Latency: months from the fusion forming in a fetus to leukaemia in an infant, against years or a decade for every other backtracked fusion

Small series, and the detection requirement biases toward cases with a large clone at birth, i.e. toward short latency. This is a selected, not a representative, latency distribution.

• Latency measured from backtrack-positive cases is measured on a biased subset, and the bias runs toward short latency. A Guthrie punch is a few microlitres, so only newborns whose clone was already above the assay threshold produce a positive backtrack; those are the larger birth clones, which are plausibly the ones that progress fastest. Burjanivova et al. quantify how selective this is: running the identical patient-specific approach with an assay calibrated to detect 1-3 positive cells in the spot, they recovered the clonotypic marker in 3 of 12 non-infant ALL cases and 0 of 13 AML cases, including 0 of the 3 whose marker was an MLL fusion. The published latency distribution is therefore conditioned on detectability at birth and cannot be read as the latency distribution of fusion-positive newborns. Read: abstract only. Burjanivova T et al. 2006

A single twin pair with a rare partner, t(1;11)(p32;q23), not KMT2A::AFF1; extreme presentations are more likely to be published.

• Seven weeks is not the floor, and a case series rather than a twin pair shows it. A clinical and molecular study of 77 NEONATAL leukaemias alongside 380 infant leukaemias in Brazil found KMT2A rearrangements to be the frequent abnormality in the non-Down-syndrome neonatal cases - mainly KMT2A-AFF1 in ALL and KMT2A-MLLT3 in AML - i.e. disease presenting in the first days to weeks of life. The lower bound on latency is therefore effectively zero, the distribution of latencies is what needs describing, and a single published twin pair is a point on that distribution selected for being early. Moura SV et al. 2015

n=1; the co-twin's three years of disease-free follow-up sets a lower bound on the alternative outcome, not an upper bound.

• Base-rate context for a single discordant pair. In a population-based California registry linkage covering 255 twin pairs with leukaemia, AML diagnosed under one year of age was 25% concordant on raw counts and 100% on the monozygotic estimate, while lymphoid leukaemia at ages 1-10 was 0.7%. Infant concordance is thus the high-concordance extreme of the disease, which makes a documented non-concordant infant pair a genuine outlier - and outliers are what get published. The pair bounds what CAN happen; it does not bound how often the carrying co-twin escapes. Nickels EM et al. 2025

RUNX1::RUNX1T1 AML, not a KMT2A fusion; 2 of 10 patients were over 10 years old at diagnosis.

• A disagreeing cohort, using the same backtracking method with a stated detection limit. Applying one identical approach to Guthrie cards from 12 non-infant childhood ALL and 13 childhood AML patients, at a sensitivity requiring only 1-3 positive cells in the blood spot, markers were found in 3 of 12 ALL cases and in NONE of the 13 AML cases - including cases with RUNX1/ETO, PML/RARA and CBFB/MYH11 markers. The authors conclude that prenatal origin in AML is either less frequent or accompanied by a much lower birth clone load. So the >10-year latency inference in t(8;21) AML rests on the minority of AML cases in which a prenatal clone is detectable at all. Burjanivova T et al. 2006

Deliberately selected rare cases diagnosed outside the normal age range; establishes that long latency is possible for KMT2A, not how often.

• Short latency is not the same as no second event; it is compatible with a fast second event, and the human experiments say a second event is needed. Montes et al. expressed MLL-AF4 in human cord-blood CD34+ HSPCs and obtained enhanced repopulating function and clonogenic potential but no immortalisation and no leukaemia in transplanted animals over 16 weeks, and adding FLT3 tyrosine-kinase-domain mutation or FLT3 overexpression did not rescue it; their conclusion was that additional oncogenic lesions are required. The inference this entry endorses -- that a short latency plus near-total twin concordance plus 1.3 non-silent mutations leaves little room for an obligatory postnatal second event -- has to accommodate the fact that the best available human model of the fusion-positive newborn does not develop leukaemia at all. The second event may be epigenetic or contextual rather than mutational, which is not the same as absent. Read: abstract only. Montes R et al. 2014