Oncofusion

start / graph / epi-jplsg-mll-10-infant-all-risk-without-partner

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

JPLSG MLL-10 (n=90 infants): a 38-point EFS split built from age, CNS status and MRD, with the partner never entering the model

also called Tomizawa 2020 Blood MLL-10 · Japanese Pediatric Leukemia/Lymphoma Study Group MLL-10

A Japanese trial sorted babies into risk groups using their age, whether the leukemia had reached the brain, and how fast it cleared. Those three things split survival from 94% to 57%, and which partner gene was involved was never used.

The same thing, for a scientist

MLL-10 stratified 90 eligible infants with ALL into low risk (KMT2A-germline, n=15), intermediate risk (n=19) and high risk (n=56) using KMT2A status, age and presence of CNS leukaemia - explicitly not the 3-prime partner. Three-year EFS was 93.3% (SE 6.4) for KMT2A-germline, 66.2% (SE 5.6) for all KMT2A-rearranged patients (IR+HR), and within the rearranged group 94.4% (SE 5.4) for IR against 56.6% (SE 6.8) for HR. In multivariable analysis, female sex and MRD at or above 0.01% at end of early consolidation were the significant adverse factors. The KMT2A-rearranged EFS of 66.2% is the best reported for an infant protocol and is roughly 20 points above Interfant-06's whole-trial figure. The partner is absent from the risk algorithm, absent from the multivariable model and absent from the published abstract.

Go deeper
The relevant number is the 38-point split that partner did not produce

Inside the KMT2A-rearranged population, intermediate-risk infants reached 94.4% 3-year EFS and high-risk infants 56.6% 1. That separation was generated by age and CNS involvement, variables that are cheap to measure and were applied prospectively. No partner-based variable in any infant cohort in this beat comes close to a 38-point prospective split. If a beat is asking what stratifies infant KMT2A-r ALL, the honest answer from this trial is: not the partner, and the things that do are already known.

Bounded absence, with the retrieval recorded

PMID 32845001 returned pmcid: null from the Europe PMC core record, so no fullTextXML could be fetched and only the abstract was read. The abstract contains no per-partner arm. The trial registration is UMIN000004801 rather than a ClinicalTrials.gov NCT, so the registry route for arm definitions is also non-standard. Recorded as an absence in the retrievable record.

Why this matters more than it looks

If age, CNS status and MRD carry the stratification that a partner variable would have to compete with, then a partner effect in infants must be large to be detectable and worth acting on. Two infant cohorts that did test partner found nulls 2, 3. The combination - strong non-partner stratifiers that work, plus tested partner nulls - is the strongest case in this beat for the class abstraction in the infant population specifically.

Facts and figures
metric3-year event-free survival by protocol risk group (no partner term)
value66.2
unit% (all KMT2A-rearranged, SE 5.6)
populationinfants with ALL enrolled on JPLSG MLL-10
n total90
partner used in stratificationfalse
registryUMIN000004801
regionJapan
year2020
prespecifiedrisk stratification prospective; no partner comparison specified
sourcePMID 32845001
arms
low risk (KMT2A-germline)15.0low risk (KMT2A-germline): 15intermediate risk (KMT2A-…19.0intermediate risk (KMT2A-r): 19high risk (KMT2A-r)56.0high risk (KMT2A-r): 56
arms — the numbers behind the chart
[
  {
    "group": "low risk (KMT2A-germline)",
    "n": 15,
    "efs_3y_pct": 93.3,
    "se": 6.4
  },
  {
    "group": "intermediate risk (KMT2A-r)",
    "n": 19,
    "efs_3y_pct": 94.4,
    "se": 5.4
  },
  {
    "group": "high risk (KMT2A-r)",
    "n": 56,
    "efs_3y_pct": 56.6,
    "se": 6.8
  }
]
stratification variables
[
  "KMT2A status",
  "age",
  "CNS leukaemia"
]
multivariable significant
[
  "female sex",
  "MRD >=0.01% at end of early consolidation"
]
retrieval record
{
  "europepmc_core_pmcid": null,
  "fulltextxml_fetched": false,
  "conclusion": "no per-partner arm in the retrievable record"
}
How this could bite us back
  • If age, CNS status and MRD already deliver a 38-point split in infants, a menin inhibitor added on a class rationale will be given to a 94.4%-EFS intermediate-risk group in which any added toxicity is pure loss.
  • A trial that stratifies on partner instead of on the variables that demonstrably work would be worse designed than MLL-10 was in 2010.
Still open
  1. What is the partner distribution within MLL-10's intermediate- and high-risk arms? If AFF1 is evenly spread across a 38-point split, that is a stronger partner null than any published comparison.
  2. Does the MLL-10 risk algorithm outperform a partner-based one head to head in the same patients? Nobody has run that comparison.
  3. Is the 66.2% KMT2A-rearranged EFS reproducible outside Japan, or does it reflect cohort composition rather than the algorithm?
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    JPLSG MLL-10 stratified 90 infants with ALL by KMT2A status, age and CNS leukaemia - not by 3-prime partner - into low (n=15), intermediate (n=19) and high (n=56) risk; 3-year EFS was 93.3% (SE 6.4) for KMT2A-germline, 66.2% (SE 5.6) for all KMT2A-rearranged, 94.4% (SE 5.4) for intermediate risk and 56.6% (SE 6.8) for high risk, with female sex and MRD >=0.01% at end of early consolidation the significant multivariable factors.

    PMID 32845001DOI
    What this evidence cannot show

    90 infants in a single national group, so the subgroup standard errors are 5-6 percentage points and the low-risk arm has 15 patients. Not randomised - this is a risk-stratified single-arm protocol with historical comparison. No per-partner arm is reported in the abstract and the full text is not open access (Europe PMC core record returned pmcid null for PMID 32845001), so whether the trial analysed partner at all is unknown. Female sex emerging as a prognostic factor in a 90-patient cohort is the kind of finding that usually does not replicate.

    Evidence that argues against this (1)
    • Interfant-06, with 651 infants, found KMT2A rearrangement, age, WBC and prednisone response prognostic and reported whole-trial 6-year EFS of 46.1%, roughly 20 points below MLL-10's KMT2A-rearranged figure - so the MLL-10 result may reflect cohort selection or national differences as much as the risk algorithm. Pieters R et al. 2019
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    CCG 1953 tested the partner comparison directly in 115 molecularly genotyped infants and found no significant EFS difference between the three major partners, concluding that molecular 11q23 translocation partners do not predict outcome.

    PMID 16556894DOI
    What this evidence cannot show

    Small arms, older protocol era, per-arm n not published; an unpowered null.

    Evidence that argues against this (1)
    • In children over 1 year with the same fusion class, reported 5-year EFS spans 41.7% to 91.2% across partner-by-lineage cells, so partner nulls are specific to infancy rather than general to KMT2A-r ALL. Attarbaschi A et al. 2023
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    A 175-infant BCP-ALL cohort genotyped by RNA-seq found KMT2A::AFF1 2-year EFS 36% (95% CI 21-59) versus 37% (23-60) for all other partners pooled, P=0.9.

    PMID 40000924DOI
    What this evidence cannot show

    Single centre over 13 years, wide confidence intervals, non-AFF1 partners pooled into one comparator arm.

    Evidence that argues against this (1)
    • A 465-case cross-age B-ALL study found AFF1 associated with slower MRD clearance (P=7.0E-04) with fusion partner independent of age and maturity in the adjusted model (P=0.03), so partner is not inert for early response even where survival looks flat. Hartmann AM et al. 2026
Connected to