Oncofusion

start / graph / epi-adult-kmt2a-aml-ell-survival-advantage

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

KMT2A::ELL in adult AML: the one partner that separates, in two independent adult cohorts

also called Shen 2026 Cancers KMT2A fusion partner · ELL-rearranged adult AML outcome

Two recent studies of adults found that one particular partner, ELL, went with noticeably better survival than the others. It is the only partner-specific survival signal that has shown up twice in adults.

The same thing, for a scientist

Two 2025-2026 adult cohorts report a survival advantage for KMT2A::ELL. In a 181-patient adolescent and adult series (median age 33, screened from 3,468 AML patients, all intensively treated), partner distribution was KMT2A::MLLT3 n=39, KMT2A::AFDN n=27, KMT2A::MLLT10 n=25, KMT2A::ELL n=24, other n=12; whole-cohort 3-year OS and EFS were 42.0% and 32.1%, and KMT2A::ELL had superior OS versus all other subtypes (3-year OS 59.8% versus 39.3%, P=0.023). Concomitant mutations did not affect prognosis in that cohort, and age and HSCT in CR1 were the independent factors. Separately, a 125-patient KMT2A-altered AML study found MLLT3/ELL-rearranged cases had superior outcomes unless a concurrent KMT2A-PTD was present, and proposed a three-tier model in which MLLT3/ELL without PTD had 3-year OS 78.1% versus 50.5% (any PTD) versus 34.9% (other KMT2A-r), P=0.044; corresponding EFS 71.0%, 40.1%, 24.9%, P=0.003.

Go deeper
Why ELL is the interesting exception

Two adult cohorts, from different groups, both place ELL on the favourable side 1, 2. Neither is large and neither was pre-specified, but the agreement matters because the two largest adult series report no partner effect at all 3. The most economical reading is that a partner effect exists in adults, is confined to one or two partners, and is too small for a global test across all partners to detect.

The band-versus-gene trap is doing work here

ELL sits at 19p13.1; MLLT1 sits at 19p13.3. Both produce a t(11;19) karyotype. A cytogenetics-only series that records t(11;19) without sub-band resolution therefore pools a partner that looks favourable in adult AML with one that is in the I-BFM high-risk group for childhood AML. If those two are averaged, a real effect in each direction cancels. This is a concrete mechanism by which a well-run adult study could report a true null while a partner effect is present. See concept-cytogenetic-band-versus-molecular-partner-misassignment.

The PTD modifier

The 125-patient study reports that a concurrent KMT2A partial tandem duplication abrogates the MLLT3/ELL survival advantage 2. If that replicates, partner prognosis in adults is not a property of the partner alone but of partner plus a second KMT2A lesion - which would explain some of the inconsistency across series, since most do not test for PTD.

Facts and figures
metric3-year overall survival, KMT2A::ELL versus other partners in adults
value59.8
unit% (KMT2A::ELL)
comparator value39.3
comparator unit% (non-ELL KMT2A-r)
p value0.023
populationadolescents and adults with KMT2A-rearranged AML, intensively treated, median age 33 (13-65)
n total181
replicationsecond cohort (n=125) placed MLLT3/ELL in a favourable tier, 3-year OS 78.1% versus 34.9% for other KMT2A-r (P=0.044), abolished by concurrent KMT2A-PTD
regionChina
year2026
prespecifiedno - retrospective, one partner versus pooled remainder
sourcePMID 41681872
partner counts
other12.0other: 12KMT2A::ELL24.0KMT2A::ELL: 24KMT2A::MLLT1025.0KMT2A::MLLT10: 25KMT2A::AFDN27.0KMT2A::AFDN: 27KMT2A::MLLT339.0KMT2A::MLLT3: 39
partner counts — the numbers behind the chart
{
  "KMT2A::MLLT3": 39,
  "KMT2A::AFDN": 27,
  "KMT2A::MLLT10": 25,
  "KMT2A::ELL": 24,
  "other": 12
}
whole cohort 3y
{
  "os_pct": 42,
  "efs_pct": 32.1
}
How this could bite us back
  • If ELL-rearranged adults already reach 59.8% or 78.1% 3-year OS on chemotherapy plus transplant, they are the worst possible population in which to demonstrate menin-inhibitor benefit, and the best population in which an added toxicity would show up as net harm.
  • A partner effect confined to one partner is exactly the pattern that a class-level eligibility criterion cannot express, so a menin inhibitor labelled for all KMT2A-r will be used in a group where its incremental value differs and is not measured.
Still open
  1. Were the t(11;19) cases sub-banded to distinguish ELL (19p13.1) from MLLT1 (19p13.3) in every adult series reporting a null? If not, the two partners may be cancelling.
  2. Does the ELL advantage survive adjustment for age, given that ELL is reported as enriched in younger patients in some series?
  3. Does KMT2A-PTD status explain the discordance between the adult cohorts that see a partner effect and those that do not?
Every source behind this entry (3)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In 181 adolescents and adults with KMT2A-rearranged AML (MLLT3 n=39, AFDN n=27, MLLT10 n=25, ELL n=24, other n=12), whole-cohort 3-year OS and EFS were 42.0% and 32.1%, and KMT2A::ELL had superior 3-year OS versus non-ELL subtypes (59.8% versus 39.3%, P=0.023); concomitant mutations did not significantly affect prognosis, while age and HSCT in CR1 did.

    PMID 41681872DOI
    What this evidence cannot show

    Retrospective single-cohort study screened from 3,468 AML patients; 24 ELL patients drive the positive finding, and the comparison is one partner against all others pooled rather than a pre-specified multi-arm test. Median follow-up for survivors was only 17.5 months against a 3-year endpoint. Age range 13-65 with median 33, so this is a young adult cohort and does not represent older AML.

    Evidence that argues against this (1)
    • The HARMONY adult AML database found OS similar across translocations in 205 patients, and located the prognostic weight in DNMT3A, KRAS and TP53 mutations rather than partner - directly contradicting the claim that concomitant mutations do not matter in adult KMT2A-r AML. Hernández-Sánchez A et al. 2024
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In 125 patients with KMT2A-altered AML, MLLT3/ELL-rearranged cases showed superior outcomes but a concurrent KMT2A partial tandem duplication abrogated the advantage; a three-tier model gave 3-year OS 78.1% (MLLT3/ELL without PTD) versus 50.5% (any PTD) versus 34.9% (other KMT2A-r), P=0.044, and EFS 71.0% / 40.1% / 24.9%, P=0.003.

    PMID 41121446DOI
    What this evidence cannot show

    125 patients split three ways gives small tiers; the risk model was derived and evaluated in the same cohort with no external validation, so the P values are optimistic. MLLT3 and ELL are pooled into a single favourable category, so the ELL-specific claim is not separable here. KMT2A-r, PTD and dual-alteration cases are combined in one study population.

    Evidence that argues against this (1)
    • An MD Anderson series of 172 adults found translocation subtype had no independent effect on survival at all, so a derived three-tier partner model has not been shown to hold outside the cohort that produced it. Issa GC et al. 2021
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The two largest adult KMT2A-r AML series (n=172 and n=205) both report no partner effect on survival, which is the baseline against which the ELL signal must be judged.

    PMID 38965370DOI
    What this evidence cannot show

    Neither series reports per-partner n in its abstract, so neither null carries a stated power. Both are retrospective with heterogeneous treatment.

    Evidence that argues against this (1)
    • Neither of those series stratified by KMT2A-PTD status, which a later study reports as the modifier that abolishes the MLLT3/ELL advantage - so a null in an unstratified cohort is compatible with a partner effect present only in the PTD-negative subset. Chen L et al. 2025
Connected to