What the treated brain looks like later: neurocognitive late effects after CNS-directed therapy in infancy
also called neurocognitive outcome infant ALL · learning problems after infant leukemia · CNS-directed therapy late effects · IQ after infant leukemia
Babies treated for leukemia get chemotherapy injected directly into the fluid around the brain and spinal cord at the exact age the brain is building its wiring, and many of the survivors later have trouble with attention, thinking speed and school. Stopping the use of skull radiation helped a great deal, but it did not make the problem disappear.
The same thing, for a scientist
CNS-directed therapy for infant acute leukemia is delivered during the peak period of myelination and synaptic refinement. In the only dedicated infant-leukemia late-effects series, 50% of survivors diagnosed at or under 12 months had learning problems and only 24% were free of any late effect, with the odds of academic difficulty rising 18% for every month younger a child was at cranial radiotherapy [E1]. Contemporary protocols, including Interfant-06, have removed cranial irradiation entirely and substituted 12 intrathecal injections plus high-dose systemic methotrexate; the radiation-driven component of injury has therefore largely gone, and studies comparing 18 Gy cranial radiotherapy with high-dose intravenous methotrexate in children treated at 1-5 years show the methotrexate groups scoring near population means on 17 of 18 measures [E3]. What remains after chemotherapy alone is a narrower but real deficit: attention, processing speed and academic achievement are selectively impaired even on radiation-free regimens [E4], and decades out, adult survivors of childhood ALL show 28.6%-58.9% impairment rates across domains with chemotherapy-only patients impaired in every domain tested [E2]. Almost all of this quantitative evidence comes from children treated at older ages, so the infant-specific magnitude is an extrapolation, not a measurement.
The target is under construction1 deeper
The first year of life is when the human brain lays down most of its myelin and prunes most of its synapses. CNS-directed leukemia therapy in an infant therefore arrives during construction rather than after it, and the damage is expressed as a failure to acquire function rather than a loss of function already gained. That is why age at treatment, not just dose, is a determinant of outcome.
The cleanest quantification of the age gradient comes from the St Jude infant-leukemia series: among 34 five-year survivors diagnosed at 12 months or younger, the estimated odds of academic difficulty rose by 18% for each month younger the child was at the time of cranial radiotherapy (P = 0.032) 1. In the same cohort, learning problems affected 50% of survivors and only 8 of 34 (24%) survived with no late effect at all 1. Survivors who received chemotherapy alone had significantly fewer complications and less academic difficulty than those who received cranial radiotherapy with or without transplant (P = 0.004 for academic difficulty) 1.
The same direction of effect appears in older-child cohorts. In the St Jude Lifetime Cohort, the impact of cranial radiotherapy on intelligence, academic and memory function was dependent on young age at diagnosis; risk of executive-function problems increased with time since treatment in a radiation-dose-dependent way 2. In a 20-year follow-up of 64 population-based survivors, the irradiated group had a mean verbal IQ of 91 against 100 in the non-irradiated survivors and 109 in healthy controls, and performance IQ 100 / 111 / 118 respectively (both P < 0.001) 5.
What contemporary infant therapy actually delivers to the CNS
Interfant-06 abolished cranial irradiation on the grounds that the CNS relapse rate in Interfant-99 had been low without it. CNS-directed therapy in the current backbone is therefore: 12 therapeutic intrathecal injections in CNS-negative infants (at least 14 if CNS-positive, CNS2 or after a traumatic lumbar puncture), with intrathecal methotrexate dosed at 6 mg per injection for children under one year and not corrected for body-surface area; plus high-dose systemic methotrexate 5,000 mg/m² x 2 = 10,000 mg/m² in the MARMA course and high-dose cytarabine 3,000 mg/m² x 8 doses = 24,000 mg/m² [E1 caveat; protocol doses are cited in the companion cardiotoxicity node]. Total body irradiation is likewise excluded from the transplant conditioning regimen, explicitly "because of its neurotoxic side effects in these young children".
The design intent is clear: the two therapies with the largest neurocognitive footprint, cranial radiotherapy and TBI, have been deliberately removed from infant protocols. This is the one late-effect domain where the field has already acted.
Did removing radiation fix it? Partly2 deeper
Two lines of evidence say the radiation-free era is genuinely better, and two say it is not clean.
Better. In a retrospective cohort of 79 children with high-risk ALL diagnosed between 1.0 and 4.9 years and assessed a mean of 10.5 years later, the groups given high-dose (8 g/m² x 3) or very-high-dose (33.6 g/m² x 3) intravenous methotrexate instead of 18 Gy cranial radiotherapy scored near the population mean on 17 of 18 neurocognitive measures, and the two methotrexate doses were indistinguishable from each other; the irradiated group performed worse on most measures 3. Methotrexate dose escalation, within that range, did not cost cognition.
Not clean. On St Jude Total Therapy XV, which omitted prophylactic cranial irradiation in every patient, function two years after therapy was "largely age appropriate" but the cohort still showed significant attention deficits and a significantly higher frequency of learning problems than national norms (all P <= 0.005), and higher-intensity CNS-directed chemotherapy conferred elevated risk for attention, processing-speed and academic difficulty (all P <= 0.01) 4. End-of-therapy attention problems predicted lower academic scores two years later 4. Decades out, in 567 adult survivors of childhood ALL at a mean of 26 years from diagnosis, impairment rates ran from 28.6% to 58.9% across domains and survivors treated with chemotherapy only showed increased impairment in all domains (all P < 0.006); in survivors who received no cranial radiotherapy at all, dexamethasone exposure was associated with impaired attention (RR 2.12, 95% CI 1.11-4.03) and executive function (RR 2.42, 95% CI 1.20-4.91) 2.
| Comparison | Cohort | Age at treatment | Finding |
|---|---|---|---|
| Chemo only vs CRT +/- BMT | 34 infant-leukemia survivors, median 13 y follow-up | <= 12 months | Academic difficulty far commoner with CRT (P = 0.004); 18% higher odds per month younger at CRT 1 |
| 18 Gy CRT vs HD/VHD IV MTX | 79 high-risk ALL, mean 10.5 y after diagnosis | 1.0-4.9 y | MTX groups near population mean on 17/18 measures; CRT worse on most 3 |
| No CRT in anyone | St Jude Total XV, 211 assessed 2 y off therapy | median childhood | Age-appropriate overall, but excess attention deficits and learning problems vs norms (P <= 0.005) 4 |
| Chemo only vs CRT, 26 y out | 567 SJLIFE adult ALL survivors | childhood | 28.6%-58.9% impaired; chemo-only impaired in all domains 2 |
| Irradiated vs not, 20 y out | 64 population-based ALL survivors | childhood | VIQ 91 vs 100 vs 109 (controls); PIQ 100 vs 111 vs 118 5 |
Deficits do not stay still
Two independent cohorts describe progression rather than a fixed scar. In the 20-year Finnish follow-up, verbal and performance IQ declined over the observation period in irradiated survivors, and verbal IQ declined even in the chemotherapy-only survivors 5. In the St Jude Lifetime Cohort, self-reported behaviour problems increased by 5% per year from diagnosis (RR 1.05, 95% CI 1.01-1.09) and the risk of executive-function problems grew with survival time 2. Neurocognitive late effects therefore have to be priced as a lifetime trajectory, not as a single post-treatment score.
The downstream cost is economic, not only psychometric
In the St Jude Lifetime Cohort, neurocognitive impairment was associated with reduced educational attainment and with unemployment 2. This is the number that matters for the argument that a targeted agent has to beat intensive chemotherapy: the endpoint is not a test score at age eight, it is whether a 25-year-old can hold a job.
A clean, infant-ALL-specific special-education rate is not available in the literature reviewed here. The 50% learning-problem figure from the infant series 1 is the closest available proxy and it is not the same measurement. Anyone quoting a special-education percentage for infant KMT2A-rearranged ALL survivors is almost certainly quoting an older-child or brain-tumour cohort.
How much of this is extrapolation (read this before quoting a number)
Infant KMT2A-rearranged ALL survivors are few: with roughly 90 US infant ALL cases a year and historical event-free survival below 50%, the entire global population of adult survivors treated as infants on a modern protocol is small, and no cohort of that population has been given a formal neuropsychological battery at scale.
Therefore:
- Infant-specific and directly measured: the 50% learning-problem rate, the 24% late-effect-free rate, and the 18%-per-month age gradient 1. But n = 34, mixed ALL and AML, treated between the 1970s and 1990s, and 24 of 34 received cranial radiotherapy which no infant receives today. Its main value is the age gradient, not the absolute rates.
- Extrapolated from children treated at 1-5 years: the finding that replacing cranial radiotherapy with high-dose methotrexate largely preserves cognition 3.
- Extrapolated from childhood ALL at all ages: the residual attention/processing-speed signature on chemotherapy-only regimens 4, the decades-out impairment rates 2, and the progressive decline 5.
The honest statement is: we know the direction and we do not know the magnitude for a baby treated on Interfant-06.
| kind | late_effect_domain |
|---|---|
| domain | neurocognitive |
| evidence quality note | No published neuropsychological series exists for survivors of infant KMT2A-rearranged ALL treated on a radiation-free contemporary protocol. All magnitude estimates here are borrowed from older-child cohorts. |
exposures
[
{
"exposure": "cranial radiotherapy",
"status_in_current_infant_protocols": "removed",
"note": "Interfant-06 gives no CNS irradiation; TBI is excluded from transplant conditioning explicitly for neurotoxicity in young children"
},
{
"exposure": "intrathecal methotrexate",
"dose": "6 mg per injection for age <1 year, not corrected for body-surface area",
"n_injections": 12,
"n_injections_if_cns_positive": 14,
"source": "Interfant-06 protocol section 1.3.8"
},
{
"exposure": "high-dose systemic methotrexate",
"dose_mg_per_m2_per_course": 5000,
"courses": 2,
"cumulative_mg_per_m2": 10000,
"course": "MARMA",
"source": "Interfant-06 protocol section 5.5"
},
{
"exposure": "high-dose cytarabine",
"dose_mg_per_m2_per_dose": 3000,
"doses": 8,
"cumulative_mg_per_m2": 24000,
"course": "MARMA second phase",
"source": "Interfant-06 protocol section 5.5"
},
{
"exposure": "dexamethasone",
"note": "associated with impaired attention (RR 2.12) and executive function (RR 2.42) in unirradiated ALL survivors",
"cohort": "St Jude Lifetime Cohort"
}
]infant specific outcomes
{
"cohort": "St Jude infant leukemia survivors (Leung 2000)",
"n": 34,
"age_at_diagnosis": "<=12 months",
"median_follow_up_years": 13,
"follow_up_range_years": [
5.7,
29
],
"learning_problems_pct": 50,
"growth_problems_pct": 66,
"free_of_any_late_effect_pct": 24,
"odds_increase_academic_difficulty_per_month_younger_at_crt_pct": 18,
"p_value": 0.032
}extrapolated outcomes — the numbers behind the chart
[
{
"cohort": "St Jude Lifetime Cohort, adult childhood-ALL survivors",
"n": 567,
"mean_years_since_diagnosis": 26,
"mean_age_years": 33,
"impairment_rate_range_pct": [
28.6,
58.9
],
"note": "chemotherapy-only survivors impaired in all domains (P < 0.006)"
},
{
"cohort": "Toronto, high-risk ALL diagnosed 1.0-4.9 y",
"n": 79,
"mean_years_since_diagnosis": 10.5,
"crt_dose_gy": 18,
"hd_mtx_g_per_m2_x3": 8,
"vhd_mtx_g_per_m2_x3": 33.6,
"note": "MTX groups near population mean on 17/18 measures; no difference between MTX doses"
},
{
"cohort": "Oulu, population-based childhood ALL, ~20 y follow-up",
"n": 64,
"verbal_iq": {
"irradiated": 91,
"non_irradiated": 100,
"controls": 109
},
"performance_iq": {
"irradiated": 100,
"non_irradiated": 111,
"controls": 118
}
}
]- A menin inhibitor or other targeted agent that reduces systemic chemotherapy but keeps the intrathecal schedule may leave most of the neurocognitive exposure untouched: CNS prophylaxis is retained precisely because CNS relapse is the failure mode nobody will risk.
- If a targeted agent has poor CNS penetration, the CNS becomes a sanctuary site and the intrathecal burden may have to go up, not down - trading systemic toxicity for more of the exposure that drives this node.
- Because the neurocognitive benefit of removing cranial radiotherapy has already been banked, the marginal neurocognitive gain available to a new agent is smaller than a naive reading of the historical literature implies; quoting 1990s IQ deficits as the bar to beat overstates the win.
- Improved survival mechanically increases the number of survivors living with residual attention and processing-speed deficits, so a successful drug will make the aggregate late-effect burden in the population look worse before it looks better.
- What is the full-scale IQ, processing-speed and working-memory profile of survivors of infant KMT2A-rearranged ALL treated on Interfant-06 with no cranial irradiation? No such series has been published, so the single most decision-relevant number in this node does not exist.
- Does intrathecal methotrexate dosed by age rather than body-surface area over-expose the smallest infants relative to older children, and is there a measurable neurocognitive dose-response across the 12-injection schedule?
- Is the progressive decline described in irradiated survivors also present in chemotherapy-only infant survivors, or does the trajectory flatten once radiation is removed?
- What fraction of infant-ALL survivors require special educational support, measured as such rather than proxied by 'learning problems'?
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Among 34 survivors of leukemia diagnosed at 12 months or younger (median follow-up 13 years), learning problems affected 50%, growth problems 66%, hypothyroidism 15% and pubertal problems 12%; only 24% had no late effect, and the odds of academic difficulty increased by 18% for each month younger the child was at cranial radiotherapy (P = 0.032).
Late effects in survivors of infant leukemia · Leung W, Hudson M, Zhu Y, Rivera GK, Ribeiro RC, Sandlund JT, Bowman LC, Evans WE, Kun L, Pui CH · Leukemia 2000What this evidence cannot show
Small single-institution series: 34 survivors (19 ALL, 15 AML), diagnosed across an era when 24 of 34 received cranial radiotherapy and 7 received bone marrow transplant. No infant on a current protocol receives cranial irradiation, so the absolute rates overstate what a child treated on Interfant-06 faces. The cohort mixes ALL and AML and does not report KMT2A status. It is nonetheless the only dedicated infant-leukemia late-effects cohort identified, and the per-month age gradient is its most transferable result.
Evidence that argues against this (1)
- Interfant-06 removed cranial irradiation entirely on the basis that CNS relapse was already low without it, so the dominant exposure driving academic difficulty in this cohort no longer exists; the age gradient may therefore now act on a much smaller injury. source 2006
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In 567 adult survivors of childhood ALL assessed a mean of 26 years after diagnosis (mean age 33), neurocognitive impairment ranged from 28.6% to 58.9% by domain; chemotherapy-only survivors showed increased impairment in all domains (all P < 0.006); in unirradiated survivors dexamethasone was associated with impaired attention (RR 2.12) and executive function (RR 2.42); and impairment was associated with reduced educational attainment and unemployment.
Neurocognitive outcomes decades after treatment for childhood acute lymphoblastic leukemia: a report from the St Jude Lifetime Cohort Study · Krull KR, Brinkman TM, Li C, Armstrong GT, Ness KK, Srivastava DK, Gurney JG, Kimberg C, Krasin MJ, Pui CH, Robison LL, Hudson MM · Journal of Clinical Oncology 2013What this evidence cannot show
Single-institution survivor cohort treated largely before 2000, so exposures include cranial radiotherapy at doses no longer used. Participation was 72.8% of eligible with 76.8% of those evaluated, leaving selection bias in an unknown direction. Critically, this cohort is childhood ALL at all ages, not infant ALL: median age at diagnosis is not in the infant range and infants are a negligible fraction. Applying these impairment percentages to an infant-treated survivor is an extrapolation.
Evidence that argues against this (1)
- A prospective cohort treated on a fully radiation-free protocol found neurocognitive function largely age-appropriate two years off therapy, which argues that the very high impairment rates here are partly an artefact of an older, radiation-heavy treatment era. Jacola LM et al. 2016
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In 79 children with high-risk ALL diagnosed between 1.0 and 4.9 years and tested a mean of 10.5 years after diagnosis, those given high-dose (8 g/m2 x 3) or very-high-dose (33.6 g/m2 x 3) intravenous methotrexate instead of 18 Gy cranial radiotherapy scored near the population mean on 17 of 18 neurocognitive measures with no difference between methotrexate doses, while the irradiated group performed worse on most measures.
Comparison of long-term neurocognitive outcomes in young children with acute lymphoblastic leukemia treated with cranial radiation or high-dose or very high-dose intravenous methotrexate · Spiegler BJ, Kennedy K, Maze R, Greenberg ML, Weitzman S, Hitzler JK, Nathan PC · Journal of Clinical Oncology 2006What this evidence cannot show
Retrospective, non-randomised: the three CNS-prophylaxis modalities map onto three treatment eras, so era effects and secular changes in supportive care are confounded with modality. Only 79 of 120 eligible children participated. The youngest patients are 12 months, so this cohort does not include infants; it is the closest available proxy for 'treated very young' and is used here as an upper bound on how much of the deficit radiotherapy accounted for. It also says nothing about intrathecal methotrexate burden, which was given to all groups.
Evidence that argues against this (1)
- A radiation-free prospective cohort still found excess attention and processing-speed deficits relative to norms and a dose-effect of CNS-directed chemotherapy intensity, so 'near population mean' should not be read as 'no injury from methotrexate'. Jacola LM et al. 2016
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
On St Jude Total Therapy XV, which omitted prophylactic cranial irradiation in all patients, neurocognitive function two years after therapy was largely age-appropriate but the cohort showed significant attention deficits and a greater frequency of learning problems than national norms (all P <= 0.005); higher-intensity CNS-directed chemotherapy conferred elevated risk for attention, processing-speed and academic difficulty (all P <= 0.01); and end-of-therapy attention problems predicted lower academic scores two years later.
Longitudinal Assessment of Neurocognitive Outcomes in Survivors of Childhood Acute Lymphoblastic Leukemia Treated on a Contemporary Chemotherapy Protocol · Jacola LM, Krull KR, Pui CH, Pei D, Cheng C, Reddick WE, Conklin HM · Journal of Clinical Oncology 2016What this evidence cannot show
Follow-up is only two years off therapy, which is far too early to capture the progressive component that longer cohorts describe, and the sample shrinks and shifts between the three assessment points (142 / 243 / 211). Total XV is a childhood-ALL protocol, not an infant protocol: its CNS-directed intensity, steroid exposure and methotrexate schedule differ from Interfant-06. 'Higher-intensity CNS-directed chemotherapy' is a protocol risk stratum, not a measured methotrexate dose, so the exposure-response reading is indirect.
Evidence that argues against this (1)
- No identified study disagrees that attention and processing speed are selectively affected on radiation-free protocols. The open disagreement is about magnitude and permanence, not direction. Krull KR et al. 2013
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In 64 population-based childhood ALL survivors tested a mean of 20 years after diagnosis, mean verbal IQ was 91 in irradiated survivors, 100 in non-irradiated survivors and 109 in healthy controls, and performance IQ 100, 111 and 118 respectively (both P < 0.001), with a significant decline in both indices over follow-up in the irradiated group and in verbal IQ in the non-irradiated group.
Progressive neurocognitive impairment in young adult survivors of childhood acute lymphoblastic leukemia · Harila MJ, Winqvist S, Lanning M, Bloigu R, Harila-Saari AH · Pediatric Blood & Cancer 2009What this evidence cannot show
Small (64 survivors, 20 of them chemotherapy-only) and from a single Finnish region treated in an era of routine cranial irradiation. The control group is 45 healthy young adults rather than siblings, so socioeconomic confounding is not controlled; the controls' mean verbal IQ of 109 is above the population mean of 100, which inflates the apparent gap. Serial comparison was possible in only 45 of 64. No infants. Use for the direction of the trajectory, not for the absolute IQ values.
Evidence that argues against this (1)
- A contemporaneous cohort of 24 adults who received only 18 Gy found average full-scale IQ despite below-average performance on speed-dependent computerised tasks, suggesting global IQ can look normal while specific processing deficits persist - i.e. IQ point differences are a crude and possibly era-specific readout. Edelstein K et al. 2011
- ← supportsNine survivors in ten carry a long-term complication, and it is worse the younger they were
38.8% neurodevelopmental complications with only 28.4% formally assessed
uncosted late effect
MEF2C haploinsufficiency causes severe intellectual disability and epilepsy in humans
what survivors live with, which sets the bar for any new therapy
total body irradiation or busulfan conditioning in infancy carries neurocognitive late effects
the stated reason for banning TBI in infants, and the endpoint with the thinnest supporting data