Join the KMT2A breakpoint archive to the Interfant MRD and outcome database, on patient
also called breakpoint-outcome join · intron of origin as a covariate
Every baby who gets the DNA-junction leftover-disease test has the exact position of their broken gene sequenced as a side effect of building the test. Nobody has ever put that position next to how the baby did, even though both pieces of information already sit in databases run by collaborating groups.
The same thing, for a scientist
Interfant-06's MRD relied on genomic KMT2A breakpoints characterised in Frankfurt or Catania [E1], and the Frankfurt network maintains a database of breakpoint information for 3,401 patients with per-patient intron of origin, partner gene, reciprocal status and complex-rearrangement class, which the authors state can be made available to scientists on request [E2]. Interfant holds the MRD values and the outcomes [E1]. The two are linked by patient identity for every infant whose MRD target came from that network. The analysis is to join them and fit breakpoint position, reciprocal allele status and complex-rearrangement class as covariates against MRD at each protocol timepoint and against disease-free survival, adjusting for the Interfant risk group. **Interfant-06's multivariable model did not include breakpoint position** [E3], and that is an absence of testing rather than a negative result.
Why this is the cheapest unexploited analysis in the beat
Three things that usually block a retrospective genotype-outcome study are already solved.
- The genotype is already measured, in every patient. Breakpoint position is not an extra assay; it is a by-product of MRD assay design and exists for every infant with a breakpoint-based MRD target 2.
- The outcome is already curated to trial standard with protocol-defined timepoints and adjudicated events 1.
- The two datasets share patients by construction, because the trial's MRD targets were produced by the breakpoint network 1.
What is being tested is not a new hypothesis. The recombinome already reports that intron 11 breakpoints associate with ALL and younger age while intron 9 predominates in AML and older patients 4, and that a breakpoint upstream of or within intron 11 changes the PHD1-3 architecture of the resulting fusion protein with possible consequences for outcome 2. The missing step is the join.
The secondary question is reciprocal status. AF4-MLL expression carried a hazard ratio of 3.42 in 43 t(4;11) infants on Interfant 99/06, with the Interfant risk group itself non-significant in the same model 5. A larger cohort with genomic reciprocal-allele status - which the breakpoint archive holds - would test whether that survives, and would separate the genomic presence of the reciprocal allele from its expression, which the 43-patient study could not.
What could go wrong, stated in advance
Confounding by partner and age. Partner, age, lineage and breakpoint position are mutually associated 4, and the KMT2A-r infant population is dominated by AFF1. Any breakpoint effect must be estimated within partner strata or it will be a partner effect relabelled.
Multiple comparisons. Intron of origin is a many-level categorical variable and reciprocal status has several classes. Pre-specify a small number of contrasts - intron 11 versus intron 9-10, reciprocal present versus absent, complex versus simple - or the analysis will produce a finding by construction.
Selection. Only patients whose DNA reached the network are in the archive. Interfant-06's own regional survival gap 1 means that selection is likely correlated with outcome, so the join is not a random sample of the trial.
Reciprocal status is partly breakpoint position. Cases expressing no reciprocal transcript are largely those with a spliced-fusion breakpoint upstream of the partner's exon 1 2, so the two covariates are not independent and must be modelled jointly.
| script | analyses/mrd-breakpoint-outcome-join/run.py |
|---|---|
| status | planned |
| join key | patient identity, available because the trial's MRD targets were produced by the breakpoint network |
| adjustment | Interfant risk group; stratified or modelled within partner gene |
| result summary | not yet run |
inputs
[ "Frankfurt DCAL KMT2A breakpoint database (per-patient intron, partner, reciprocal status, complex class)", "Interfant-06 MRD values at TP2/TP4/TP5 and disease-free survival", "Interfant-06 risk group, age, WBC, prednisone response" ]
primary contrasts
[ "intron 11 versus introns 9-10", "reciprocal allele present versus absent", "complex versus simple rearrangement" ]
outcomes
[ "MRD level at each protocol timepoint", "6-year disease-free survival" ]
prespecified hazards
[ "confounding by partner and age", "multiple comparisons across intron levels", "selection into the referral archive correlates with region and outcome", "reciprocal status is partly a relabelling of breakpoint position" ]
tests hypotheses
[ "hyp-breakpoint-position-prognostic-interfant" ]
- A positive result would make breakpoint position a stratifier, which means more infants allocated to transplant in first remission on the basis of an intron. The functional argument for intron 11 rests on PHD1-3 dimerisation in vitro, which is a long way from a treatment allocation.
- The archive is a referral collection and its selection is likely correlated with the same regional factors that produced a 10-12 point survival gap in the trial. A breakpoint effect estimated on it could be a geography effect.
- Three recombinome publications over seventeen years have stated that breakpoint characterisation allows outcome correlation without performing one. That the analysis is obvious and undone is itself a signal worth taking seriously before assuming it is easy.
- Does the Interfant database retain a breakpoint-position field, or only the fact that a patient-specific assay was built? If the latter, the join must go through the Frankfurt archive's own patient records.
- Can genomic reciprocal-allele status, which the archive holds, be separated from reciprocal transcript expression, which the 43-patient prognostic study measured? They are different variables and the literature conflates them.
- Is there enough power within the AFF1 stratum alone to test intron 11 versus intron 9-10, given that partner-stratified analysis is mandatory to avoid confounding?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In Interfant-06 the KMT2A genomic breakpoint fusion sequence was determined at the Frankfurt or Catania reference centres and used as the main MRD-PCR target, MRD was measured at end of induction, end of consolidation and after MARMA in 249 KMT2A-rearranged infants, and 6-year disease-free survival by MRD category was reported at each timepoint; 6-year event-free survival in West European and North American groups was 10-12 percentage points higher than in other countries.
Clinical Implications of Minimal Residual Disease Detection in Infants With KMT2A-Rearranged Acute Lymphoblastic Leukemia Treated on the Interfant-06 Protocol. · Stutterheim J, van der Sluis IM, de Lorenzo P, et al. · J Clin Oncol 2021What this evidence cannot show
Read in full text for the MRD and breakpoint-centre statements; the regional survival gap is reported in the primary trial publication rather than in this paper. Whether Interfant's database retains a breakpoint-position field, as opposed to merely having used the sequence to build a primer, is unknown and is the main feasibility question.
Evidence that argues against this (1)
- MRD was non-mandatory in Interfant-06 and available for 249 of 439 eligible patients, so the joinable subset is smaller again than the trial, and the power of any breakpoint-stratified analysis within it will be modest. Stutterheim J et al. 2021
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The Frankfurt diagnostic network characterised genomic KMT2A breakpoints in 3,401 acute leukaemia patients, recording intron of origin, partner gene, reciprocal fusion alleles, complex rearrangements and out-of-frame fusions in a database the authors state can be made available to scientists on request; breakpoints within KMT2A intron 11 destroy the dimerisation capacity of the PHD1-3 region, which the authors suggest may explain altered outcome; and cases expressing no reciprocal transcript are largely those with a spliced-fusion breakpoint upstream of the partner gene's exon 1.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al. · Leukemia 2023What this evidence cannot show
Read in full text. A referral-based series, so the archive is not a population sample. The data-availability statement is an offer, not a deposited dataset with an accession, so access depends on a collaboration agreement. The PHD1-3 functional consequence is cited from other work rather than measured here.
Evidence that argues against this (1)
- The same paper notes that the intron 11 breakpoint shift observed in adults with MLLT3 fusions was clearly linked to therapy-induced leukaemia and was absent for ELL, EPS15 and SEPTIN6, so breakpoint position is not a uniform variable across partners and a pooled analysis risks averaging opposite effects. Meyer C et al. 2023
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Interfant-06 reported KMT2A rearrangement as the strongest prognostic factor for event-free survival followed by age, white blood cell count and prednisone response, and its risk model and reported prognostic analysis contain no breakpoint-position or reciprocal-status term.
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. · Pieters R, De Lorenzo P, Ancliffe P, et al. · J Clin Oncol 2019What this evidence cannot show
Abstract only; not open access and no PMCID. That breakpoint position is absent from the reported factors is an inference from its non-appearance in the abstract, and the full text may list candidate variables that were considered. This is an absence of testing, and this entry treats it as such rather than as evidence that position does not matter.
Evidence that argues against this (1)
- An absence from a published model may mean the variable was never in the dataset rather than that it was tested and dropped, and if Interfant never recorded a breakpoint-position field the proposed join may not be possible from trial records alone. Stutterheim J et al. 2021
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Across 2,345 acute leukaemia patients, KMT2A breakpoints showed an age-dependent shift with intron 11 breakpoints associated with acute lymphoblastic leukaemia and younger patients and intron 9 breakpoints predominating in acute myeloid leukaemia or older patients, and the authors state that molecular characterisation of breakpoints allows correlation of functional domains of the MLL gene with clinical outcome.
The MLL recombinome of acute leukemias in 2017. · Meyer C, Burmeister T, Gröger D, et al. · Leukemia 2018What this evidence cannot show
The same referral network as E2 with an overlapping patient pipeline, so these are not independent series. The association is cross-sectional and confounded by the strong mutual association of age, lineage and partner gene.
Evidence that argues against this (1)
- The authors state that breakpoint characterisation allows outcome correlation; they do not perform one. The prognostic claim has been available as a stated possibility across three recombinome publications spanning 2006 to 2023 and has not been tested. Meyer C et al. 2023
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In 43 t(4;11)-positive infant B-ALL patients treated on Interfant 99/06, the reciprocal AF4-MLL fusion was expressed in 19 (45%), AF4-MLL/HOXA-expressing patients had 4-year event-free survival 62.4% versus 11.7% (P = 0.001) and overall survival 73.7% versus 25.2% (P = 0.016), and in a Cox model adjusting for Interfant-06 risk stratification the hazard ratio for lacking AF4-MLL expression was 3.42 (95% CI 1.35-8.63, P = 0.01) while risk group was not significant (HR 1.34, P = 0.49).
Unraveling the cellular origin and clinical prognostic markers of infant B-cell acute lymphoblastic leukemia using genome-wide analysis. · Agraz-Doblas A, Bueno C, Bashford-Rogers R, et al. · Haematologica 2019What this evidence cannot show
Read in full text. 43 patients with wide standard errors. AF4-MLL expression and HOXA cluster expression were analysed as a composite, so the prognostic variable cannot be attributed to the reciprocal fusion alone. Expression is an RNA measurement and conflates allele absence with transcriptional silence.
Evidence that argues against this (1)
- Because the reciprocal transcript is absent by construction in spliced-fusion architectures, an RNA-defined reciprocal-negative group mixes patients with no reciprocal allele and patients with a silent one, and in t(4;11) the spliced architecture is vanishingly rare (1 of 1262 cases), so the non-expressing group must be transcriptionally rather than structurally defined. Meyer C et al. 2023
joins the breakpoint archive to the trial outcomes on patient identity
the analysis that entry identifies as cheapest and undone
the outcome database one side of the join comes from