CIBMTR post-transplant AML (n=426 KMT2A-r, curated cytogenetics): no outcome difference by fusion partner
also called Menghrajani 2022 Blood Advances · CIBMTR KMT2A partner post-HCT null
In the largest registry study of transplant outcomes in this leukemia, with the chromosome reports checked by hand, the partner gene made no difference to what happened after transplant.
The same thing, for a scientist
From 8,709 AML patients in the CIBMTR database, 3,779 transplanted in first complete remission entered the final analysis after selection and manual curation of cytogenetics: 2,384 intermediate-risk, 969 adverse-risk and 426 KMT2A-rearranged. In adjusted multivariable analysis, KMT2A-rearranged patients had increased relapse risk versus intermediate-risk (HR 1.27, P=.01), inferior leukaemia-free survival (HR 1.26, P=.002) and inferior overall survival (HR 1.32, P<.001), each comparable to adverse-risk disease. The authors state plainly that **no differences in outcome were detected when patients were stratified by KMT2A fusion partner**. This is the largest post-transplant partner comparison in existence, and the manual cytogenetic curation partially addresses the band-ambiguity problem that weakens most registry analyses.
Why the curation matters
Registry cytogenetics are usually taken as submitted, which is exactly the setting in which t(11;19) and t(10;11) pool two partners each. This study manually curated the cytogenetics before classification 1, which removes the most common mechanical explanation for a partner null. That makes this null stronger than the adult AML nulls that do not state their assignment method.
But the setting selects heavily
Every patient here reached first complete remission and then transplant. Partners that differ in induction failure rate, in early relapse, or in transplant eligibility are filtered out before the analysis starts. If the partner effect acts mainly on achieving and holding CR1 - which is where the I-BFM high-risk group's excess relapse sits - then conditioning on transplant in CR1 would remove exactly that effect and leave a null. This is collider-type selection and it is not a flaw in the study; it is a limit on what the study can answer.
Where it sits in the beat
Together with the two adult AML series and the two infant cohorts, this is the fifth reported partner null. Against it stand four positive datasets, all paediatric and none post-transplant. The pattern - positives in children pre-transplant, nulls in adults and in post-transplant settings - is consistent both with a real paediatric partner effect and with selection and power differences, and this beat cannot separate them.
| metric | post-transplant outcome stratified by KMT2A fusion partner |
|---|---|
| value | no differences detected by fusion partner |
| unit | stratified comparison |
| population | AML patients transplanted in first complete remission, CIBMTR registry |
| n screened | 8709 |
| n analysed | 3779 |
| n kmt2a r | 426 |
| n intermediate risk | 2384 |
| n adverse risk | 969 |
| cytogenetics handling | manual curation before risk classification |
| selection caveat | conditioned on CR1 and transplant, so partner effects acting before transplant are selected out |
| region | international (CIBMTR) |
| year | 2022 |
| prespecified | no - retrospective registry analysis; the partner stratification is a subgroup analysis |
| source | PMID 34551064 |
kmt2a versus intermediate — the numbers behind the chart
{
"relapse_hr": 1.27,
"relapse_p": 0.01,
"lfs_hr": 1.26,
"lfs_p": 0.002,
"os_hr": 1.32,
"os_p": 0.001
}- Menin inhibitors are being positioned largely as a bridge to transplant. If partner differences are erased by transplant, then a partner-stratified response endpoint measured in a bridge-to-transplant population will be measured in exactly the setting where partner stops mattering.
- A large, well-curated null is the most persuasive evidence available for class-wide menin use, and this one is conditioned on a selection step that would remove the effect it failed to find.
- How many of the 426 KMT2A-rearranged patients carried each partner, and how many had no partner identified? Without that the null has no stated power.
- Does a partner effect reappear if the analysis is run from diagnosis rather than from transplant, using the same registry linked to pre-transplant records?
- Is the partner-by-transplant interaction seen in childhood AML present in adults? If so, transplanted cohorts are structurally the wrong place to look for partner effects.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In 3,779 AML patients transplanted in CR1 from the CIBMTR database with manually curated cytogenetics, including 426 with KMT2A rearrangement, no differences in outcome were detected when patients were stratified by KMT2A fusion partner; KMT2A-rearranged disease carried increased relapse risk (HR 1.27, P=.01), inferior leukaemia-free survival (HR 1.26, P=.002) and inferior overall survival (HR 1.32, P<.001) versus intermediate-risk AML.
Risk classification at diagnosis predicts post-HCT outcomes in intermediate-, adverse-risk, and KMT2A-rearranged AML. · Menghrajani K, Gomez-Arteaga A, et al. · Blood Adv 2022What this evidence cannot show
Conditioned on reaching CR1 and undergoing transplant, so any partner effect acting on induction response, early relapse or transplant eligibility is selected out before analysis. Per-partner n are not reported in the abstract, so the null has unstated power; 426 KMT2A-r patients split across partners with an unknown-partner fraction leaves small arms. Registry data with heterogeneous conditioning regimens and donor types, across an unstated span of years. Adults predominate.
Evidence that argues against this (2)
- In childhood KMT2A-r AML, allo-SCT in CR1 reduced relapse risk specifically within the high-risk fusion-partner group (HR 0.5, 95% CI 0.4-0.8, P=.00096) and not within the non-high-risk group, which implies a partner-by-transplant interaction that would flatten partner differences in a transplanted-only cohort exactly as observed here. van Weelderen RE et al. 2023
- In a 181-patient adult cohort, KMT2A::ELL showed superior 3-year OS (59.8% versus 39.3%, P=0.023) in a population where 74 of 181 were transplanted in CR1, so a partner effect can survive in a partly transplanted cohort. Shen H et al. 2026
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Allogeneic transplant in CR1 reduces relapse risk within the high-risk fusion-partner group but not within the non-high-risk group in childhood KMT2A-r AML (HR 0.5, 95% CI 0.4-0.8, P=.00096 versus HR 0.6, 95% CI 0.4-1.0, P=.058), and does not improve overall survival in either.
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
Non-randomised transplant allocation within a retrospective cohort, so transplanted and non-transplanted patients differ systematically; a 90-day landmark analysis was used but residual confounding by indication remains. Transplant rates differed between risk groups (27.6% high-risk versus 17.0% non-high-risk).
Evidence that argues against this (1)
- In non-infant childhood KMT2A-r ALL, allo-HSCT did not improve EFS over chemotherapy alone in either t(4;11) B-ALL or KMT2A-r T-ALL, so a partner-by-transplant interaction is not a general feature of KMT2A-rearranged disease. Attarbaschi A et al. 2023
no outcome difference by fusion partner in 426 transplanted KMT2A-r patients
the largest curated partner null, in a transplanted population
a partner-by-transplant interaction in children would flatten partner differences here
the partner-gene question in post-transplant outcome; answered as a null in AML, unasked in infant ALL