The analysis that would settle it: a partner-by-lineage interaction term in pooled individual-patient data
also called pooled IPD partner interaction test · partner-by-lineage interaction analysis
Every study so far has looked at partners inside one disease or one age group. The decisive test is to put them all together and ask directly whether the partner effect changes depending on the disease and the age.
The same thing, for a scientist
No published analysis has fitted a partner-by-lineage or partner-by-age interaction term. The datasets needed already exist and are held by groups that have already collaborated: I-BFM childhood KMT2A-r AML (n=1,130), Ponte-di-Legno non-infant KMT2A-r ALL (n=629), Interfant-99 and Interfant-06 infant ALL (n=482 and n=651), the international 11q23 paediatric AML series (n=756), and cross-age KMT2A-r B-ALL cohorts (n=465). Pooled, that is well over 3,500 patients with partner assignment. The proposed analysis is a Cox model for EFS with terms for partner (categorical, reference KMT2A::AFF1), lineage, age band, protocol era and, where available, RAS-pathway mutation status, plus partner-by-lineage and partner-by-age interaction terms. Primary readout: whether the interaction terms are significant, and the partner main effect after adjustment.
Why this and not a new trial
The question is observational and the data are already collected. A prospective trial powered on partner arms in this disease would need thousands of patients per rare partner and will never be run. The only realistic route to an answer is pooled individual-patient data, and the collaborations that would have to agree - I-BFM, Ponte-di-Legno, Interfant - overlap heavily in membership and have pooled before.
The controls that make or break it
Protocol era must be in the model: the datasets span 1983-2016 and outcome improved substantially over that period independently of anything about partners. Comutation status should be included where available; the beat found that adult cohorts with NGS locate the prognostic signal in KRAS, DNMT3A and TP53 rather than in the partner, and no paediatric partner analysis has ever been adjusted for RAS-pathway status. Partner assignment method (karyotype band, sub-band, RT-PCR panel, RNA-seq) should be a covariate or a sensitivity stratum, because it determines which partners can be seen at all.
The cells that carry the information
The estimate will be driven by the off-modal cells - adults with KMT2A::MLLT3, infants with rare partners, t(11;19) B-ALL, t(9;11) B-ALL. Those cells are small and are the reason the interaction has not been testable within any single cohort. A power calculation on the pooled counts should be the first step, and if the off-modal cells are too few the honest output is a bound on the interaction rather than an estimate.
| design | pooled individual-patient-data Cox model for EFS with partner (categorical, reference KMT2A::AFF1), lineage, age band, protocol era and RAS-pathway mutation status where available, plus partner-by-lineage and partner-by-age interaction terms |
|---|---|
| readout | significance and magnitude of the interaction terms; partner main effect after adjustment; a bound on the interaction if the off-modal cells are too small to estimate it |
| controls | protocol era as a covariate; partner assignment method (karyotype band / sub-band / RT-PCR panel / RNA-seq) as a sensitivity stratum; pre-registered power calculation on pooled cell counts before any model is fitted |
| cost tier | low |
| timeline | 12-18 months, dominated by data-sharing agreements rather than by analysis |
candidate datasets — the numbers behind the chart
[
{
"cohort": "I-BFM childhood KMT2A-r AML",
"n": 1130,
"pmid": "36996387",
"title": "Measurable Residual Disease and Fusion Partner Independently Predict Survival and Relapse Risk in Childhood KMT2A-Rearranged Acute Myeloid Leukemia: A Study by the International Berlin-Frankfurt-Münster Study Group.",
"authors": "van Weelderen RE, Klein K, Harrison CJ, et al.",
"venue": "J Clin Oncol",
"year": 2023,
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
},
{
"cohort": "Ponte-di-Legno non-infant KMT2A-r ALL",
"n": 629,
"pmid": "36256911",
"title": "Outcomes of Childhood Noninfant Acute Lymphoblastic Leukemia With 11q23/KMT2A Rearrangements in a Modern Therapy Era: A Retrospective International Study.",
"authors": "Attarbaschi A, Möricke A, Harrison CJ, et al.",
"venue": "J Clin Oncol",
"year": 2023,
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
},
{
"cohort": "international 11q23 paediatric AML",
"n": 756,
"pmid": "19528532",
"title": "Novel prognostic subgroups in childhood 11q23/MLL-rearranged acute myeloid leukemia: results of an international retrospective study.",
"authors": "Balgobind BV, Raimondi SC, Harbott J, et al.",
"venue": "Blood",
"year": 2009,
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
},
{
"cohort": "Interfant-06",
"n": 651,
"pmid": "31283407",
"title": "Outcome of Infants Younger Than 1 Year With Acute Lymphoblastic Leukemia Treated With the Interfant-06 Protocol: Results From an International Phase III Randomized Study.",
"authors": "Pieters R, De Lorenzo P, Ancliffe P, et al.",
"venue": "J Clin Oncol",
"year": 2019,
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
},
{
"cohort": "Interfant-99",
"n": 482,
"pmid": "17658395",
"title": "A treatment protocol for infants younger than 1 year with acute lymphoblastic leukaemia (Interfant-99): an observational study and a multicentre randomised trial.",
"authors": "Pieters R, Schrappe M, De Lorenzo P, et al.",
"venue": "Lancet",
"year": 2007,
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
},
{
"cohort": "cross-age KMT2A-r B-ALL",
"n": 465,
"pmid": "41743263",
"title": "Intra-subtype heterogeneity shapes treatment response in KMT2A-rearranged ALL across all age groups.",
"authors": "Hartmann AM, Bastian L, Barz MJ, et al.",
"venue": "Hemasphere",
"year": 2026,
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
},
{
"cohort": "COG AAML0531/1031 MLLT10",
"n": 2080,
"pmid": "38306602",
"title": "Structural variants involving MLLT10 fusion are associated with adverse outcomes in pediatric acute myeloid leukemia.",
"authors": "Abla O, Ries RE, Triche T Jr, et al.",
"venue": "Blood Adv",
"year": 2024,
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
}
]critical cells
[ "adult KMT2A::MLLT3", "infant non-AFF1 partners", "t(11;19) B-ALL", "t(9;11) B-ALL" ]
tests hypotheses
[ "hyp-partner-outcome-separation-versus-menin-class-assumption" ]
- If the interaction is null, the result will be read as vindicating the class abstraction when it may only mean the off-modal cells were too rare to see anything - the power calculation must be published alongside the estimate.
- Pooling across 1983-2016 protocols risks an era effect that mimics or masks a partner effect, because partner detection methods and treatment both changed over that span.
- Are the off-modal cells large enough in the pooled data to estimate an interaction, or will the honest output be a bound?
- Do enough of the contributing cohorts hold RAS-pathway mutation status to include it as a covariate, or would that restrict the analysis to a small subset?
- Should partner assignment method be a covariate or an inclusion criterion? Restricting to molecularly assigned partners loses power but removes the band-ambiguity problem entirely.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The largest childhood AML dataset reports the partner as a grouped binary risk variable rather than per-partner curves, so it cannot on its own support a partner-by-lineage interaction test.
Measurable Residual Disease and Fusion Partner Independently Predict Survival and Relapse Risk in Childhood KMT2A-Rearranged Acute Myeloid Leukemia: A Study by the International Berlin-Frankfurt-Münster Study Group. · van Weelderen RE, Klein K, et al. · J Clin Oncol 2023What this evidence cannot show
The underlying per-partner data exist within the I-BFM collaboration even though the publication reports the grouped variable; this is a reporting limitation rather than a data limitation.
Evidence that argues against this (1)
- The source study for that classification did report per-subgroup hazard ratios for four individual translocations, so partner-level granularity has been published once within the same collaboration and could be reproduced. Balgobind BV et al. 2009
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The one dataset that does span the full age range reports an adjusted partner effect for MRD clearance (drop-one P=0.03) rather than for survival, establishing both that the adjusted analysis is feasible and that the survival version has not been done.
Intra-subtype heterogeneity shapes treatment response in KMT2A-rearranged ALL across all age groups. · Hartmann AM, Bastian L, et al. · Hemasphere 2026What this evidence cannot show
B-ALL only, 465 cases, and the MRD-clearance analysis ran on a subset (n=97 with clearance categories). Extending to survival would require follow-up data that the assembled cohort may not uniformly hold.
Evidence that argues against this (1)
- Two infant cohorts found no partner effect on survival despite the partner effect on early response, so an adjusted survival model may simply return a null and the interaction terms may be unestimable at realistic n. Ilyasova K et al. 2025
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
The partner-by-lineage contrast that motivates the interaction test is currently a juxtaposition of two cohorts: MLLT1 is in the childhood AML high-risk group and is the best-outcome partner-by-lineage cell (91.2%, n=34) in non-infant childhood T-ALL.
Outcomes of Childhood Noninfant Acute Lymphoblastic Leukemia With 11q23/KMT2A Rearrangements in a Modern Therapy Era: A Retrospective International Study. · Attarbaschi A, Möricke A, et al. · J Clin Oncol 2023What this evidence cannot show
n=34 in the favourable cell, no B-lineage comparator for the same partner, retrospective and not pre-specified; the contrast is a motivation for the analysis rather than a result.
Evidence that argues against this (1)
- An MLLT10-centred analysis found that swapping the 5-prime partner entirely, including replacing KMT2A with PICALM, did not change outcome - so fusion-component identity does not always translate into outcome and the interaction may be null. Abla O et al. 2024
the sign flip is the observation the pooled interaction analysis would confirm or kill
the pooled interaction analysis that would falsify or support it
- related to →KMT2A-rearranged acute lymphoblastic leukaemia
pooled individual-patient data spanning ALL and AML cohorts
one of the candidate contributing datasets
- depends on →The protocol and era confound: the partner datasets span 1983 to 2016 and outcome moved under them
era and study group must be covariates or the estimate is uninterpretable