paediatric TARGET data, KRAS and SETD2 mutations occurred more often in t(10;11)(p12;q23), the partner independently flagged as adverse. Comutation and partner are therefore not independent variables, and in each of these datasets the comutation carries the significance.

Go deeper
The AFDN example is the cleanest illustration

t(6;11)/KMT2A::AFDN is one of the most reproducible adverse partners: independent HR 2.2 (P<.001) in 756 children 3, and a member of the I-BFM high-risk group. In a 159-patient adult series it showed only a trend to inferior OS - and carried significantly higher KRAS mutation frequency than other rearrangement groups, while KRAS itself was independently adverse 2. Either AFDN is adverse because it is adverse, or it is adverse because it comes with KRAS. No published analysis has adjusted a paediatric partner comparison for RAS-pathway status, so the question is open in the population where the partner effect is largest.

KMT2A-PTD as a modifier, not a confounder

A 125-patient study reports that a concurrent KMT2A partial tandem duplication abolishes the survival advantage of MLLT3/ELL rearrangement 4. That is an interaction rather than a confound: the partner's meaning depends on a second KMT2A lesion. If it replicates, it explains part of why adult partner results disagree, because most series do not test for PTD.

Where this points

Every strong partner effect in this beat comes from paediatric cohorts, which are the cohorts least likely to have comprehensive comutation data; every partner null comes from adult cohorts, which are the ones with NGS panels and which find the signal in the mutations instead. That pattern is consistent with the partner effect being real and consistent with it being comutation in disguise. Distinguishing them requires a paediatric partner analysis adjusted for RAS-pathway status, which is the analysis this beat is recommending.

Facts and figures
confoundco-occurring mutations and KMT2A-PTD
key gapno paediatric partner comparison has been adjusted for RAS-pathway mutation status
sourcePMIDs 38965370, 40530668, 35833755, 41121446
comutations with independent prognostic weight
KRAS0.001KRAS: 0.001KRAS0.005KRAS: 0.005DNMT3A0.047DNMT3A: 0.047KRAS0.050KRAS: 0.05
comutations with independent prognostic weight — the numbers behind the chart
[
  {
    "gene": "KRAS",
    "cohort": "HARMONY adult, n=205",
    "hr_os": 2,
    "p": 0.005
  },
  {
    "gene": "DNMT3A",
    "cohort": "HARMONY adult, n=205",
    "hr_os": 2.1,
    "p": 0.047
  },
  {
    "gene": "KRAS",
    "cohort": "adult de novo, n=159",
    "os_2y_mut_pct": 24.6,
    "os_2y_wt_pct": 50.5,
    "p": 0.001
  },
  {
    "gene": "KRAS",
    "cohort": "paediatric TARGET, n=105",
    "hr_efs": 2.1,
    "p": 0.05
  },
  {
    "gene": "TP53",
    "cohort": "HARMONY adult de novo <60",
    "effect": "CR 50% vs 86%, median OS 7 vs 30 months, both P<0.001"
  }
]
partner comutation associations
[
  {
    "partner": "KMT2A::MLLT4 (AFDN)",
    "enriched_mutation": "KRAS",
    "cohort_pmid": "40530668"
  },
  {
    "partner": "t(10;11)(p12;q23) KMT2A::MLLT10",
    "enriched_mutations": [
      "KRAS",
      "SETD2"
    ],
    "cohort_pmid": "35833755"
  }
]
modifier
{
  "lesion": "KMT2A partial tandem duplication",
  "effect": "abrogates the MLLT3/ELL survival advantage",
  "cohort_pmid": "41121446"
}
How this could bite us back
  • RAS-pathway mutation is both a partner-associated confounder and a documented mechanism of menin-inhibitor resistance, so the partners most enriched for KRAS are simultaneously the ones with the worst prognosis and the ones least likely to respond - a compounding that a class-level trial will read as a single flat response rate.
  • If the partner effect is mostly comutation, then partner-stratified reporting in a menin trial will produce a spurious mechanistic story while the actionable variable goes unmeasured.
  • A menin inhibitor that works well against the partner but not against the co-mutation will show partner-specific response that is really mutation-specific, and the field will draw the wrong conclusion about the interface.
Still open
  1. Does the I-BFM partner risk group remain prognostic after adjustment for RAS-pathway mutation status? This is the single most informative unrun analysis in this beat and the data to run it already exist.
  2. Is the AFDN adverse effect fully mediated by KRAS co-mutation, partially mediated, or independent? A formal mediation analysis has never been reported.
  3. Would stratifying the published adult partner nulls by KMT2A-PTD status recover a partner effect in adults?
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In 205 adults with KMT2A-r AML (185 with NGS), OS was similar across translocations (t(9;11) P=0.756) while DNMT3A (HR 2.1, P=0.047) and KRAS (HR 2.0, P=0.005) mutations were independent adverse factors, and de novo patients under 60 with a KRAS or TP53 mutation had CR 50% versus 86% (P<0.001) and median OS 7 versus 30 months (P<0.001).

    PMID 38965370DOI
    What this evidence cannot show

    Retrospective, assembled across studies so treatment is heterogeneous. Per-translocation n not reported in the abstract, so the partner null has unstated power and could reflect small arms rather than absent effect. Mutation panel content varied across contributing studies.

    Evidence that argues against this (1)
    • A 181-patient adult KMT2A-r AML cohort reported the opposite: concomitant mutations did not significantly affect prognosis, while KMT2A::ELL did (3-year OS 59.8% versus 39.3%, P=0.023) - so which of partner or comutation carries the signal is not settled even within adult disease. Shen H et al. 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In 159 patients with de novo KMT2A-r AML, KRAS mutation gave 2-year OS 24.6% versus 50.5% (P=.001) and CIR 56.3% versus 34.6% (P=.018), persisted in the transplanted subgroup, and was independent on multivariable analysis; the KMT2A::MLLT4/AFDN subgroup had significantly higher KRAS frequency than other rearrangement groups and showed only trends toward inferior OS and higher relapse.

    PMID 40530668DOI
    What this evidence cannot show

    Retrospective clinicogenomic series; the partner-versus-KRAS attribution is observational and the study did not report a formal mediation analysis. The OS hazard ratio for KRAS is reported as 0.51 in the source, in the opposite direction from the CIR hazard ratio of 1.80, which reflects a coding convention rather than a contradiction but means the direction should be read from the survival percentages rather than from the HR sign. NRAS and FLT3 showed no association, so this is a KRAS-specific claim.

    Evidence that argues against this (1)
    • In 756 children, t(6;11)(q27;q23)/KMT2A::AFDN was an independent adverse predictor with HR 2.2 (P<.001) in a model that contained no mutational covariates at all - so the partner effect was established in a setting where the KRAS explanation could not have been detected either way. Balgobind BV et al. 2009
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In paediatric TARGET data on 105 children with 11q23/KMT2A-rearranged AML, 58% carried RAS-pathway mutations and both KRAS and SETD2 mutations occurred more often in t(10;11)(p12;q23); KRAS was associated with worse 5-year EFS (P=0.001) and OS (P=0.009) and was an independent predictor on multivariable analysis (EFS HR 2.10, P=0.05; OS HR 2.39, P=0.054).

    PMID 35833755DOI
    What this evidence cannot show

    105 patients from a public dataset, so the cohort is defined by data availability rather than by protocol; the multivariable P values sit at or just above 0.05. Partner-specific mutation enrichment is reported descriptively without formal adjustment for multiple comparisons across partners and genes.

    Evidence that argues against this (1)
    • In 1,130 children the partner-based risk group remained independently prognostic after adjustment for end-of-induction MRD, which is a downstream integrator of biological aggressiveness including comutation, suggesting the partner effect is not entirely mediated by mutational background. van Weelderen RE et al. 2023
  4. E4weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    A concurrent KMT2A partial tandem duplication abrogates the survival advantage of MLLT3/ELL-rearranged AML: in a 125-patient three-tier model, MLLT3/ELL without PTD had 3-year OS 78.1% versus 50.5% for any PTD and 34.9% for other KMT2A-r (P=0.044), with EFS 71.0% / 40.1% / 24.9% (P=0.003).

    PMID 41121446DOI
    What this evidence cannot show

    125 patients split into three tiers with the model derived and evaluated in the same cohort and no external validation, so the P values are optimistic and the tier boundaries are not independently supported. MLLT3 and ELL are pooled. Whether other series' partner nulls would change after PTD stratification is untested.

    Evidence that argues against this (1)
    • An MD Anderson series of 172 adults found no independent effect of translocation subtype on survival at all, without invoking PTD status, so the PTD interaction has not been shown to rescue a partner effect in an independent cohort. Issa GC et al. 2021
Connected to