COG AALL0631 — lestaurtinib in infant KMT2A-rearranged ALL: negative overall, positive where the drug actually hit
also called AALL0631 · NCT00557193 · CEP-701 infant ALL trial
This trial added a FLT3-blocking drug to chemotherapy for babies with KMT2A-rearranged leukemia. Overall it did not help, but the babies in whom the drug measurably switched FLT3 off did far better, which suggests the idea was right and the drug was too weak.
The same thing, for a scientist
AALL0631 randomised highest-risk infants with KMT2A-rearranged ALL to intensive chemotherapy with or without the FLT3 inhibitor lestaurtinib (CEP-701). Three-year EFS was 36±6% with lestaurtinib (n=67) versus 39±7% without (n=54), p=0.67 — negative on the primary comparison. But in correlative analyses, infants in whom potent FLT3 inhibition was pharmacodynamically demonstrated had EFS 59±10% versus 28±7% in those without (p=0.009), and infants whose blasts were FLT3-inhibitor-sensitive ex vivo had EFS 52±8% versus 5±5% if resistant (p<0.001). The 17 patients who were both PD-inhibited and ex-vivo sensitive had EFS 88±8%. The trial also demonstrated, in a separate analysis, that the Interfant-99-style induction was far safer than the COG P9407-style induction (1.6% vs 15.4% induction mortality).
The result, and the honest way to read it1 deeper
| Comparison | EFS | p |
|---|---|---|
| Lestaurtinib + chemo (n=67) vs chemo alone (n=54), 3-year | 36±6% vs 39±7% | 0.67 |
| PD-inhibited vs not (correlative) | 59±10% vs 28±7% | 0.009 |
| Ex-vivo FLT3-sensitive vs resistant (correlative) | 52±8% vs 5±5% | <0.001 |
| Both PD-inhibited AND ex-vivo sensitive (n=17) | 88±8% | — |
All values are ±standard error, as reported 1.
The trap. The correlative findings are exploratory, post-hoc and confounded in a specific and serious way: achieving potent pharmacodynamic FLT3 inhibition requires adequate drug exposure, and adequate drug exposure correlates with organ function, absence of early toxicity, absence of intercurrent infection and ability to stay on protocol — all of which independently predict survival. A patient who is well enough to be PD-inhibited is a patient who is well enough to survive. The 88±8% figure in 17 hand-picked patients is the most exciting and least trustworthy number in the paper.
Why it is still worth taking seriously. The ex-vivo sensitivity split is less confounded, because it is a property of the leukemia measured before treatment rather than a property of the patient measured during it. A 52% versus 5% EFS gap between ex-vivo sensitive and resistant blasts is a biomarker signal, not just a fitness signal. It says the FLT3 dependency in KMT2A-rearranged infant ALL is real but heterogeneous, and lestaurtinib — a weak, poorly-exposed, first-generation multikinase inhibitor — was the wrong tool.
Why FLT3 in KMT2A-rearranged ALL at all
Unlike AML, where FLT3 is activated by ITD or TKD mutation, KMT2A-rearranged infant ALL characteristically overexpresses wild-type FLT3 to extreme levels as part of the fusion's transcriptional program. That is a different pharmacological problem: you are not inhibiting a constitutively active mutant kinase with a low-nanomolar dependency, you are trying to suppress an overabundant wild-type receptor whose signalling can be restored by ligand. Modern FLT3 inhibitors (gilteritinib, quizartinib) were developed against mutant FLT3 and their activity against wild-type overexpression in ALL has never been tested in a randomised paediatric trial.
The other, quieter result: induction backbone matters more than the targeted agent
AALL0631 amended its induction partway through because of excessive early mortality. Comparing the two cohorts, induction mortality fell from 15.4% (4/26) on the COG P9407-style 3-week induction to 1.6% (2/123) on the Interfant-99-style 5-week induction with enhanced supportive care (p=0.009), sterile-site infections fell from 57.7% to 19.5% (p=0.0002), and the complete response rate at end of induction intensification rose from 68% to 94% (p=0.0012) 2. All induction deaths but one were infection-related. A near-tenfold reduction in early death, with a higher CR rate, dwarfs anything the randomised targeted agent did — and it came from giving less chemotherapy and better supportive care, not more.
This is a recurring lesson in infant leukemia and the lab should hold it: in a population where a meaningful fraction of children die of treatment rather than of disease, reducing induction mortality is a therapeutic intervention with a larger effect size than most drugs. It is also the strongest argument for adding a non-myelosuppressive agent (menin inhibitor, T-cell engager) rather than more chemotherapy.
| nct | NCT00557193 |
|---|---|
| phase | Phase 3 |
| status | Completed |
| n | 218 |
| n randomised | 121 |
| population | Newly diagnosed infants with KMT2A-rearranged ALL, highest-risk stratum |
| pediatric eligible | true |
| open now | false |
| readout | Negative for unselected FLT3 inhibition; strong ex-vivo sensitivity biomarker signal; induction choice had a larger effect than the targeted agent |
arms
[
{
"name": "Chemotherapy + lestaurtinib",
"n": 67,
"efs_3y_pct": 36,
"se": 6
},
{
"name": "Chemotherapy alone",
"n": 54,
"efs_3y_pct": 39,
"se": 7
}
]primary result
{
"randomised_p": 0.67,
"verdict": "negative"
}correlative results — the numbers behind the chart
{
"pd_inhibited_efs_pct": 59,
"pd_inhibited_se": 10,
"pd_uninhibited_efs_pct": 28,
"pd_uninhibited_se": 7,
"pd_p": 0.009,
"exvivo_sensitive_efs_pct": 52,
"exvivo_sensitive_se": 8,
"exvivo_resistant_efs_pct": 5,
"exvivo_resistant_se": 5,
"exvivo_p": 0.001,
"both_n": 17,
"both_efs_pct": 88,
"both_se": 8
}induction mortality — the numbers behind the chart
{
"p9407_style_pct": 15.4,
"p9407_n": 26,
"interfant99_style_pct": 1.6,
"interfant99_n": 123,
"p": 0.009
}- Essential-function toxicity: FLT3 is required for normal lymphoid and dendritic progenitor development, so sustained FLT3 inhibition in an infant targets a developmental pathway, not just the leukemia.
- Resistance: FLT3 in KMT2A-rearranged ALL is overexpressed wild-type receptor, which can be re-activated by ligand; ligand-mediated rescue is a well-described escape route for FLT3 inhibitors and would be expected here.
- Selection artefact as a hazard: the PD-inhibited subgroup result is the kind of finding that launches an underpowered confirmatory trial. Any successor must randomise within the ex-vivo-sensitive stratum, not select on achieved exposure.
- Delivery: lestaurtinib's exposure problem in infants was the proximate cause of failure; oral bioavailability and exposure variability in babies is a recurring, under-modelled barrier for every oral targeted agent including menin inhibitors.
- Would gilteritinib or quizartinib, which achieve real target coverage, reproduce the ex-vivo-sensitive subgroup benefit? Nobody has run this in infant ALL, and the ex-vivo assay that identified responders already exists.
- Is the ex-vivo FLT3 sensitivity signal a FLT3 dependency or a general drug-sensitivity phenotype? Testing ex-vivo sensitivity to an unrelated agent in the same samples would separate the two and has not been done.
- Does FLT3 overexpression in KMT2A-rearranged ALL predict menin-inhibitor response or resistance? FLT3-ITD is associated with menin-inhibitor failure in adult AML, and the wild-type-overexpression case in infants is unexamined.
- Given a near-tenfold difference in induction mortality between two backbones, how much of the historical 'infant ALL is refractory' narrative is actually early treatment-related death rather than resistant disease?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In COG AALL0631, adding lestaurtinib to chemotherapy in newly diagnosed KMT2A-rearranged infant ALL gave 3-year EFS 36±6% (n=67) versus 39±7% (n=54) without, p=0.67; correlative analyses showed EFS 59±10% versus 28±7% (p=0.009) for pharmacodynamically FLT3-inhibited patients and 52±8% versus 5±5% (p<0.001) for ex-vivo FLT3-inhibitor-sensitive versus resistant blasts, with 88±8% EFS in the 17 patients who were both.
FLT3 inhibitor lestaurtinib plus chemotherapy for newly diagnosed KMT2A-rearranged infant acute lymphoblastic leukemia: Children's Oncology Group trial AALL0631 · Brown PA, Kairalla JA, Hilden JM, et al. · Leukemia 2021What this evidence cannot show
The primary randomised comparison is negative. All subgroup findings are correlative and post-hoc, and the pharmacodynamic-inhibition subgroup is confounded by patient fitness: the ability to achieve target drug exposure is itself a survival predictor independent of FLT3 biology. Values are standard errors, not 95% confidence intervals. A published erratum exists (PMID 33846544, DOI 10.1038/s41375-021-01245-x). Lestaurtinib is a weak, poorly bioavailable multikinase inhibitor and these results do not generalise to gilteritinib or quizartinib.
Evidence that argues against this (1)
- The correlative rescue - 'the drug worked in patients where FLT3 was actually inhibited' - has been run before in adults and did not survive. In the UK AML15/AML17 randomisation of lestaurtinib added to first-line chemotherapy in 500 FLT3-mutated AML patients there was no difference in 5-year overall survival (46% vs 45%, HR 0.90, p=0.3), yet correlative analysis again showed that patients achieving sustained greater-than-85% FLT3 inhibition had significantly better outcomes. The same post-hoc signal appearing in a second, larger, independent negative trial is the argument that it reflects drug-exposure-as-fitness-marker rather than FLT3 biology, exactly the confound this entry's caveats raise. Knapper S et al. 2017
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Within AALL0631, de-intensifying induction from a COG P9407-based 3-week regimen (Cohort 1) to an Interfant-99-based 5-week regimen with enhanced supportive care (Cohort 2) reduced induction mortality from 15.4% (4/26) to 1.6% (2/123), p=0.009, and sterile-site infections from 57.7% to 19.5% (p=0.0002), while raising the end-induction-intensification complete response rate from 68% (17/25) to 94% (94/100), p=0.0012.
Decreased induction morbidity and mortality following modification to induction therapy in infants with acute lymphoblastic leukemia enrolled on AALL0631: a report from the Children's Oncology Group · Salzer WL, Jones TL, Devidas M, et al. · Pediatric Blood & Cancer 2015What this evidence cannot show
A comparison of sequential, non-randomised induction cohorts within one trial, with small numbers in the earlier cohort (n=26). Two changes were made at once — de-intensification AND enhanced supportive care with mandated hospitalisation — so their contributions cannot be separated. Era effects in supportive care also contribute. All induction deaths but one were infection-related, which is consistent with the interpretation but does not prove it.
Evidence that argues against this (1)
- Reading this as 'de-intensified induction is the safe choice' is contested by the JPLSG MLL-10 experience, which went the other way - adding high-dose cytarabine to early intensification for KMT2A-rearranged infants (explicitly modelled on AALL0631) - and reported 3-year EFS of 66.2% in KMT2A-rearranged infants without an offsetting mortality signal. That is, the AALL0631 induction modification reduced early death but the trial's own successor evidence suggests the survival ceiling is set by what happens after induction, not by how gentle induction is. Tomizawa D et al. 2020
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.trial_registry
AALL0631 is registered as NCT00557193, a Children's Oncology Group phase 3 trial with enrolment 218 that opened in January 2008 and randomised highest-risk infants to intensive chemotherapy with or without FLT3 inhibition with lestaurtinib (CEP-701).
Randomization of Highest Risk Infants to Intensive Chemotherapy +/- FLT3 Inhibition (CEP-701, Lestaurtinib) · Children's Oncology Group · ClinicalTrials.gov 2021What this evidence cannot show
Registry enrolment (218) covers the whole trial including non-randomised strata; the randomised comparison involved 121 patients.
Evidence that argues against this (1)
- The registered primary outcome is far narrower than the trial's description suggests, and this matters for what the record can be cited for. Raw ClinicalTrials.gov v2 API, retrieved 2026-09-19: the SOLE primary outcome of NCT00557193 is 'Percent Probability for Event-free Survival (EFS) for Patients on Arm C at Dose Level 2 (DL2)', assessed from the start of post-induction therapy for up to 10 years — a single-arm estimate in one arm at one dose level, not a randomised comparison of chemotherapy with versus without lestaurtinib. The record carries ACTUAL enrolment 218, primary completion 2017-09-30, completion 2024-06-30, and results posted since 2018-09-26. So the registry does establish that results exist, but it does not establish that the FLT3-inhibition randomisation was the prespecified primary question, and the entry should not lean on the registry for that. Children's Oncology Group 2008
randomised lestaurtinib trial; negative overall but positive in the pharmacodynamically inhibited subgroup
a negative trial, and a lesson about targeting overexpression rather than mutation
- ← related toWhich MRD timepoint drives which decision in infant ALL, and how good the evidence for each is
flow MRD at three timepoints, prognostic in KMT2A-rearranged infants only
COG uses an age cut of 90 days with a published biological argument, unlike the WBC line
lestaurtinib + chemotherapy; 3-year EFS 36% vs 39%, p=0.67
FLT3 inhibition in KMT2A-rearranged infant ALL; negative overall, positive in ex-vivo-sensitive blasts
AALL0631 established that unselected FLT3 inhibition does not help in this genotype