Oncofusion

start / graph / drug-thioguanine

Drugstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q51-backbone

6-Thioguanine - the drug a 1,498-child randomised trial removed from maintenance and left in intensification, which is exactly where Interfant-06 uses it

also called 6-thioguanine · thioguanine · 6-TG · TG · tioguanine · Lanvis · 2-amino-6-mercaptopurine · thiopurine (class)

Thioguanine is mercaptopurine's more potent sibling. A large randomised trial found it caused serious, sometimes permanent liver damage in about one child in nine when used long-term, so it was dropped from the two-year maintenance phase - but kept for the short intensive blocks, and that is how the infant protocol uses it.

The same thing, for a scientist

6-Thioguanine is a thiopurine that reaches thioguanine nucleotides in one fewer metabolic step than mercaptopurine and produces higher DNA-thioguanine incorporation per milligram. Interfant-06 gives it **60 mg/m2/day orally on OCTADA(D) part I days 1-28 and part II days 36-49** [E1] - short intensification blocks - and uses **mercaptopurine**, not thioguanine, for protocol IB, MARMA and the two years of maintenance [see `drug-mercaptopurine`]. **That division is not arbitrary and the evidence behind it is unusually good.** UK ALL97 randomised 1,498 children between 6-thioguanine and 6-mercaptopurine for interim maintenance and continuing therapy, with **all patients receiving 6-thioguanine during intensification courses**: at a median 6 years there was no difference in event-free or overall survival; thioguanine lowered isolated CNS relapse (OR 0.53, 95% CI 0.30-0.92) but raised death in remission (OR 2.22, 1.20-4.14) mainly from infection; **95 patients developed hepatic veno-occlusive disease, 82 of them on thioguanine - 11% of all thioguanine recipients** - and about 5% of thioguanine recipients show long-term non-cirrhotic portal hypertension from periportal fibrosis or nodular regenerative hyperplasia [E2]. The Children's Oncology Group CCG-1952 trial reached the same conclusion independently [E3], and the hepatotoxicity risk tracks **TPMT genotype** [E4]. **This drug is named in one single entry of this graph** - the least represented backbone agent of the beat.

Go deeper
Where it sits, and why the phase matters more than the drug

Interfant-06 1:

phasethioguanine
OCTADA(D) part I, days 1-2860 mg/m2/day PO, once daily
OCTADA(D) part II, days 36-4960 mg/m2/day PO, once daily
protocol IB, MARMA, maintenancemercaptopurine, not thioguanine

ALL97's design is the reason this split exists. That trial randomised the maintenance thiopurine and gave thioguanine to everyone during intensification 2 - so its veno-occlusive disease finding is a finding about prolonged thioguanine, and its conclusion ('6-mercaptopurine should remain the thiopurine of choice for continuing therapy') is explicitly about continuing therapy. Short intensification exposure was never the thing found harmful, and no trial has removed it.

Interfant-06 gives thioguanine for 28 + 14 = 42 days total, against up to two years of mercaptopurine. The protocol is therefore aligned with the evidence, and this entry exists partly to record that a division that looks like an arbitrary quirk of a complicated protocol is in fact the exact shape of a 1,498-child randomised result.

The toxicity, and the genotype that predicts it

ALL97, 1,498 children, median follow-up 6 years 2:

  • event-free and overall survival: no difference
  • isolated CNS relapse: lower with thioguanine, OR 0.53 (0.30-0.92), P = 0.02
  • death in remission: higher, OR 2.22 (1.20-4.14), P = 0.01, mainly infection during continuing therapy
  • hepatic veno-occlusive disease: 95 patients, 82 of them randomised to thioguanine = 11% of all thioguanine recipients
  • long-term: about 5% of thioguanine recipients with non-cirrhotic portal hypertension from periportal fibrosis or nodular regenerative hyperplasia

Why the CNS result is worth keeping for this disease. KMT2A-rearranged infant ALL has high CNS involvement at diagnosis and CNS relapse is a recurrent problem; the one advantage thioguanine demonstrated in a randomised trial is exactly the one this disease needs. It was rejected because the death-in-remission cost exceeded it - in children aged 1-18, on a two-year exposure, with no genotype selection.

And the risk is partly predictable. TPMT genotype is associated with thioguanine-related hepatic veno-occlusive disease in children with ALL 4. That turns a 11% population risk into a stratifiable one, and no trial has re-asked the ALL97 question with genotype-guided dosing.

Facts and figures
targetdirect conversion to thioguanine nucleotides incorporated into DNA (one fewer metabolic step than mercaptopurine, and higher DNA-thioguanine incorporation per milligram). No HPRT1, TPMT or NUDT15 node exists in this graph as of cycle 28.
modalitysmall molecule, purine antimetabolite (thiopurine), oral
interfant06 positionOCTADA(D) part I days 1-28 and part II days 36-49 at 60 mg/m2/day - 42 days total. NOT used in protocol IB, MARMA or maintenance, where mercaptopurine is used instead.
interfant06 dose60 mg/m2/day PO
infant dose adjustmentx 2/3 under 6 months, x 3/4 from 6 through 12 months
protocol monitoring gapInterfant-06 chapter 7 gives toxicity-modification guidance for nine agents and NONE for thioguanine, so the drug whose signature toxicity is hepatic veno-occlusive disease has no hepatic stopping rule in this protocol.
kmt2a specific evidenceNONE. Thioguanine is in the protocol and that is all. The thiopurine class shows no infant-versus-older-children difference in the MTT assay but relative resistance in proB ALL of all ages (see drug-mercaptopurine E2); no thioguanine-specific measurement in KMT2A-rearranged or infant cells was found.
named in graph entries1
randomised evidence
{
  "trial": "UK ALL97 (ISRCTN26727615), 1,498 children aged 1-18",
  "randomised_phase": "interim maintenance and continuing therapy only - ALL patients received thioguanine during intensification, which is Interfant-06's use of it",
  "efs_and_os": "no difference at median 6 years",
  "isolated_cns_relapse_or": 0.53,
  "isolated_cns_relapse_ci": "0.30 to 0.92",
  "death_in_remission_or": 2.22,
  "death_in_remission_ci": "1.20 to 4.14",
  "veno_occlusive_disease": "95 patients, 82 on thioguanine = 11% of thioguanine recipients",
  "long_term_portal_hypertension_pct": 5,
  "conclusion": "6-mercaptopurine should remain the thiopurine of choice for CONTINUING therapy"
}
key ae
[
  "hepatic veno-occlusive disease (11% of thioguanine recipients on prolonged exposure in ALL97)",
  "non-cirrhotic portal hypertension from periportal fibrosis or nodular regenerative hyperplasia (about 5% long-term)",
  "myelosuppression, severe in TPMT or NUDT15 variant carriers",
  "increased death in remission from infection on prolonged exposure"
]
resistance
[
  "loss of HPRT preventing activation",
  "mismatch repair deficiency preventing thioguanine-nucleotide-triggered apoptosis",
  "no infant- or KMT2A-specific mechanism identified in this beat"
]
trials
[
  "NCT00550992",
  "ISRCTN26727615"
]
How this could bite us back
  • **About one in nine children given prolonged thioguanine developed hepatic veno-occlusive disease and about one in twenty has lasting non-cirrhotic portal hypertension.** Interfant-06 gives 42 days of it to infants and specifies no liver monitoring or stopping rule, while specifying such rules for nine other drugs.
  • The one randomised advantage this drug has - nearly halved isolated CNS relapse - is the endpoint that matters most in a disease with frequent CNS involvement at diagnosis. Anyone proposing to reinstate thioguanine for that reason is proposing to reintroduce a hepatotoxin to babies on the strength of a subgroup logic from a trial that enrolled nobody under 1.
  • ALL97 was designed to answer a maintenance question and gave thioguanine to both arms during intensification. That means the use Interfant-06 makes of thioguanine has never been randomised against anything, and the trial most often cited about this drug is silent on the only way this protocol uses it.
  • TPMT explains part of the hepatotoxicity risk and NUDT15, unknown when that work was done, explains much of the rest with a very different ancestral distribution. A stratification built on TPMT alone would fail hardest in the populations least represented in the trials that generated the risk estimate.
  • This drug is named in one file of a 1,553-entry graph, and doxorubicin - which is given nowhere in this protocol - is named in twenty. The graph's attention has been allocated by what the literature discusses rather than by what an infant receives, and this beat exists because of that asymmetry.
Still open
  1. What is the hepatic veno-occlusive disease rate from 42 days of thioguanine in an infant? The 11% figure comes from two years of it in children aged 1-18, and the protocol that gives the short course specifies no hepatic monitoring at all.
  2. Would genotype-guided thioguanine be worth revisiting for CNS control in KMT2A-rearranged infant ALL? Thioguanine nearly halved isolated CNS relapse in ALL97 and was rejected on a toxicity trade-off computed without TPMT or NUDT15 selection, in a population at far lower CNS-relapse risk than this one.
  3. Why does Interfant-06 give no toxicity-modification guidance for thioguanine when it gives it for nine other agents? If that is an oversight rather than a judgement, it is a correctable one.
  4. Is the relative thiopurine resistance of proB ALL - the immunophenotype of this disease - true of thioguanine as well as of mercaptopurine? The one measurement was at class level and was unquantified.
Every source behind this entry (5)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Interfant-06 gives 6-thioguanine 60 mg/m2/day orally on OCTADA(D) part I days 1-28 and part II days 36-49, and uses 6-mercaptopurine rather than thioguanine in protocol IB, MARMA and maintenance; the body-surface-area dose is multiplied by 2/3 under 6 months and 3/4 from 6 through 12 months.

    What this evidence cannot show

    A trial protocol document, not a peer-reviewed publication, and specifically the SKION (Dutch) national version of April 2006; the trial was amended during its course (asparaginase provisions in MARMA and OCTADA(D) among others), so later versions may differ. Doses given are PLANNED doses; delivered cumulative dose per patient is lower wherever a course was omitted, delayed or curtailed by a stopping rule, and the trial report does not publish delivered dose per patient. Retrieved and text-extracted by this beat on 2026-09-20 (90 pages, 194,979 characters); page numbers are the PDF's own. The protocol's chapter 7 toxicity-modification section gives specific guidance for L-asparaginase, cyclophosphamide, high-dose cytarabine, daunorubicin, the glucocorticoids, etoposide, high-dose methotrexate, mitoxantrone and vincristine, and NO guidance for thioguanine - so the protocol specifies no hepatic monitoring or stopping rule for the agent whose signature toxicity is hepatic.

    Evidence that argues against this (1)
    • Total thioguanine exposure in Interfant-06 is 42 days at 60 mg/m2/day multiplied by an age factor, which is far shorter than the interim-maintenance-plus-continuing-therapy exposure that produced the 11% veno-occlusive disease rate in ALL97. Applying that rate to this protocol would be a scope error; the risk at 42 days has not been separately measured in any population, and not at all in infants. Vora A et al. 2006
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    In UK ALL97, 1,498 children with lymphoblastic leukaemia were randomised to 6-thioguanine (n = 750) or 6-mercaptopurine (n = 748) during interim maintenance and continuing therapy, with all patients receiving 6-thioguanine during intensification courses; at a median 6 years of follow-up there was no difference in event-free or overall survival, thioguanine conferred a significantly lower risk of isolated CNS relapse (OR 0.53, 95% CI 0.30-0.92, P = 0.02) offset by an increased risk of death in remission (OR 2.22, 95% CI 1.20-4.14, P = 0.01) mainly from infection during continuing therapy, 95 patients developed hepatic veno-occlusive disease of whom 82 were randomised to thioguanine (11% of all thioguanine recipients), and on long-term follow-up about 5% of thioguanine recipients had evidence of non-cirrhotic portal hypertension from periportal liver fibrosis or nodular regenerative hyperplasia; the authors conclude that 6-mercaptopurine should remain the thiopurine of choice for continuing therapy.

    Toxicity and efficacy of 6-thioguanine versus 6-mercaptopurine in childhood lymphoblastic leukaemia: a randomised trial. · Vora A, Mitchell CD, Lennard L, Eden TO, Kinsey SE, Lilleyman J, et al. · Lancet 2006
    PMID 17046466DOI
    What this evidence cannot show

    **Children aged 1-18 - no infants.** The randomised exposure is interim maintenance plus continuing therapy, not intensification: every patient in both arms received thioguanine during intensification courses, which is the use Interfant-06 makes of it, so the trial contains no comparison of intensification thioguanine against anything. The conclusion is explicitly about continuing therapy. Doses and schedules are those of ALL97, not Interfant.

    Evidence that argues against this (1)
    • The trial's one positive finding for thioguanine - a nearly halved risk of isolated CNS relapse - is the endpoint that matters most in KMT2A-rearranged infant ALL, where CNS involvement at diagnosis is common. ALL97 rejected thioguanine on a toxicity trade-off calculated without pharmacogenetic selection and in a population at much lower CNS-relapse risk than this one. Lennard L et al. 2006
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The Children's Oncology Group CCG-1952 trial independently compared oral 6-mercaptopurine with oral 6-thioguanine in children with standard-risk acute lymphoblastic leukaemia and reported on veno-occlusive disease.

    PMID 20124218DOI
    What this evidence cannot show

    Cited at title level for the existence of an independent North American randomised comparison reaching the same class of conclusion; this beat did not obtain the full text and states no effect size, incidence or outcome from it. Standard-risk childhood ALL, so again no infants. It is recorded because a single-trial result of this importance should not stand alone in the graph.

    Evidence that argues against this (1)
    • Two trials agreeing that prolonged thioguanine causes hepatic veno-occlusive disease in children aged 1-18 is strong evidence about prolonged thioguanine in children aged 1-18 and says nothing about 42 days of it in an infant. Neither trial enrolled the population this graph is about, and the liver of a 4-month-old is not the liver of a 6-year-old. de Kluis T et al. 2025
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The thiopurine methyltransferase genetic polymorphism is associated with thioguanine-related veno-occlusive disease of the liver in children with acute lymphoblastic leukaemia.

    PMID 17015055DOI
    What this evidence cannot show

    Cited at title level; this beat did not obtain the full text and states no odds ratio, allele, or incidence by genotype. It is recorded to establish that the signature toxicity of this drug has a known pharmacogenetic determinant, which converts a population risk into a stratifiable one - not to specify how it should be stratified. The study population is children with ALL, not infants.

    Evidence that argues against this (1)
    • A TPMT association does not make the risk fully predictable: NUDT15 was not known at the time of this work and is now recognised as a major determinant of thiopurine toxicity with a very different ancestral distribution, so a TPMT-only stratification would miss a substantial fraction of susceptible patients, particularly outside European populations. Zhou Y et al. 2024
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    6-Thioguanine is named in exactly one of the 1,553 node files in this graph as of 2026-09-20, the least of any drug in the Interfant-06 backbone - against 66 files naming cytarabine, 62 prednisone, 37 dexamethasone, 27 asparaginase, 20 each for methotrexate, doxorubicin and daunorubicin, 18 mitoxantrone, 18 prednisolone, 17 vincristine, 15 cyclophosphamide and 6 mercaptopurine.

    What this evidence cannot show

    A measurement of this repository, not of the literature, taken 2026-09-20 by case-insensitive regex over all node JSON files (`thioguanine|6-tg` for this drug; each drug counted as the number of FILES containing at least one match, not the number of mentions). The counts are sensitive to the alias set chosen: `6-tg` is a short token that could collide, and `ara-c` inflates the cytarabine count relative to a `cytarabine`-only query. Reported as an inventory fact about the graph.

    Evidence that argues against this (1)
    • A low mention count is not the same as low importance, and the converse is also true: doxorubicin is named in 20 files of this graph and is given nowhere in this protocol. Mention counts measure what has been written about, which is what this beat exists to correct, and should not be used as a proxy for clinical weight in either direction. ClinicalTrials.gov 2007
Connected to