Oncofusion

start / graph / concept-who-gets-transplanted-and-whether-it-helps

Explainermoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q53-transplant

Who gets transplanted in infant KMT2A-rearranged ALL, and whether it helps - four criteria, and a literature that points both ways

also called transplant indication infant ALL · Interfant-06 high-risk criteria · SCT eligibility infant leukaemia · HSCT in first remission infant ALL · does transplant help in t(4;11)

Only a specific group of babies is offered a transplant: those under six months old at diagnosis, with the KMT2A gene rearrangement, with a very high white cell count, and in a first remission. Whether the transplant actually helps them is genuinely disputed - the largest study ever done found that children with this leukaemia did WORSE after a transplant, and other studies found the opposite, and no randomised trial has ever been run.

The same thing, for a scientist

Interfant-06 restricts transplant eligibility to four simultaneous criteria, stated verbatim in the protocol: age at diagnosis under 6 months (under 183 days), MLL rearrangement, initial WBC at or above 300 x10^9/L, and first complete remission - and only if donor criteria are met [E1]. The protocol's own justification is unusually candid: 'The role of high dose chemotherapy and SCT in infant ALL is unclear. There are no data to support extensive use of SCT in infants. No randomised studies have been performed that studied the role of SCT in infant ALL' [E1]. That remains true in 2026. The four principal datasets disagree in direction: Pui 2002 found transplant associated with significantly WORSE survival in 256 patients with t(4;11) [E2]; Mann 2010 found a 64% reduction in relapse risk confined to a 33% subgroup of Interfant-99 [E3]; Kosaka 2004 and Sanders 2005 report favourable single-arm outcomes without chemotherapy controls or waiting-time correction [E4][E5]; and JPLSG MLL-10 restricted transplant, did well, and recommended restricting it further [E6]. The graph records the contest, not a verdict.

Go deeper
The four criteria, verbatim1 deeper

From Interfant-06 chapter 6.1, 'The primary eligibility criteria for SCT are' 1:

  • Age at diagnosis: < 6 months (i.e. <183 days)
  • MLL Rearrangement
  • Initial WBC >= 300 x 10^9/L
  • First complete remission

All four, together, plus an acceptable donor. The donor hierarchy is HLA-identical sibling, then 10/10 matched unrelated or family donor, then 9/10; cord blood is permitted if HLA-identical or closely matched, and the preferred stem cell source is unmanipulated bone marrow. Timing: 'after MARMA so before OCTADA(D) or during OCTADA(D) but not later than 8 months after initial diagnosis', with risk-adjusted bridging chemotherapy if conditioning must be delayed.

Note what is not a criterion. Not minimal residual disease. Not partner gene. Not breakpoint position. Not CNS involvement. Not steroid response - which was the sole stratifier in Interfant-99 and was dropped as a transplant criterion in Interfant-06.

The result is a stratum of roughly 15% of enrolled infants, of whom - per this graph's existing entry concept-hsct-in-kmt2a-r-leukemia - only 46% actually reached a transplant.

Three protocols, three different definitions of who qualifies
protocoltransplant criteria
Interfant-99prednisone poor response only 1
Interfant-06KMT2A-r AND age <6 months AND WBC >=300 x10^9/L AND CR1 1
JPLSG MLL-10KMT2A-r AND (age <180 days OR CNS leukaemia) 6

The Japanese high-risk definition uses OR where Interfant-06 uses AND, and substitutes CNS involvement for white-cell count. These are not the same patients. Any comparison of transplant outcomes between Interfant-06 and MLL-10 is therefore a comparison between differently-constituted groups, and the apparent superiority of MLL-10's transplanted cohort (3-year EFS 56.8% against Interfant-06's high-risk 6-year EFS of 20.9%) is partly a denominator effect before it is anything else.

The contest, laid out1 deeper
studyndesigndirectionthe number
Pui 2002 2256 with t(4;11)retrospective intergroup, 11 study groups, 1983-1995against transplantany transplantation: DFS HR 1.61 (1.10-2.35, p=0.014); OS HR 1.76 (1.08-2.45, p=0.004) - both worse
Kosaka 2004 444 infants, 29 transplanted in CR1single-arm Japanese protocolfor transplant3-year post-transplant EFS 64.4% (46.4-82.4)
Sanders 2005 540 infants, 39 TBI-conditionedsingle-centre, 1982-2003for transplantDFS 76% in CR1; MLL status not prognostic
Mann 2010 3277 KMT2A-r infants in CR1, 37 transplantedInterfant-99 analysisfor transplant, in 33% of patients onlyHR 0.36 (0.15-0.86) for relapse or death, restricted to age <6 months plus poor steroid response or WBC >=300; no advantage at all in the rest
JPLSG MLL-10 656 high-risk, 43 transplanted in CR1single-arm, restricted indicationfor a restricted indication3-year EFS 56.8%, OS 80.2% - and the authors recommend restricting further or eliminating

The largest single dataset in the table is the one against. Pui 2002 pooled 497 children with 11q23 abnormalities from eleven groups, and in the 256-patient t(4;11) subgroup transplantation of any kind was associated with worse disease-free and overall survival, with the paper stating that 'even transplantation with stem cells from HLA-matched related or HLA-matched unrelated donors tended to be associated with a worse outcome than chemotherapy alone' 2.

And it is the oldest. Its patients were treated between 1983 and 1995, with the transplant practice and supportive care of that era, and transplant was then given preferentially to patients who were doing badly - confounding by indication that a retrospective analysis cannot remove. Interfant-06's own protocol cites it as one of the reasons for restricting transplant rather than for abandoning it 1.

Mann 2010 is the most useful, and it is a subgroup finding. The benefit was confined to 97 of 297 patients (33%), and in the remaining two thirds 'there was no advantage of HSCT over chemotherapy at all' 3. That is simultaneously the best argument for transplanting some infants and the best argument against transplanting most of them.

Why this cannot be settled retrospectively

Every comparison in that table is vulnerable to guarantee-time bias: to be transplanted you must survive to transplant, so the transplanted group is selected for not having relapsed or died in the interval. Interfant-99's own analysis states that after correction for waiting time to SCT, the event-free survival curves of poor-responders who did and did not receive a transplant 'were not significantly different' 1.

And the bias runs both ways. Where transplant is offered preferentially to poor responders, the transplanted group is selected for worse biology, which pushes the estimate against transplant - which is the most likely reading of Pui 2002. Guarantee-time bias inflates the apparent benefit; confounding by indication deflates it. The two errors are both present, in opposite directions, in every study in the table, and their net is unknown.

The protocol's conclusion follows: 'In view of the uncertainties about the efficacy and risks of SCT only infants of the newly defined HR category in Interfant-06 will be eligible for SCT' 1. Restriction under uncertainty, not a claim of benefit.

What has changed since, and what has not

Changed. The criteria are no longer the only lever. JPLSG MLL-10 found that flow-cytometric MRD at the end of consolidation was the sole factor associated with poor EFS on multivariable analysis (p<0.001), outperforming the Interfant high-risk criteria themselves 6. A response-based indication is now demonstrably more discriminating than the presentation-based one Interfant-06 uses - and Interfant-06's criteria contain no response variable at all.

Not changed. There is still no randomised trial. The sentence in the 2006 protocol - 'No randomised studies have been performed that studied the role of SCT in infant ALL' - is as true on 2026-09-20 as it was when it was written, and there is no registered study that would change it. Given how few infants there are, and how strongly the field has moved towards bridging with blinatumomab and menin inhibitors instead, there probably never will be one.

So the honest position for this graph is: the indication is a convention supported by one subgroup analysis, contradicted by the largest historical dataset, and not testable by the method that would settle it. Any entry in this graph asserting that transplant benefits high-risk KMT2A-rearranged infants is asserting more than the literature supports; so is any entry asserting that it harms them.

Facts and figures
criteria logicALL FOUR simultaneously, plus a suitable donor
timingafter MARMA, before or during OCTADA(D), and not later than 8 months after initial diagnosis; risk-adjusted bridging chemotherapy if conditioning must be postponed
preferred stem cell sourceunmanipulated bone marrow; minimum 3x10^8/kg nucleated cells or 3x10^6/kg CD34+ cells
competing criteria jplsg mll10KMT2A-r AND (age <180 days OR CNS leukaemia) - an OR where Interfant-06 uses an AND
randomised evidenceNONE. Interfant-06's protocol states 'No randomised studies have been performed that studied the role of SCT in infant ALL' and that remained true on 2026-09-20; no registered study would change it.
the two biasesGuarantee-time bias inflates apparent transplant benefit (you must survive to be transplanted); confounding by indication deflates it (in the Pui era transplant went preferentially to poor responders). Both are present in every study in the contest, in opposite directions, and their net is unknown.
interfant06 sct criteria
[
  "Age at diagnosis < 6 months (i.e. <183 days)",
  "MLL rearrangement",
  "Initial WBC >= 300 x10^9/L",
  "First complete remission"
]
not criteria
[
  "minimal residual disease",
  "KMT2A partner gene",
  "breakpoint position",
  "CNS involvement",
  "prednisone response (the sole Interfant-99 stratifier, dropped as an SCT criterion in Interfant-06)"
]
donor hierarchy
[
  "HLA-identical sibling (MSD)",
  "10/10 identical unrelated or family donor (MD)",
  "9/10 identical unrelated or family donor (MD)",
  "umbilical cord blood if HLA-identical or closely matched; >7/8 high-resolution match if no MD identified"
]
evidence directions
{
  "against": "Pui 2002, n=256 t(4;11): DFS HR 1.61 (1.10-2.35), OS HR 1.76 (1.08-2.45) - transplant worse",
  "for_a_subgroup": "Mann 2010, Interfant-99: HR 0.36 (0.15-0.86) restricted to 97/297 (33%); no advantage in the other two thirds",
  "single_arm_favourable": "Kosaka 2004 3-y post-transplant EFS 64.4%; Sanders 2005 DFS 76% in CR1 (39/40 TBI-conditioned)",
  "restricted_and_recommending_further_restriction": "JPLSG MLL-10: 3-y EFS 56.8%, OS 80.2%, authors recommend restricting or eliminating"
}
How this could bite us back
  • The indication and its evidence are not independent. Interfant-06's four criteria were derived from the Interfant-99 subgroup analysis and then applied prospectively, so the prospective experience cannot validate the criteria - it can only measure what happens when they are followed.
  • Improving access to a treatment of unproven benefit increases exposure to its harms with certainty and its benefits only hypothetically. Every non-chemotherapy agent framed in this graph as a bridge to transplant is, on the Pui 2002 reading, a bridge to an intervention that made this population worse.
  • The two biases cancel to an unknown. Guarantee-time bias and confounding by indication are both present in every study here and point in opposite directions, so the field's inability to agree is a property of the evidence rather than of the reading.
  • A four-way AND produces a small and unstable stratum. Roughly 15% of enrolled infants qualify, fewer than half reach transplant, and a criterion set that intersects four conditions is maximally sensitive to how each one is measured - a white-cell count at 295 versus 305 x10^9/L changes the treatment plan.
  • The best-outcome transplanted cohort in the literature is reported by investigators recommending the practice be eliminated. If the people who did it best think it should be done less, a graph entry that treats transplant as standard of care is not representing the field.
Still open
  1. Should the transplant indication be response-based rather than presentation-based? End-of-consolidation MRD was the sole multivariable predictor of poor EFS in MLL-10 and outperformed the Interfant criteria on the same cohort, yet Interfant-06's four criteria contain no response variable.
  2. Can the Pui 2002 result be re-analysed with waiting-time correction? It is the largest dataset in the contest and the only one pointing against transplant, and its hazard ratio has never been recomputed with the landmark or time-dependent methods that would separate guarantee-time bias from confounding by indication.
  3. Does the transplant benefit, if any, differ by KMT2A partner gene? This graph already records that a CIBMTR post-transplant analysis of KMT2A-rearranged AML found no separation by partner, and no equivalent analysis exists in infant ALL.
  4. If bridging therapy (blinatumomab, menin inhibitors) succeeds in getting more high-risk infants to transplant, does the transplant still help - or does raising the fraction reaching transplant from 46% simply expose more infants to conditioning toxicity for a benefit that was never demonstrated?
Every source behind this entry (6)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.book

    Interfant-06 restricts stem-cell transplant eligibility to infants meeting all four criteria of age at diagnosis under 6 months (under 183 days), MLL rearrangement, initial WBC at or above 300 x10^9/L and first complete remission, with a suitable donor; the protocol states that the role of SCT in infant ALL is unclear, that there are no data to support its extensive use, that no randomised studies have been performed, and that after correction for waiting time to SCT the Interfant-99 data showed no difference in outcome for poor prednisone responders who did or did not receive a transplant.

    INTERFANT-06: International collaborative treatment protocol for infants under one year with acute lymphoblastic or biphenotypic leukemia (protocol document, SKION version 01, April 2006) · Pieters R, de Lorenzo P, Valsecchi MG, Peters C, and the Interfant-06 study group · Interfant-06 study group / SKION protocol document 2006
    What this evidence cannot show

    Sections 1.2.3, 1.3.9 and 6.1-6.2 of the 2006 SKION protocol document, retrieved and read in full on 2026-09-20. This is a protocol and states eligibility, not practice: the published Interfant-06 outcome paper reports that only 46% of high-risk infants actually received a transplant, so the criteria describe who was offered one rather than who received one. Later protocol amendments were not retrieved, and it is possible the criteria were modified during the trial's accrual. The waiting-time-corrected Interfant-99 analysis is summarised in the protocol's own words and its underlying numbers (n=16 transplanted, n=56 not) are small.

    Evidence that argues against this (1)
    • A response-based criterion outperformed these presentation-based criteria when both were tested on the same cohort. In JPLSG MLL-10, univariable analysis showed both the Interfant high-risk criteria and flow-cytometric MRD at end of induction and end of consolidation were associated with poorer EFS, but on multivariable analysis positive MRD at end of consolidation was SOLELY associated with poor EFS (p<0.001). A transplant indication built from age and white-cell count at diagnosis is being outperformed by a measurement taken months later. Takachi T et al. 2021
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In a retrospective intergroup analysis of 497 children and young adults with 11q23 abnormalities treated by 11 study groups between 1983 and 1995, any type of transplantation in the largest subgroup - 256 patients with t(4;11) - was associated with significantly WORSE disease-free survival (HR 1.61, 95% CI 1.10-2.35, p=0.014) and overall survival (HR 1.76, 1.08-2.45, p=0.004) than chemotherapy alone, and even HLA-matched related or unrelated donor transplantation tended to be associated with a worse outcome than chemotherapy alone.

    Outcome of treatment in childhood acute lymphoblastic leukaemia with rearrangements of the 11q23 chromosomal region. · Pui CH, Gaynon PS, Boyett JM, Chessells JM, Baruchel A, Kamps W, Silverman LB, Biondi A, et al. · The Lancet 2002
    PMID 12057554DOI
    What this evidence cannot show

    Retrospective, pooled across eleven groups with heterogeneous chemotherapy backbones, and treating patients from 1983-1995 - two to four decades before current transplant practice, HLA typing resolution, supportive care and GVHD prophylaxis. It is almost certainly confounded by indication in the direction of its own conclusion: in that era transplant was offered preferentially to children responding poorly, so the transplanted group carried worse biology. The analysis is not restricted to infants; it covers children and young adults with t(4;11), and age was its strongest prognostic factor. A hazard ratio of 1.61 for 'any type of transplantation' pools related, unrelated and autologous grafts.

    Evidence that argues against this (1)
    • The Interfant-99 analysis found a real benefit in a defined subgroup using a contemporary protocol and prospective data collection: of 277 KMT2A-rearranged infants in first remission, 37 transplanted and 240 on chemotherapy alone, the disease-free survival advantage (p=0.03) was restricted to the 97 of 297 (33%) with age under 6 months plus poor steroid response at day 8 or WBC at or above 300 x10^9/L, where HSCT was associated with a 64% reduction in the risk of relapse or death in CR (HR 0.36, 95% CI 0.15-0.86). Mann G et al. 2010
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In Interfant-99, of 277 KMT2A-rearranged infants in first complete remission, 37 received HSCT and 240 chemotherapy only; the disease-free survival advantage (p=0.03) was restricted to the 97 of 297 (33%) with age under 6 months plus either poor steroid response at day 8 or WBC at or above 300 x10^9/L, where HSCT was associated with a 64% reduction in the risk of relapse or death in first remission (HR 0.36, 95% CI 0.15-0.86), with no advantage of HSCT over chemotherapy in all remaining patients.

    PMID 20592248DOI
    What this evidence cannot show

    A subgroup analysis of a non-randomised comparison: transplant was not allocated, so the comparison is observational within a trial. Thirty-seven transplanted patients across the whole cohort and a benefit confined to a subgroup of 97 means the decisive hazard ratio rests on a small number of events, and its confidence interval (0.15-0.86) is correspondingly wide. This is the analysis that DEFINED the Interfant-06 high-risk criteria, so Interfant-06's eligibility rule and this evidence are not independent - the criteria were derived from this subgroup and then applied prospectively, which is a legitimate design and not a validation.

    Evidence that argues against this (1)
    • The largest historical dataset points the other way in the same disease: in 256 patients with t(4;11), any type of transplantation was associated with worse disease-free survival (HR 1.61, 95% CI 1.10-2.35) and worse overall survival (HR 1.76, 1.08-2.45) than chemotherapy alone. A subgroup benefit of HR 0.36 in 97 patients and a whole-cohort harm of HR 1.61 in 256 patients are both on the record, in different eras, with opposite confounding. Pui CH et al. 2002
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In 44 infants with MLL-rearranged ALL treated with intensive chemotherapy followed by HSCT between 1998 and 2002, remission induction was 91.0% and 3-year overall and event-free survival were 58.2% and 43.6%; the 3-year post-transplant EFS for the 29 patients transplanted in first remission was 64.4% (95% CI 46.4-82.4), and only the timing of HSCT (first remission versus later) was significant on multivariable analysis.

    Infant acute lymphoblastic leukemia with MLL gene rearrangements: outcome following intensive chemotherapy and hematopoietic stem cell transplantation. · Kosaka Y, Koh K, Kinukawa N, Wakazono Y, Isoyama K, Oda T, Hayashi Y, Ohta S, et al. · Blood 2004
    PMID 15297313DOI
    What this evidence cannot show

    Single-arm with no chemotherapy-only comparator and no correction for waiting time to transplant, so the 64.4% post-transplant EFS in CR1 patients is subject to guarantee-time bias in full force - reaching CR1 transplant requires surviving to it. Forty-four patients, 29 transplanted. The finding that 'only the timing of HSCT was a significant risk factor' is close to a restatement of that bias. The authors' own conclusion is hedged towards less toxicity and better patient selection, not towards transplanting more infants: 'early introduction of HSCT, possibly with a less toxic conditioning regimen, may improve the prognosis' and 'identification of subgroups or patients who respond well to intensified chemotherapy alone should have a high priority'.

    Evidence that argues against this (1)
    • Interfant-06's protocol committee read this exact paper and declined to act on it, naming the reasons: 'these studies did not have a control arm in which patients only received chemotherapy and the data were not corrected for waiting time to SCT'. A favourable single-arm outcome in transplanted patients is the canonical shape of a guarantee-time artefact, and the group that would most have liked to act on it did not. Pieters R et al. 2006
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In 40 infants diagnosed before 12 months and transplanted between 1982 and 2003, disease-free survival was 76% for the 17 transplanted in first complete remission versus 45% in CR2/CR3 and 8% in relapse (p<0.001); 26 of 33 with cytogenetic data had MLL rearrangement, 14 of these were in CR1 and 11 survive in remission, and MLL rearrangement was not predictive of outcome.

    Allogeneic hematopoietic cell transplantation for infants with acute lymphoblastic leukemia. · Sanders JE, Im HJ, Hoffmeister PA, Gooley TA, Woolfrey AE, Carpenter PA, Andrews RG, Bryant EM, et al. · Blood 2005
    PMID 15637143DOI
    What this evidence cannot show

    Single centre, 40 patients over 21 years, 17 in the CR1 group that carries the headline number, no chemotherapy comparator and no waiting-time correction. Thirty-nine of 40 were conditioned with cyclophosphamide plus total body irradiation, a regimen that is now prohibited in this age group by Interfant-06, so the efficacy result cannot be transferred to current infant conditioning. Donors were mismatched related in 16 and unrelated in 16, reflecting pre-modern HLA matching. The finding that 'having the MLL gene was not a factor predictive of outcome' is based on 26 MLL-rearranged patients and is a null in a very small sample.

    Evidence that argues against this (1)
    • The same intergroup era produced the opposite conclusion at ten times the sample size: in 256 patients with t(4;11), any transplantation was associated with worse disease-free and overall survival than chemotherapy alone. And the conditioning that produced the 76% figure - cyclophosphamide plus TBI in 39 of 40 infants - is the exact regimen Interfant-06 cites as causing 'severe late effects in infants' and forbids. Pui CH et al. 2002
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In JPLSG MLL-10, transplant was restricted to high-risk patients defined as KMT2A-rearranged AND either age under 180 days OR central nervous system leukaemia; 43 of 56 high-risk patients received HSCT in first remission, giving 3-year event-free survival of 56.8% (95% CI 42.4-68.8) and overall survival of 80.2% (67.1-88.5), with positive end-of-consolidation MRD the sole multivariable predictor of poor EFS, and the authors concluded the indication should be further restricted or even eliminated in future.

    Hematopoietic stem cell transplantation for infants with high-risk KMT2A gene-rearranged acute lymphoblastic leukemia · Takachi T, Watanabe T, Miyamura T, Moriya Saito A, Deguchi T, Hori T, Yamada T, Ohmori S, et al. · Blood Advances 2021
    PMID 34500465DOI
    What this evidence cannot show

    Single-arm, so the 56.8% EFS has no chemotherapy comparator and is subject to the same guarantee-time bias as every other study in this contest. The high-risk definition uses OR where Interfant-06 uses AND and substitutes CNS leukaemia for white-cell count, so this cohort is NOT the Interfant-06 high-risk population and the outcomes are not directly comparable. The conditioning also differs (targeted busulfan, etoposide, cyclophosphamide rather than busulfan, cyclophosphamide, melphalan), so protocol, criteria and conditioning all differ simultaneously and no single factor can be credited.

    Evidence that argues against this (1)
    • The authors' own concluding sentence argues against reading their favourable numbers as support for the practice: 'considering the high rate of potentially life-threatening toxicities and the risk of late effects, its indication should be further restricted or even eliminated in the future by introducing more effective therapeutic modalities with minimal toxicities.' The best-performing transplanted infant cohort in the literature is reported by investigators recommending that fewer infants be transplanted. Takachi T et al. 2021
Connected to