CNS disease at diagnosis in infants: a real excess, measured by a procedure that partly creates it
also called CNS involvement infant ALL · CNS2 CNS3 infant · traumatic lumbar puncture infant · CNS staging artefact
About 14% of infants have leukaemia in the fluid around the brain at diagnosis, against 2-5% of older children - a real and large excess. But how many are found depends on when the spinal tap is done: one trial delayed it by five days and after steroid treatment found fewer cases, without any change in the disease.
The same thing, for a scientist
CNS involvement at diagnosis was reported in 13.8% of Interfant-06 patients against 2-5% in non-infant paediatric ALL [E1]. The excess is real and is concentrated in KMT2A::AFF1 pro-B disease [E1]. Three things nonetheless make the number partly procedural. **First, timing.** CCCG-ALL-2015 deferred the first lumbar puncture to day 5 of induction, after four days of oral dexamethasone, and reported a lower CNS involvement rate as a consequence, with a later isolated CNS relapse rate of 2.2% against 11.9% in Interfant-06 [E2]. A prophase that has already been given changes what the tap finds. **Second, whether a tap is done at all.** In the TCCSG infant cohort only 36 of 62 infants underwent lumbar puncture; CNS involvement was evident in 9 of those 36 (25%), and 3 of the 9 had no neurological symptoms [E3]. A denominator of the tapped rather than of the enrolled inflates the rate. **Third, blood contamination.** A traumatic tap in an infant with a leucocyte count in the hundreds of thousands introduces circulating blasts into the specimen; CCCG report an approximately 10% traumatic-tap rate [E2]. CNS2/CNS3 classification is itself a set of administrative cell-count thresholds, not a biological boundary.
The comparison, laid out1 deeper
| Cohort | CNS involvement at diagnosis | How and when ascertained |
|---|---|---|
| Interfant-06 | 13.8% 1 | at diagnosis, per protocol |
| Non-infant paediatric ALL | 2-5% 1 | at diagnosis |
| TCCSG infants 1977-1995 | 9 of 36 tapped (25%); 9 of 62 enrolled (14.5%) 3 | LP performed in only 58% of the cohort |
| Single US hospital, infants <6 months | 6 of 9 (66.6%) 4 | n = 9 |
| CCCG-ALL-2015 | described as lower; a numeric rate is not given in the letter 2 | LP deferred to day 5, after 4 days of dexamethasone |
Note how far apart the numbers are: 13.8%, 14.5%, 25% and 66.6% are all 'CNS involvement at diagnosis in infants'. The spread is dominated by who was tapped and when, not by the populations differing fourfold in biology.
The CCCG natural experiment
CCCG-ALL-2015 gave four days of dexamethasone before the first lumbar puncture, an intervention introduced for safety in an infant with a high blast count. Their isolated CNS relapse rate was 2.2% and combined marrow-plus-CNS relapse 0.7%, against 11.9% isolated CNS relapse in Interfant-06, and the authors state that performing the LP after the blast count has fallen might be a safer approach and that dexamethasone prophase may protect against CNS relapse 2.
Two readings are available and the data cannot separate them. Either the prophase treats occult CNS disease and genuinely lowers relapse, or it hides CNS disease at staging so that patients are under-staged into lower risk groups and the denominator for 'CNS relapse' shifts. The second would be a staging artefact that looks like a therapeutic win. CCCG's cohort also under-enrolled high-risk infants on their own account 2, which pushes the same direction.
Cut-points again
CNS1/CNS2/CNS3 are defined by cells per microlitre of cerebrospinal fluid and the presence of blasts, with thresholds chosen for operational convenience. Dichotomising a continuous quantity - CSF blast burden - makes the apparent size of any effect depend on where the line was drawn, and the line differs between protocols and eras. A claim that infants have four times the CNS involvement of older children is a claim about two populations measured with the same ruler; a claim that CCCG has less CNS disease than Interfant is a claim about two different rulers.
The relapse side, which is not artefactual
Whatever the staging noise at diagnosis, the relapse pattern is robust: in Interfant-06, 213 of 442 patients in first complete remission relapsed, and 50 of those relapses (24% of relapses; 11% of the 442) involved the CNS - 11% isolated CNS and 13% combined 1. The proportion of relapses involving CNS is not higher than in non-infant ALL, but the absolute number is, because the relapse rate itself is much higher 1. Blinatumomab does not cross into the CNS, and in the Interfant blinatumomab pilot all four relapses involved the CNS despite no CNS involvement at diagnosis 15. That is the single most operationally important consequence of this section.
| metric | CNS involvement at diagnosis |
|---|---|
| population | infants <1 year with ALL |
cohorts — the numbers behind the chart
[
{
"cohort": "Interfant-06",
"pct": 13.8,
"pmid": "40207717",
"note": "as reported in the Interfant investigators' 2025 review",
"title": "KMT2A-rearranged acute lymphoblastic leukemia in infants: current progress and challenges.",
"authors": "Stutterheim J, Van der Sluis IM, Vrenken KS, et al.",
"venue": "Haematologica",
"year": 2025,
"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."
},
{
"cohort": "non-infant paediatric ALL",
"pct_range": [
2,
5
],
"pmid": "40207717",
"title": "KMT2A-rearranged acute lymphoblastic leukemia in infants: current progress and challenges.",
"authors": "Stutterheim J, Van der Sluis IM, Vrenken KS, et al.",
"venue": "Haematologica",
"year": 2025,
"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."
},
{
"cohort": "TCCSG",
"positive": 9,
"tapped": 36,
"enrolled": 62,
"pmid": "10530057",
"title": "Clinical and biological aspects of acute lymphoblastic leukemia in 62 infants: retrospective analysis of the Tokyo Children's Cancer Study Group.",
"authors": "Isoyama K, Okawa H, Hayashi Y, et al.",
"venue": "Pediatr Int",
"year": 1999,
"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."
},
{
"cohort": "single US hospital <6 months",
"positive": 6,
"n": 9,
"pmid": "18800262",
"title": "Infant acute lymphoblastic leukemia: a 20-year children's hospital experience.",
"authors": "Murray RA, Thom G, Gardner RV, et al.",
"venue": "Fetal Pediatr Pathol",
"year": 2008,
"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."
}
]relapse
{
"cohort": "Interfant-06",
"in_cr1": 442,
"relapses": 213,
"cns_involved": 50,
"isolated_cns_pct": 11,
"combined_pct": 13,
"pmid": "40207717",
"title": "KMT2A-rearranged acute lymphoblastic leukemia in infants: current progress and challenges.",
"authors": "Stutterheim J, Van der Sluis IM, Vrenken KS, et al.",
"venue": "Haematologica",
"year": 2025,
"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."
}artefact sources
[ "timing of first LP relative to steroid prophase", "proportion of cohort actually tapped", "traumatic tap with circulating blasts", "CNS1/2/3 cell-count thresholds are administrative" ]
- A CNS-inactive immunotherapy can convert a marrow-relapse problem into a CNS-relapse problem without changing overall relapse-free survival, which would make the regimen look better on marrow endpoints than it is.
- Steroid prophase before staging can move patients into lower risk groups without treating them, so a protocol change that improves apparent stratification can worsen real under-treatment.
- Omitting high-dose methotrexate in favour of blinatumomab removes the most CNS-penetrant agent in the backbone at the same time as adding one that does not penetrate at all.
- Intensifying intrathecal therapy to cover the CNS trades against neurocognitive late effects in a brain that is still myelinating, and infants are the group least able to absorb that cost.
- Does a steroid prophase before the first lumbar puncture treat occult CNS disease or merely under-stage it? A trial randomising day-1 against day-5 tap with paired cytospin and flow cytometry would separate them.
- What is the traumatic-tap rate in infants across Interfant centres, and how many CNS2/CNS3 assignments in infants are attributable to blood contamination at counts above 300 x10^9/L?
- If blinatumomab and CAR T-cell therapy displace marrow relapse, does the absolute CNS relapse rate rise, stay flat or fall? The pilot's four events cannot answer this.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.review
CNS involvement was reported in 13.8% of infants at diagnosis in Interfant-06 against 2-5% in non-infant paediatric ALL; of 213 relapses among 442 patients in first complete remission, 50 (24%) involved the CNS, and in the Interfant blinatumomab pilot all four relapses involved the CNS despite no CNS involvement at diagnosis.
KMT2A-rearranged acute lymphoblastic leukemia in infants: current progress and challenges. · Stutterheim J, Van der Sluis IM, Vrenken KS, Pieters R, Stam RW · Haematologica 2025What this evidence cannot show
A narrative review by the Interfant investigators summarising their own trial, so it is a secondary report of the Interfant-06 numbers; the primary paper's full text is paywalled and its baseline table was not read. The blinatumomab pilot observation rests on four events. Read in the open-access full text at PMC12399962.
Evidence that argues against this (1)
- The same review notes that the proportion of relapses involving the CNS in infants (24%) is not higher than in non-infant ALL (28% in AALL1331, 20% in DCOG ALL11), so the infant CNS problem is a consequence of a higher overall relapse rate rather than of a special CNS tropism. Pieters R et al. 2019
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
CCCG-ALL-2015 deferred the first lumbar puncture to day 5 of induction after four days of dexamethasone and reported lower CNS involvement, an isolated CNS relapse rate of 2.2% and combined marrow-CNS relapse of 0.7% against 11.9% isolated CNS relapse in Interfant-06, with an approximately 10% traumatic lumbar puncture rate.
Outcome of infants with acute lymphoblastic leukemia treated with the Chinese Children's Cancer Group Acute Lymphoblastic Leukemia 2015 study protocol. · Leung AWK, Cai J, Wan Z, Qin J, Fang Y, Sun L, et al. · Haematologica 2024What this evidence cannot show
A non-randomised between-trial comparison across countries and eras. The cohort under-enrolled high-risk infants by the authors' own account (18.4% under six months against 50-68% elsewhere; KMT2A-r 53.0% against 74-83%), which independently predicts less CNS disease. The letter gives no numeric CNS involvement rate at diagnosis, only the statement that it was lower. Read in the open-access full text at PMC11290514.
Evidence that argues against this (1)
- The lower relapse figure is equally compatible with under-staging: a prophase given before the diagnostic tap can clear CSF blasts, moving patients out of the CNS-positive category without altering their disease, which would lower the apparent CNS relapse rate by changing the classification rather than the outcome. Stutterheim J et al. 2025
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In 62 infants with ALL in the TCCSG cohort, CNS involvement was evident in 9 of the 36 infants who underwent lumbar puncture, and 3 of those 9 had no neurological symptoms at diagnosis.
Clinical and biological aspects of acute lymphoblastic leukemia in 62 infants: retrospective analysis of the Tokyo Children's Cancer Study Group. · Isoyama K, Okawa H, Hayashi Y, Hanada R, Okimoto Y, Maeda M, et al. · Pediatr Int 1999What this evidence cannot show
Only 58% of the cohort was tapped and the paper does not say how the 36 were selected; if symptomatic infants were preferentially tapped the 25% is badly inflated. 1977-1995, before modern CNS classification. Abstract read; full text not retrieved.
Evidence that argues against this (1)
- That a third of the CNS-positive infants had no neurological symptoms cuts the other way and argues the tapped-only denominator may under-count as well as over-count, since asymptomatic disease would be missed in the 26 who were not tapped. Stutterheim J et al. 2025
- E4weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper
In nine infants under 6 months with ALL at a single US hospital, 6 of 9 (66.6%) had central nervous system disease.
Infant acute lymphoblastic leukemia: a 20-year children's hospital experience. · Murray RA, Thom G, Gardner RV, Craver RD · Fetal Pediatr Pathol 2008What this evidence cannot show
n = 9; the 95% confidence interval on 6/9 spans roughly 30% to 93%. Single institution, pathology review, definition of CNS disease not stated.
Evidence that argues against this (1)
- This figure is nearly five times Interfant-06's 13.8% on nine patients; it should be read as consistent with the trial figure rather than as contradicting it, and is included to show how unstable small-series CNS rates are. Stutterheim J et al. 2025
- E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Adding one post-induction course of blinatumomab to the Interfant-06 backbone raised 2-year disease-free survival from 49% to 82% and overall survival from 66% to 93%.
Blinatumomab Added to Chemotherapy in Infant Lymphoblastic Leukemia. · van der Sluis IM, de Lorenzo P, Kotecha RS, Attarbaschi A, Escherich G, Nysom K, et al. · N Engl J Med 2023What this evidence cannot show
A single-arm pilot of 30 infants compared with a historical Interfant-06 control, not a randomised comparison; the historical control differs in era and in supportive care. The CNS-relapse observation that follows from it rests on four relapse events. **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)
- Because blinatumomab does not cross into the CNS, a regimen that suppresses marrow relapse can shift the relapse burden into the CNS rather than reduce it, which is what all four pilot relapses involving the CNS suggests. Stutterheim J et al. 2025
infant-specific rate and the procedural sources of its variance
- related to →Interfant-06 — 651 infants, and the definitive answer that myeloid-style consolidation does not help
13.8% CNS involvement at diagnosis; 50 of 213 relapses involved CNS
- related to →Blinatumomab in infant KMT2A-rearranged ALL — the largest single outcome gain in thirty years
CNS-inactive agent may displace rather than remove relapse
CNS disease at diagnosis in infant ALL
33.6% CNS-positive against Interfant-06's 13.8% in the same disease
- ← related toWhen infant leukaemia comes back, it comes back early and while treatment is still running
50 of 213 relapses involved the CNS
- ← supportsHypothesis: the spread in infant presenting-feature frequencies is mostly measurement, not biology
the CNS case is the hypothesis's strongest instance