External and historical controls: the design infant leukemia actually uses
also called historical control · external control arm · synthetic control
When there are too few babies to split into two groups, researchers compare the new treatment against records of similar babies treated earlier. That is how the recent blinatumomab result in infants was produced, and it is both the reason the study could happen and the reason its answer is not definitive.
The same thing, for a scientist
An externally controlled trial compares a single-arm cohort to patients treated outside the study, either historically or concurrently. In infant leukemia this is not a shortcut, it is the standing practice: the blinatumomab infant study compared 30 newly diagnosed infants to historical Interfant-06 controls, and the CCG-1953 and Interfant series form a deep, protocol-defined, MRD-annotated control resource that few rare diseases possess. FDA's 2023 draft guidance sets out when an external control is acceptable and where it fails, and the failure modes are precisely the ones this disease is most exposed to: era effects in supportive care, differences in MRD assay and stratification, and outcome ascertainment differences.
Why this disease is unusually well placed for it2 deeper
Most rare diseases lack a usable external control. Infant KMT2A-rearranged leukemia has one of the best in paediatric oncology.
- Interfant-99, Interfant-06 and Interfant-21 are sequential international protocols with defined eligibility, uniform therapy and published outcomes (
trial-interfant-99,trial-interfant-06,trial-interfant-21). - MRD is measured and standardised, and Interfant-06 published 6-year disease-free survival stratified by end-of-induction MRD at 60.2%, 45.0% and 33.8% for negative, intermediate and high 2. That gives an external control conditional on a prognostic covariate, which is far stronger than a marginal historical rate.
- Outcomes have been flat for decades 3, which weakens the usual objection that a historical control is unfairly pessimistic because care improved. When the control era's results have not moved, era bias is smaller.
The blinatumomab study used exactly this resource: 30 newly diagnosed infants, single arm, historical Interfant-06 comparison, primary endpoint toxicity, and better outcomes than the historical controls at a median follow-up of 26 months 3.
What FDA asks for, and what it will not accept
FDA's draft guidance on externally controlled trials sets expectations that map directly onto this setting 1: the external population must be comparable in disease definition, prognostic factors, standard of care and outcome ascertainment; the analysis must be pre-specified including the covariate adjustment; and the design is most defensible where the treatment effect is large, the outcome is objective and the natural history is well characterised and stable.
Infant KMT2A-rearranged leukemia satisfies the last three: the outcome (event-free survival, MRD clearance) is objective, the natural history is well characterised, and the effect sought is large. It is exposed on the first: diagnostic sensitivity, MRD assay, transplant policy and supportive care have all moved between Interfant-99 and today.
Practical consequence: the strongest available design in this disease is a single-arm study with a pre-specified, covariate-adjusted external control drawn from the immediately preceding Interfant protocol, not from the whole historical record, with MRD stratum as the adjustment variable.
The honest limitation
An external control cannot distinguish a drug effect from a selection effect. Infants enrolled in a new single-arm study are enrolled at centres willing and able to run it, by clinicians who judged them fit enough, in an era with better antifungals and better intensive care. Every one of those pushes the observed result in the favourable direction.
The blinatumomab result is a case in point: 30 patients, single arm, historical comparison, 26 months of follow-up. It is a genuine and important signal and it is not proof. The graph records it as such (epi-infant-leukemia-incidence), and any menin-inhibitor result obtained the same way should be recorded the same way.
The mitigation that is available and under-used is a concurrent external control: infants treated on the current Interfant protocol at non-participating centres during the same calendar period. That removes era bias entirely and leaves only centre and selection effects, which are at least measurable.
| design class | externally controlled single-arm |
|---|---|
| recommended adjustment covariate | end-of-induction MRD stratum |
| preferred improvement | concurrent rather than historical external control, drawn from contemporaneous Interfant-21 patients at non-participating centres |
available control resources
[ "trial-interfant-99", "trial-interfant-06", "trial-interfant-21", "trial-cog-aall0631", "trial-cog-aall15p1" ]
interfant 06 dfs by mrd PCT — the numbers behind the chart
{
"negative": 60.2,
"intermediate": 45,
"high": 33.8
}worked example
{
"study": "blinatumomab in infant ALL",
"n": 30,
"control": "historical Interfant-06",
"pmid": "37099340",
"title": "Blinatumomab Added to Chemotherapy in Infant Lymphoblastic Leukemia.",
"authors": "van der Sluis IM, de Lorenzo P, Kotecha RS, et al.",
"venue": "N Engl J Med",
"year": 2023,
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
}- Every successive single-arm study against historical controls raises the apparent bar for the next one, because the previous study's favourably-selected result becomes the new comparator. A field can ratchet itself into claiming improvements that never happened.
- If a menin inhibitor is approved in infants on an externally controlled basis, the randomised trial that would have settled the question becomes unethical to run, and the uncertainty is frozen permanently. That is the real cost of the design, and it should be decided deliberately rather than by default.
- Would the Interfant consortium release a covariate-annotated patient-level control dataset for use as a pre-specified external control, and under what governance?
- How much has outcome in the Interfant control population actually moved between the 06 and 21 eras, which is the direct measure of the era bias an external control would carry?
- Could a concurrent external control be assembled prospectively at non-participating centres, and would a regulator accept it in place of randomisation for a menin inhibitor in infants?
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
FDA guidance on externally controlled trials for drugs and biological products sets out the conditions under which an external control is interpretable, including comparability of the external population in disease definition, prognostic factors, standard of care and outcome ascertainment, and pre-specification of the analysis.
Considerations for the Design and Conduct of Externally Controlled Trials for Drug and Biological Products: Guidance for Industry · US Food and Drug Administration, Center for Drug Evaluation and Research · FDA guidance documents 2023What this evidence cannot show
Issued as draft guidance; guidance is non-binding and reflects current thinking rather than requirement. The URL was not re-fetched at the time of writing this node, so the document's current status (draft versus final) should be confirmed before citation in a regulatory submission.
Evidence that argues against this (1)
- FDA guidance is non-binding and states conditions — comparability in disease definition, prognostic factors, standard of care and outcome ascertainment — that historical infant-leukaemia cohorts meet poorly: Interfant-era controls differ in supportive care, transplant practice and MRD methodology from any contemporary arm. U.S. Food and Drug Administration 2026
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In Interfant-06, end-of-induction minimal residual disease stratified 6-year disease-free survival at 60.2% (negative), 45.0% (intermediate) and 33.8% (high), providing a covariate-conditional external control resource.
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. · Journal of Clinical Oncology 2021What this evidence cannot show
MRD was measurable in a subset of enrolled infants, so the stratification excludes patients without a trackable marker; DFS is not overall survival; assay methods have since evolved, which is itself a source of external-control incomparability.
Evidence that argues against this (1)
- MRD strata are an observational gradient, so using them as a covariate-conditional external control imports the confounding as well as the covariate: high-MRD patients differ in disease biology before any treatment decision, and the thresholds separating the three strata are chosen cut-points on a continuous measurement. Statistics in Medicine 2006
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
A single-arm study of 30 newly diagnosed infants given one course of blinatumomab added to Interfant-06 chemotherapy met its primary endpoint with no protocol-defined toxic effects and showed better outcomes than historical Interfant-06 controls at a median follow-up of 26 months, in a disease whose outcomes had not improved in recent decades.
Blinatumomab Added to Chemotherapy in Infant Lymphoblastic Leukemia · van der Sluis IM, de Lorenzo P, Kotecha RS, et al. · New England Journal of Medicine 2023What this evidence cannot show
Thirty patients, single arm, historical rather than randomised comparator, median follow-up 26 months. The survival benefit is suggestive, not established, and the comparison inherits every era and selection effect described in this node. **Regraded strong -> established on 2026-09-20** under `concept-the-confidence-grading-rule`: this is the trial's own reported result, which is established as a fact about what the trial found.
Evidence that argues against this (1)
- The strongest available check on this particular external-control comparison has now been run and it held. With median follow-up extended to 4.2 years, the same 30 KMT2A-rearranged infants had 4-year disease-free survival of 83.3% versus 44.0% and overall survival of 93.3% versus 60.2% against the selected Interfant-06 historical cohort. Durability at four years is much harder to generate by favourable-patient selection than a 26-month readout, so this is evidence that the design's central weakness did not, in this instance, produce a false positive. It remains a comparison against a SELECTED historical cohort, and the randomised Interfant-21 is still the test that matters. Vieira Martins M et al. 2026
Cohort disagreement (Biondi 2018, PMID 30501871): EsPhALL2010 with continuous imatinib and only 38% transplanted reached 5-year EFS 57.0% and OS 71.8%, below the 70% 5-year DFS quoted, with 61% of deaths occurring in continuous remission.
- ← related toPML::RARA, t(15;17)(q24;q21)
Cohort disagreement (Park 2011, PMID 21653939): SEER's 1400 unselected APL patients gave a 17.3% early-death rate, not 29%, so registry early-death figures are ascertainment-dependent and bound rather than fix the number.
Counter-evidence (NCT06317662, read 2026-09-18): the NCI opened a RANDOMIZED frontline infant ALL trial in June 2025, 153 infants under 365 days, with the randomisation sited inside the KMT2A-rearranged stratum. The node's conclusion that a randomised frontline infant trial 'cannot accrue and a reviewer will know' is empirically false as of 2025, so the single-arm-plus-historical-control route is a choice, not a necessity.
why single-arm external-control designs dominate
- depends on →Interfant-06 — 651 infants, and the definitive answer that myeloid-style consolidation does not help
the external control resource
- related to →Interfant-21 (NCT05327894) — the randomised test of blinatumomab in infant KMT2A-rearranged ALL, open now
the contemporaneous cohort a concurrent external control would draw on
- supports →Blinatumomab in infant KMT2A-rearranged ALL — the largest single outcome gain in thirty years
worked example: 30 infants against historical Interfant-06 controls