Oncofusion

start / graph / concept-risk-group-cut-points-are-administrative-choices-infant-all

Explainerstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q7-clinical-infant

WBC 300, age 6 months, MRD 0.01%: three lines drawn by committees, not by biology

also called WBC 300 threshold · age 6 months cut-point · MRD 0.01% threshold · dichotomisation artefact infant ALL

Infant leukaemia trials sort babies into risk groups using round numbers - a white count of 300, an age of six months, a leukaemia level of 0.01% - and different trials use different numbers for the same purpose. Because each of those quantities is really a smooth scale, how big a risk difference appears depends on where somebody decided to put the line.

The same thing, for a scientist

Every risk stratification in infant ALL rests on dichotomising continuous variables. Interfant-06 defined high risk as KMT2A-rearranged **and** age **under** 6 months **and** either WBC >=300 x10^9/L or poor prednisone response [E-interfant06-age-direction]. (This entry previously said *older* than 6 months, which was backwards; see the correction section.) CCCG-ALL-2015 used age <6 months with KMT2A-r and WBC >=300, but its own multivariable analysis within KMT2A-rearranged patients found **age <3 months** to be the age factor associated with inferior EFS, and MRD >0.01% at day 19 was the only factor surviving adjustment [E2]. JPLSG MLL-10 stratified on KMT2A status, age and CNS leukaemia, and found MRD >=0.01% at end of early consolidation prognostic [E3]. COG AALL0631 defines an infant as under 366 days [E4]. Interfant-99 found age <6 months and 'very high WBC' independently adverse without a stated threshold [E5]. Five protocols, five different lines. Dichotomisation has three known consequences: it discards information, it makes the estimated effect size depend on the cut-point's position, and it creates a discontinuity in treatment allocation at a point where two patients are clinically indistinguishable - an infant at 299 x10^9/L and one at 301 x10^9/L receive different therapy.

Go deeper
The lines in use1 deeper
ProtocolAge lineWBC lineMRD lineOther
Interfant-99 56 months'very high', unstated-prednisone prophase response
Interfant-06 [E-interfant06-age-direction]< 6 months (conjunction, not a list)300 x10^9/L-prednisone response
CCCG-ALL-2015 26 months for stratification; 3 months in its own analysis300 x10^9/L0.01% at day 19day-5 dexamethasone response
JPLSG MLL-10 3used, value not in abstract-0.01% end of early consolidationCNS leukaemia present/absent
COG AALL0631 4366 days eligibility---
TCCSG 1977-1995 6-50,000/microlitre--

The TCCSG row is the clearest illustration: a threshold of 50 x10^9/L, six times lower than Interfant's, was also reported as prognostic. Both can be true, and both are statements about where a line was put.

Why this is not pedantry

Three practical consequences.

  1. Cross-trial comparison breaks. The proportion of infants who are 'high risk' differs between CCCG-ALL-2015 and Interfant-06 partly because 11.2% versus 29% exceed the WBC line, and partly because the age and MRD lines differ. Comparing high-risk EFS between them compares differently-constituted groups.
  2. A prognostic factor can vanish under adjustment. In CCCG-ALL-2015, age under 3 months was significant in univariate analysis within KMT2A-rearranged patients and did not survive multivariable adjustment, where only day-19 MRD >0.01% remained 2. The age effect was, on that analysis, a shadow of the MRD effect.
  3. The clinical discontinuity is real. Two infants either side of a line receive different consolidation and different transplant intent. No biology changes at 300.
What the graph should do with a threshold claim

When any entry in this graph states a risk difference across one of these lines, it should carry (a) the line's numerical value, (b) which protocol defined it, and (c) the fact that the effect size is conditional on the line. A statement like 'infants with WBC above 300 do worse' is a statement about Interfant-06's stratification and should not be read as a statement about the leucocyte count. Where a continuous or spline analysis exists, it should be preferred; this beat found none in infant ALL.

Correction: Interfant-06 high risk is age UNDER six months

Correction, 2026-09-20: the age criterion was recorded backwards, and it is UNDER six months. Two sources were fetched and read directly.

ClinicalTrials.gov NCT00550992, raw API, states the three groups explicitly, and the medium-risk line settles it on its own:

Medium-risk disease, defined by 1 of the following criteria: MLL status unknown; MLL rearranged AND age > 6 months; MLL rearranged AND age < 6 months AND WBC < 300 x 10^9/L AND prednisone good response

High-risk disease, defined by MLL rearrangement AND meets the following criteria: Age at diagnosis < 6 months (i.e., < 183 days); WBC >= 300 x 10^9/L AND/OR prednisone poor response

PMID 33405950, from the Interfant group itself, read in full text (PMC8196086 via efetch db=pmc, 22,863 characters after stripping): "Other high-risk factors on the Interfant-06 treatment protocol include age < 6 months at diagnosis and a high WBC count at diagnosis (>= 300 x 10^9/L) and/or a poor prednisone response."

Two further consequences.

First, the criterion is a conjunction, not a list. High risk requires KMT2A rearrangement and age under six months and (WBC >= 300 or poor prednisone response). An eight-month-old with a WBC of 400 is medium risk under this protocol, which the previous wording got wrong in the other direction too.

Second, the anomaly this graph recorded was never real. The entry asked why age appeared with opposite sign in Interfant-99 and Interfant-06 and left it open as a question about trial design. Both protocols treat the youngest infants as the highest risk and always did. The sign flip was ours. The JCO abstract is what both entries quoted, so this is a transcription or abstract-reading error and not an invention - but it stood through a beat that noticed it, flagged it, and then explained it rather than checking it, which is the more instructive part.

Facts and figures
metricstratification thresholds in use
populationinfants <1 year with ALL
thresholds
TCCSG1,999TCCSG: 1999COG AALL06312,015COG AALL0631: 2015Interfant-062,019Interfant-06: 2019JPLSG MLL-102,020JPLSG MLL-10: 2020CCCG-ALL-20152,024CCCG-ALL-2015: 2024
thresholds — the numbers behind the chart
[
  {
    "protocol": "Interfant-06",
    "wbc_x10_9_per_L": 300,
    "age_months": 6,
    "pmid": "31283407",
    "title": "Outcome of Infants Younger Than 1 Year With Acute Lymphoblastic Leukemia Treated With the Interfant-06 Protocol: Results From an International Phase III Randomized Study.",
    "authors": "Pieters R, De Lorenzo P, Ancliffe P, et al.",
    "venue": "J Clin Oncol",
    "year": 2019,
    "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."
  },
  {
    "protocol": "CCCG-ALL-2015",
    "wbc_x10_9_per_L": 300,
    "age_months_stratification": 6,
    "age_months_in_analysis": 3,
    "mrd_pct": 0.01,
    "mrd_day": 19,
    "pmid": "38634141",
    "title": "Outcome of infants with acute lymphoblastic leukemia treated with the Chinese Children's Cancer Group Acute Lymphoblastic Leukemia 2015 study protocol.",
    "authors": "Leung AWK, Cai J, Wan Z, et al.",
    "venue": "Haematologica",
    "year": 2024,
    "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."
  },
  {
    "protocol": "JPLSG MLL-10",
    "mrd_pct": 0.01,
    "mrd_timepoint": "end of early consolidation",
    "pmid": "32845001",
    "title": "A risk-stratified therapy for infants with acute lymphoblastic leukemia: a report from the JPLSG MLL-10 trial.",
    "authors": "Tomizawa D, Miyamura T, Imamura T, et al.",
    "venue": "Blood",
    "year": 2020,
    "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."
  },
  {
    "protocol": "COG AALL0631",
    "eligibility_age_days": 366,
    "pmid": "25407157",
    "title": "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.",
    "authors": "Salzer WL, Jones TL, Devidas M, et al.",
    "venue": "Pediatr Blood Cancer",
    "year": 2015,
    "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."
  },
  {
    "protocol": "TCCSG",
    "wbc_per_microlitre": 50000,
    "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."
  }
]
known consequences of dichotomisation
[
  "information loss",
  "effect size conditional on cut-point position",
  "treatment discontinuity between clinically indistinguishable patients",
  "non-comparability of high-risk groups across trials"
]
How this could bite us back
  • A targeted agent tested only in a protocol-defined high-risk group inherits that protocol's cut-points, so its apparent benefit will not transfer to a system using different lines.
  • Stratification that allocates transplant intent to a group in which only 46% reach transplant allocates an intention, not a treatment, and trial arms defined that way compare plans rather than exposures.
  • MRD thresholds set by assay limit of quantification move when the assay changes, so a risk group defined at 0.01% silently changes composition as laboratories adopt next-generation sequencing MRD.
Still open
  1. Has anyone fitted a continuous or spline model of WBC, age and MRD to pooled infant ALL data? This beat found none, and the Interfant, COG and JPLSG datasets together would support one.
  2. RESOLVED 2026-09-20, and the answer is that the question was built on our own error: Interfant-06 high risk requires age **under** 6 months, the same direction as Interfant-99. NCT00550992 raw eligibility and PMID 33405950 full text both state it. What remains genuinely open is why the Interfant-06 primary abstract reads the other way, which needs the full text (PMID 31283407, no PMCID, publisher returns 403).
  3. How many infants in Interfant-06 sat within 10% of the 300 x10^9/L line, and did their outcomes differ?
Every source behind this entry (7)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Interfant-06 defined high risk as KMT2A-rearranged and older than 6 months with WBC count 300 x10^9/L or more or a poor prednisone response, medium risk as all other KMT2A-rearranged cases, and low risk as KMT2A germline. **CORRECTED 2026-09-20: this reads 'older than 6 months' and the protocol says UNDER 6 months** - NCT00550992 raw eligibility and PMID 33405950 full text, both fetched. The criterion is also a conjunction: KMT2A-r AND age < 6 months AND (WBC >= 300 OR poor prednisone response).

    PMID 31283407DOI
    What this evidence cannot show

    The thresholds are protocol definitions reported in the abstract; the justification for choosing 300 rather than any other value is not given there and the full text is paywalled. **This sentence was wrong and is retained only to show what was corrected.** It read: the high-risk definition requires age OLDER than 6 months, while Interfant-99 found age YOUNGER than 6 months adverse, so the same variable appears with opposite sign in the two designs. **There is no opposite sign.** Interfant-06 requires age UNDER 6 months, exactly as Interfant-99 found - verified against NCT00550992 and PMID 33405950 on 2026-09-20.

    Evidence that argues against this (1)
    • The 6-year EFS of the intention-to-transplant high-risk group was 20.9% and only 46% of them actually reached transplant because of early events, so the stratification did not deliver the treatment it was designed to allocate. Pieters R et al. 2007
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    CCCG-ALL-2015 stratified on age <6 months, KMT2A-r and WBC >=300 x10^9/L, but its own multivariable analysis within KMT2A-rearranged patients retained only day-19 MRD >0.01% (p = 0.008), with age under 3 months significant only in univariate analysis.

    PMID 38634141DOI
    What this evidence cannot show

    71 KMT2A-rearranged patients, so the multivariable model is fitted on a small sample and a true age effect could be lost to power rather than to confounding. Read in the open-access full text at PMC11290514.

    Evidence that argues against this (1)
    • Failure of age to survive adjustment in 71 patients is at least as likely to be low power as to be genuine mediation by MRD; Interfant-99 with 482 patients found age independently adverse. Pieters R et al. 2007
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    JPLSG MLL-10 stratified infants by KMT2A status, age and presence of CNS leukaemia, and found MRD >=0.01% at the end of early consolidation, together with female sex, to be significant for poor prognosis in multivariable analysis.

    A risk-stratified therapy for infants with acute lymphoblastic leukemia: a report from the JPLSG MLL-10 trial. · Tomizawa D, Miyamura T, Imamura T, Watanabe T, Moriya Saito A, Ogawa A, et al. · Blood 2020
    PMID 32845001DOI
    What this evidence cannot show

    n = 90. The 0.01% MRD threshold is an assay-driven convention (the limit of reliable quantification of flow or PCR MRD), not a biological boundary, and is applied at a different timepoint than CCCG's day 19. Abstract read; full text paywalled.

    Evidence that argues against this (1)
    • MRD thresholds are set by assay sensitivity, so the same patient can be above the line on one platform and below it on another; a threshold that tracks the instrument is not a property of the disease. Leung AWK et al. 2024
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    COG AALL0631 defined its eligible population as infants under 366 days of age with newly diagnosed ALL.

    PMID 25407157DOI
    What this evidence cannot show

    An eligibility boundary. It differs in a day or two from 'under 12 months' and 'under 1 year' used elsewhere, which is immaterial clinically and material when comparing cohort compositions at the margin.

    Evidence that argues against this (1)
    • That the cooperative groups cannot agree on the definition of an infant is weak evidence that the category itself is administrative; KMT2A status, not the birthday, is what the biology tracks. Stutterheim J et al. 2025
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Interfant-99 found all types of MLL rearrangement, very high white blood cell count, age younger than 6 months and poor prednisone prophase response independently associated with inferior outcome in 482 infants.

    PMID 17658395DOI
    What this evidence cannot show

    'Very high white blood cell count' is reported without a numerical threshold in the abstract, so the effect size cannot be located on the scale. Full text paywalled.

    Evidence that argues against this (1)
    • Interfant-06, using explicit thresholds derived from this analysis, produced essentially identical survival, so the stratification did not convert into benefit. Pieters R et al. 2019
  6. E6weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    In the TCCSG infant cohort, hyperleukocytosis above 50,000 per microlitre was associated with poor prognosis.

    PMID 10530057
    What this evidence cannot show

    n = 62, retrospective, 1977-1995. A threshold six times lower than Interfant's, reported as prognostic in the same disease, which is the point being made rather than a competing estimate.

    Evidence that argues against this (1)
    • With a mean count of 205,900/microlitre in this cohort, a threshold of 50,000 puts most patients on one side of the line, so the comparison it defines is between a small low-count group and everyone else and is not the same contrast as Interfant's 300 x10^9/L. Leung AWK et al. 2024
  7. E-interfant06-age-directionestablished confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.registry

    Interfant-06 defines high risk as KMT2A rearrangement AND age at diagnosis under 6 months (under 183 days) AND either WBC at least 300 x10^9/L or a poor prednisone response. KMT2A-rearranged infants older than 6 months are explicitly medium risk, as are KMT2A-rearranged infants under 6 months with WBC below 300 and a good prednisone response.

    NCT00550992
    What this evidence cannot show

    Fetched from the ClinicalTrials.gov v2 API on 2026-09-20 and read verbatim, and independently confirmed in the full text of PMID 33405950, which is written by the Interfant group. Established as a fact about the protocol's stated definition. **The registry is not the protocol document**, and the Interfant-06 primary report (PMID 31283407) could not be opened - no PMCID and the publisher returns HTTP 403 - so the possibility that the published abstract's wording reflects a real amendment rather than an error is not formally excluded. Against that: the registry, the group's own 2021 full text and an independent full text (PMID 36861410) all agree on under 6 months.

    Evidence that argues against this (1)
    • The abstract of the Interfant-06 primary report is the source both entries quoted, and it is what produced the reversal - so a reader going back to the most authoritative-looking source for this protocol may find the wrong direction there too. That is the reason this correction records the registry and a full text rather than simply flipping the sentence. Journal of Clinical Oncology 2021
Connected to