Oncofusion

start / graph / epi-kmt2a-b-all-partner-and-mrd-clearance-across-ages

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

465 KMT2A-r B-ALL cases from 1 month to 89 years: partner predicts MRD clearance independently of age

also called Hartmann 2026 HemaSphere · intra-subtype heterogeneity KMT2Ar B-ALL

By studying leukemias from babies to eighty-nine-year-olds together, this study could ask whether the partner gene matters once age is accounted for. It does: the partner still predicted how fast the leukemia cleared, even after adjusting for age and how mature the cells were.

The same thing, for a scientist

This is the only study located in this beat that separates the partner effect from the age confound by design. 465 KMT2A-rearranged B-ALL cases spanning 1 month to 89 years were profiled by transcriptomics, genomics, ex vivo drug response and MRD kinetics. Partner and age are tightly linked in the cohort (chi-square P<0.0001): AFF1 was the commonest partner (n=221/321, 69%) and was enriched in adult/elderly patients (71% versus 48% infant/paediatric, P=7.7E-12); MLLT3r was found exclusively in infants and children; MLLT1r and MLLT10r were enriched in infants/children (61% and 92% versus 8% and 0.6% in adults, P=0.008 and P=1.5E-4). Univariately, slower MRD clearance correlated with advancing age (P=2.1E-04), lower B-cell maturity score (P=1.3E-03), and AFF1 as fusion partner (P=7.0E-04). In multivariable logistic regression with KMT2A::AFF1 as reference, maturity score (P=0.0455) and MLLT1 fusion (P=0.033) remained significant while MLLT3 did not (P=0.315); the drop-one test confirmed an independent contribution of fusion partner (P=0.03) alongside maturity (P=0.02). A decision tree placed MLLT1r as the split that predicts fast MRD clearance among low-maturity cases.

Go deeper
This is the study the partner-outcome question needed

Every other cohort in this beat is confined to one age band, so partner and age cannot be separated within it. This one spans the whole age range and reports the adjusted model 1. The answer it gives is neither of the two clean answers: partner contributes independently (P=0.03), but the contribution is carried by MLLT1 specifically (P=0.033) and not by MLLT3 (P=0.315), with developmental maturity contributing on its own terms (P=0.0455). Partner matters, but as one term among several and not uniformly across partners.

The endpoint is MRD clearance, not survival

This is the main limitation and it must not be elided. The adjusted analysis is of early response, not EFS or OS. Early response is a strong surrogate in this disease, but a partner that changes MRD kinetics without changing cure rate is entirely possible, and a menin-inhibitor programme that reads this as a survival result would be over-reading it 1.

The confounding is quantified, which is rare

AFF1 at 71% of adult/elderly versus 48% of infant/paediatric cases, MLLT3 exclusively paediatric, MLLT10 at 92% paediatric versus 0.6% adult 1. Those are the numbers that make every single-age-band partner comparison in this beat hard to interpret, and they are worth carrying into any trial design: a partner arm and an age arm are close to the same arm. See concept-partner-age-confound-in-kmt2a-outcome.

Facts and figures
metricindependent effect of KMT2A fusion partner on MRD clearance, adjusted for age and developmental maturity
value0.03
unitP value, drop-one test for fusion partner
populationKMT2A-rearranged B-ALL, age 1 month to 89 years
n total465
n partner counts321
n age correlation305
n mrd categories97
n drug response59
endpoint caveatMRD clearance, not EFS or OS
regionEurope (multi-source assembled cohort)
year2026
prespecifiedno - retrospective integrative analysis
sourcePMID 41743263
partner counts
KMT2A::MLLT1014.0KMT2A::MLLT10: 14KMT2A::MLLT328.0KMT2A::MLLT3: 28KMT2A::MLLT143.0KMT2A::MLLT1: 43KMT2A::AFF1221.0KMT2A::AFF1: 221
partner counts — the numbers behind the chart
{
  "KMT2A::AFF1": 221,
  "KMT2A::MLLT1": 43,
  "KMT2A::MLLT3": 28,
  "KMT2A::MLLT10": 14
}
multivariable terms
{
  "maturity_score_p": 0.0455,
  "fusion_MLLT1_p": 0.033,
  "fusion_MLLT3_p": 0.315,
  "wbc_p": 0.00916,
  "sex_p": 0.21,
  "age_group_infant_p": 0.0892,
  "reference": "KMT2A::AFF1"
}
age enrichment
{
  "AFF1_adult_pct": 71,
  "AFF1_infant_paediatric_pct": 48,
  "AFF1_p": "7.7E-12",
  "MLLT3": "exclusively infant/paediatric",
  "MLLT1_paediatric_pct": 61,
  "MLLT1_adult_pct": 8,
  "MLLT1_p": 0.008,
  "MLLT10_paediatric_pct": 92,
  "MLLT10_adult_pct": 0.6,
  "MLLT10_p": "1.5E-4"
}
How this could bite us back
  • If the partner term acts on MRD clearance and MRD guides treatment intensity, then partner already changes therapy indirectly in every MRD-guided protocol, and a class-level menin inhibitor will be layered onto differently intensified backbones without anyone recording that as a partner effect.
  • Partner and age are so collinear here that a menin-inhibitor trial powered to detect a partner-by-drug interaction would need to break that collinearity deliberately, e.g. by over-recruiting adults with MLLT3 and infants with AFF1 - both rare cells.
Still open
  1. Does the same adjusted model applied to EFS rather than MRD clearance still show an independent partner term?
  2. MLLT1 carries the partner signal here and is favourable; in childhood AML MLLT1 is high-risk. Is that a lineage interaction, and would a partner-by-lineage interaction term in a pooled ALL+AML dataset be significant?
  3. How much of the maturity-score effect is itself downstream of the partner, in which case adjusting for maturity removes part of the partner effect rather than confounding?
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In 465 KMT2A-rearranged B-ALL cases aged 1 month to 89 years, fusion partner remained an independent determinant of MRD clearance in multivariable analysis including WBC, B-cell maturity score, sex and age group (partner drop-one P=0.03; maturity P=0.02), with MLLT1 significant against an AFF1 reference (P=0.033) and MLLT3 not (P=0.315); univariately, slow clearance tracked with advancing age (P=2.1E-04), low maturity (P=1.3E-03) and AFF1 partner (P=7.0E-04).

    PMID 41743263DOI
    What this evidence cannot show

    The endpoint is MRD clearance, not EFS or OS, so this does not establish a partner effect on cure. Partner-age correlation is extreme in the cohort (P<0.0001), so the multivariable model is adjusting for a strongly collinear covariate and the partner coefficient is correspondingly unstable. Not all 465 cases contributed to every analysis: fusion-age correlation used n=305, partner counts n=321, MRD clearance categories n=97, ex vivo drug response n=59. Retrospective, assembled from multiple sources, treatment not uniform. B-ALL only, so nothing follows for AML or T-ALL.

    Evidence that argues against this (2)
    • Two infant cohorts report no partner effect on event-free survival at all (KMT2A::AFF1 2-year EFS 36% versus 37% for other partners, P=0.9), so a partner effect on MRD kinetics has not translated into a survival difference in the age group where it would matter most. Ilyasova K et al. 2025
    • An earlier 497-patient pooled analysis concluded that age was the dominant prognostic factor and that within infants any 11q23 abnormality was equally dismal, which is the competing reading in which partner is a proxy for age. Pui CH et al. 2002
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    KMT2A partner distribution is strongly age-dependent: AFF1 is 71% of adult/elderly versus 48% of infant/paediatric cases (P=7.7E-12), MLLT3 occurs exclusively in infants and children, and MLLT1 and MLLT10 are enriched in infants/children (61% and 92%) versus adults (8% and 0.6%), P=0.008 and P=1.5E-4.

    PMID 41743263DOI
    What this evidence cannot show

    B-ALL only and one assembled cohort; the exact percentages are cohort-specific and referral patterns differ between contributing centres. The denominator differs between the partner-count analysis (n=321) and the age-correlation analysis (n=305).

    Evidence that argues against this (1)
    • The reference KMT2A recombinome, built on 3,401 acute leukaemia patients, is the appropriate external check on these frequencies; partner distribution there is reported by disease and age category and any cohort-specific enrichment should be read against it rather than taken as a population value. Meyer C et al. 2023
  3. E3weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    Ex vivo, the most resistant quartile of induction-drug response was enriched for slow MRD clearance, low maturity score, AFF1 rearrangement and older age (MRD clearance chi-square P=0.003, age group P=0.025, fusion P=0.153, maturity group P=0.064), in n=59 patients with drug-response profiling.

    PMID 41743263DOI
    What this evidence cannot show

    n=59 for the drug-response analysis and the fusion-partner association is not significant (P=0.153); only MRD clearance and age group reach significance. Cited as a direction, not a result. Ex vivo drug response is an imperfect surrogate for clinical response.

    Evidence that argues against this (1)
    • In the same study MLLT3 as a partner had no significant effect on MRD clearance in the adjusted model (P=0.315), so the apparent enrichment of AFF1 among resistant cases does not resolve into a clean partner ordering. Hartmann AM et al. 2026
Connected to