Oncofusion

start / graph / concept-lymphoid-vs-myeloid-consolidation

Explainerstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q6b-therapeutics2

Lymphoid-style versus myeloid-style consolidation — the question asked twice in 1,100 infants, and answered no both times

also called ADE MAE versus protocol IB · AML-type consolidation in infant ALL · hybrid protocol rationale

Because KMT2A-rearranged infant leukemia looks partly like lymphoid and partly like myeloid leukemia, doctors twice tested whether adding AML-style chemotherapy would help. Both times, across more than a thousand babies, it made no difference at all.

The same thing, for a scientist

KMT2A-rearranged infant leukemia is lineage-ambiguous: it co-expresses myeloid antigens, a substantial fraction meets criteria for mixed phenotype acute leukemia, and it switches lineage under pressure. That biology motivated two randomised tests of adding AML-directed chemotherapy blocks to an ALL backbone. Interfant-99 randomised 191 infants to an additional late intensification course: 4-year DFS 60.9% (SE 5.2) versus 57.0% (SE 5.5), p=0.81. Interfant-06 randomised 330 medium- and high-risk infants to myeloid-type ADE plus MAE versus lymphoid-type protocol IB: 6-year DFS 39.3% (SE 4.0) versus 36.8% (SE 3.9), p=0.47. Both negative, with no reported subgroup benefit. The general lesson is that lineage plasticity is not the same as drug-class sensitivity.

Go deeper
What the two regimens are, concretely

Lymphoid-style consolidation (Interfant-06 protocol IB): low-dose cytarabine, 6-mercaptopurine and cyclophosphamide. This is the standard consolidation block of a paediatric ALL regimen — moderate intensity, delivered over weeks, relying on antimetabolites.

Myeloid-style consolidation (ADE then MAE): cytarabine + daunorubicin + etoposide, then cytarabine + mitoxantrone + etoposide. These are AML induction-intensity blocks — high-dose cytarabine, two anthracycline-class drugs and a topoisomerase II poison, delivered as short intense courses with profound aplasia between them.

They differ in intensity, in drug class, and in the toxicity they impose on a baby. ADE+MAE in particular delivers both daunorubicin and mitoxantrone to a child under one year old 2.

The two answers1 deeper
TrialComparisonResultp
Interfant-99Additional late intensification course vs none, 191 randomised4-year DFS 60.9% (SE 5.2) vs 57.0% (SE 5.5)0.81
Interfant-06ADE+MAE (n=169) vs protocol IB (n=161)6-year DFS 39.3% (SE 4.0) vs 36.8% (SE 3.9)0.47

12

Neither abstract reports a subgroup — including prednisone poor responders or the high-risk stratum — in which the myeloid-style arm won. We could not access the full-text subgroup tables, so we state that no benefit was reported, not that none exists.

Interfant-06's authors also state flatly that outcome in Interfant-06 did not improve compared with Interfant-99 2. Twelve years and 651 additional infants, and the survival curve did not move.

A third, independent negative belongs alongside these: COG P9407 Cohort 3, which tested intensified chemotherapy without transplant in 147 infants, gave 5-year EFS 42.3±6% and OS 52.9±6.5%, with MLL-rearranged infants over 90 days old at 43.8±8% versus 69.1±13.6% for MLL-germline 3. Intensification, in three separate international attempts, did not solve this disease.

The one intensification lesson that did work — in reverse

The intervention on the chemotherapy backbone that produced the largest effect in this disease was de-intensification. In COG AALL0631, switching from a COG P9407-style 3-week induction to an Interfant-99-style 5-week induction with enhanced supportive care reduced induction mortality from 15.4% to 1.6% (p=0.009), cut sterile-site infections from 57.7% to 19.5% (p=0.0002), and raised the complete response rate from 68% to 94% (p=0.0012) 4.

Less chemotherapy, better supportive care, fewer deaths, more remissions. That is the single most useful thing the chemotherapy era established, and it is the reason every new agent in this disease is now evaluated on whether it can add efficacy without adding myelosuppression.

Why the seductive reasoning was wrong

The argument for myeloid-style consolidation was: this leukemia is biologically part-myeloid, so treat it with myeloid drugs. The flaw is a category error. Lineage identity and drug sensitivity are set by different things.

Lineage antigen expression reflects the transcriptional program the fusion imposes. Sensitivity to cytarabine, anthracyclines or etoposide reflects nucleoside metabolism, drug transport, DNA repair capacity and apoptotic priming — none of which is what CD13 or CD33 expression tells you. A blast can look myeloid and still be exactly as chemo-resistant as it was when it looked lymphoid, because the fusion drives self-renewal and the differentiation block, not drug efflux.

The corollary for the current era is direct: the same error is available today. 'This leukemia is lineage-plastic, so let us target both lineages immunologically' is a good argument, because antigen expression is exactly what an antibody or CAR sees. 'This leukemia is lineage-plastic, so let us use both classes of cytotoxic' was a bad argument, and it cost two decades and more than a thousand infants to establish that.

Facts and figures
questionDoes adding AML-directed (myeloid-style) chemotherapy blocks to an ALL (lymphoid-style) backbone improve outcome in infant KMT2A-rearranged leukemia?
answerNo. Two randomised tests, both negative, with no reported subgroup benefit.
lymphoid regimenprotocol IB: low-dose cytarabine, 6-mercaptopurine, cyclophosphamide
myeloid regimenADE (cytarabine/daunorubicin/etoposide) then MAE (cytarabine/mitoxantrone/etoposide)
total infants randomised521
the positive resultde-intensification: AALL0631 induction mortality 15.4% -> 1.6%, CR 68% -> 94%
general lessonlineage plasticity is not drug-class sensitivity; antigen expression predicts what an antibody sees, not what a cytotoxic does
randomised tests
[
  {
    "trial": "Interfant-99",
    "n_randomised": 191,
    "endpoint": "4-year DFS",
    "intervention_pct": 60.9,
    "control_pct": 57,
    "se": [
      5.2,
      5.5
    ],
    "p": 0.81
  },
  {
    "trial": "Interfant-06",
    "n_randomised": 330,
    "endpoint": "6-year DFS",
    "intervention_pct": 39.3,
    "control_pct": 36.8,
    "se": [
      4,
      3.9
    ],
    "p": 0.47
  }
]
supporting negative
{
  "trial": "COG P9407 Cohort 3",
  "n": 147,
  "efs_5y_pct": 42.3,
  "se": 6,
  "os_5y_pct": 52.9,
  "os_se": 6.5
}
How this could bite us back
  • Toxicity without benefit: ADE+MAE delivers daunorubicin plus mitoxantrone to an infant for no survival gain, and anthracycline exposure in the first year of life is the main driver of lifelong cardiomyopathy risk in these survivors.
  • Infection: 49% of Interfant-99 randomised-intensification patients with follow-up data had infections, and infection is the dominant cause of treatment-related death in infant ALL.
  • The saturation point: two negative randomisations in 521 infants mean cytotoxic dose intensity is a spent lever. Any future proposal to intensify chemotherapy in this disease carries the burden of explaining why it differs from these two.
  • The reasoning trap is still live: mechanistic plausibility ('the leukemia is part-myeloid', 'the leukemia is chromatin-driven') is not evidence of drug sensitivity, and this disease has already punished that inference once at enormous cost.
Still open
  1. Was there ever an interaction between fusion partner and consolidation type? KMT2A::MLLT3 and KMT2A::MLLT10 are myeloid-biased partners and KMT2A::AFF1 is lymphoid-biased; a partner-stratified reanalysis of Interfant-06 has never been reported and the data exist.
  2. Did myeloid-style consolidation reduce the rate of myeloid lineage switch at relapse, even without improving DFS? That would be a real biological effect hidden by a null primary endpoint, and it is directly relevant now that CD19-directed therapy is making lineage switch more common.
  3. How much of the residual mortality in these protocols is treatment-related rather than leukemia-related? The AALL0631 induction result implies the fraction is larger than usually assumed, which changes the calculus for every added agent.
  4. Is the same category error being repeated with epigenetic therapy — treating a chromatin-driven disease with chromatin drugs because the mechanism matches, rather than because the dependency is load-bearing?
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In Interfant-99, 191 of 445 infants in complete remission after induction were randomised to an additional late intensification course (n=95) or not (n=96), with 4-year disease-free survival 60.9% (SE 5.2) versus 57.0% (SE 5.5), p=0.81; delayed intensification of chemotherapy did not benefit patients.

    PMID 17658395DOI
    What this evidence cannot show

    From the primary abstract; full text not retrieved, so no confidence intervals on the DFS estimates and no subgroup tables were read. Only 191 of 445 eligible infants were randomised, so the comparison is underpowered relative to the cohort and subject to selection of who was offered randomisation.

    Evidence that argues against this (2)
    • The randomised subset is a survivor-selected minority, so its absolute numbers do not describe infant ALL. Only 191 of 445 infants (43%) reached this randomisation, because reaching it required complete remission after induction — the infants who died in induction or never entered remission are absent by construction, which is why the arms sit near 60% DFS. When the same intensification question was re-asked prospectively in Interfant-06 across 651 infants with 330 randomised (myeloid-type ADE+MAE n=169 vs lymphoid-type IB n=161), 6-year disease-free survival was 39.3% (SE 4.0) versus 36.8% (SE 3.9), log-rank P=.47 — again no difference, but roughly 20 percentage points lower in absolute terms, and the authors concluded outcome in Interfant-06 did not improve on Interfant-99. A null result in a selected subgroup and a null result in the full protocol are different findings that happen to agree. Pieters R et al. 2019
    • A second reason the null result is not the whole story: consolidation type INTERACTS with end-of-induction MRD, so averaging over unstratified patients hides opposite effects. In 249 infants with KMT2A-rearranged ALL on Interfant-06, among infants MRD-NEGATIVE at end of induction lymphoid-style consolidation gave 6-year DFS 78.2% (95% CI 51.4-91.3) versus 45.0% (23.9-64.1) for myeloid-style, while among infants with HIGH end-of-induction MRD the ordering reversed, 45.9% (29.4-60.9) for myeloid-style versus 23.2% (12.1-36.4) for lymphoid-style. An unstratified randomisation that reports no difference is consistent with two real and opposite subgroup effects cancelling, which is a different conclusion from 'intensification does not help'. Stutterheim J et al. 2021
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In Interfant-06, medium- and high-risk infants were randomised to myeloid-type ADE (cytarabine/daunorubicin/etoposide) plus MAE (cytarabine/mitoxantrone/etoposide) versus lymphoid-type protocol IB (low-dose cytarabine, 6-mercaptopurine, cyclophosphamide), with 6-year disease-free survival 39.3% (SE 4.0, n=169) versus 36.8% (SE 3.9, n=161), log-rank p=0.47; early intensification with postinduction myeloid-type courses did not improve outcome and outcome did not improve compared with Interfant-99.

    PMID 31283407DOI
    What this evidence cannot show

    From the primary abstract; the abstract reports standard errors, not 95% confidence intervals, and the full text with subgroup tables was not retrieved. The randomisation applied only to medium- and high-risk patients who achieved remission, so it says nothing about myeloid-style therapy given during induction or to induction failures.

    Evidence that argues against this (1)
    • The 'answered no both times' framing is contradicted by a third asking of a related question that answered yes. The JPLSG MLL-10 trial introduced high-dose cytarabine - a myeloid-type agent - into EARLY intensification under a three-tier risk stratification and reported 3-year EFS of 66.2% (SE 5.6) for KMT2A-rearranged infants, against Interfant-06's 6-year EFS of 46.1% for the whole cohort. Interfant-99 randomised a LATE course in patients already in CR and Interfant-06 randomised post-induction blocks in MR/HR patients; neither tested myeloid-type intensity early and to all comers. The counterweight is that MLL-10 is a single-country, non-randomised, 90-patient study with shorter follow-up and its comparison with Interfant is historical. Tomizawa D et al. 2020
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In COG P9407 Cohort 3, intensified chemotherapy without stem cell transplant in 147 infants gave 5-year EFS 42.3±6% and OS 52.9±6.5%; infants aged 90 days or younger had 5-year EFS 15.5±10.1% versus 48.5±6.7% for those older than 90 days (p<0.0001), and among infants over 90 days, MLL-rearranged 5-year EFS was 43.8±8% versus 69.1±13.6% for MLL-germline (p<0.0001).

    Intensified chemotherapy without SCT in infant ALL: results from COG P9407 (Cohort 3) · Dreyer ZE, Hilden JM, Jones TL, et al. · Pediatric Blood & Cancer 2015
    PMID 25399948DOI
    What this evidence cannot show

    Non-randomised cohort within a phase 3 trial; figures are standard errors, not 95% confidence intervals. Cohort 3 followed induction modifications made because of excessive early mortality in Cohorts 1 and 2, so it is not comparable to the earlier cohorts. This is an intensification study without a de-intensified randomised control.

    Evidence that argues against this (1)
    • COG P9407 Cohort 3 is a single-arm cross-protocol comparison, and the randomised test of the proposition it is used to support came out null. Interfant-06 randomly assigned 330 medium- and high-risk infants to myeloid-style post-induction courses (ADE + MAE, n=169) or the lymphoid-style course IB (n=161) and found 6-year DFS 39.3% versus 36.8% (log-rank P=.47), concluding that early intensification with post-induction myeloid-type chemotherapy did not significantly improve outcome and that Interfant-06 as a whole did not improve on Interfant-99. Intensity of consolidation is therefore not established as the lever; the P9407 age effect (5-year EFS 15.5% at 90 days or younger versus 48.5% above) is a much larger signal than anything intensification has moved. Pieters R et al. 2019
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In COG AALL0631, changing induction from a COG P9407-based 3-week regimen to an Interfant-99-based 5-week regimen with enhanced supportive care reduced induction mortality from 15.4% (4/26) to 1.6% (2/123) (p=0.009), reduced sterile-site infections from 57.7% to 19.5% (p=0.0002), and increased the end-induction-intensification complete response rate from 68% (17/25) to 94% (94/100) (p=0.0012).

    PMID 25407157DOI
    What this evidence cannot show

    Sequential non-randomised cohorts with small numbers in the earlier group (n=26). Two changes were made simultaneously — de-intensification and enhanced supportive care with mandated hospitalisation — so their contributions cannot be separated, and era effects also contribute.

    Evidence that argues against this (1)
    • This is a comparison of SEQUENTIAL historical cohorts (4/26 versus 2/123), so it confounds the induction regimen with calendar era, accrual experience and the concurrent supportive-care changes the same report describes; and a 15.4% rate resting on four deaths in 26 patients carries a very wide interval. That protocol choice may not be the operative variable is supported independently: a single-institution series of 116 infants treated 1990-2018 on five consecutive protocols found Interfant and BFM-based protocols gave comparable induction mortality (8.1% versus 7.1%, P=.852) and comparable complete-remission mortality (21.6% versus 28.6%, P=.438), with no significant reduction in relapse rate (48.5% versus 30.7%, P=.149), and KMT2A-rearrangement status the strongest prognostic factor independently of protocol strategy. Pennella CL et al. 2020
Connected to