Oncofusion

start / graph / ana-duration-adjusted-value-menin-sequencing

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q12-analogy

Analysis (runnable): value a menin inhibitor as a short-duration instrument — cohort meta-regression plus a discrete-event model of remission months

A modelling study that takes the published trial numbers and asks what really drives survival: how many patients get into remission, or how many of them get to a transplant or cell therapy while they are there.

The same thing, for a scientist

Two linked pieces. Piece A is a cohort-level meta-regression of overall survival on CR/CRh rate and on conversion-to-consolidation rate across published menin-inhibitor cohorts, with prior-therapy burden, median age and cohort size as covariates. Piece B is a discrete-event simulation of the patient pathway (treat, respond or not, remain in remission for a sampled duration, reach consolidation or not, relapse, salvage or die) parameterised from published estimates, used to compute restricted mean survival time and to run one-way and probabilistic sensitivity analyses over duration of response, conversion rate and CR rate. The output is an explicit statement of which parameter the survival benefit is most sensitive to.

Go deeper
Piece A — the meta-regression, and why it is weak on purpose

There are only a handful of published menin-inhibitor cohorts, so this regression will have single-digit degrees of freedom. That is stated up front rather than hidden. The output is a pair of coefficients with wide confidence intervals and a plot, not a p-value. Its purpose is to show whether the direction and magnitude are consistent with the hypothesis, and to document how many cohorts do not report conversion at all — which is itself the motivating finding.

Inputs: AUGMENT-101 1, KOMET-001 2, the post-failure outcomes cohort 3, plus any further cohorts published by the run date. Conversion definitions must be harmonised or the cohort dropped, and the dropped list must be reported.

Piece B — the simulation, which is where the useful output is

A discrete-event model with states: on treatment, response (CR/CRh or partial), consolidation, post-consolidation, relapse, salvage, death. Parameters from published data, with duration of response sampled from a fitted distribution rather than held at the median. Outputs: restricted mean survival time at 24 months, and tornado plots ranking sensitivity to CR rate, duration of response, conversion rate and post-consolidation cure fraction.

Run the model separately with infant parameters and adult parameters, because the consolidation options and their costs differ so much that a pooled model would be meaningless. For the infant arm, include a non-transplant consolidation branch parameterised from the frontline blinatumomab result 4, which is currently the strongest evidence that the bridge need not be a transplant.

A quality-adjusted variant, weighting infant post-transplant years for growth, endocrine and neurocognitive burden, should be run as a clearly labelled secondary analysis; the weights are contestable and should be presented as a sensitivity range, not a single number.

What this is not

It is not a health-economic evaluation and produces no cost figures. It is not causal: the meta-regression is ecological and the simulation inherits every assumption in its parameters. It cannot tell you whether to give a menin inhibitor to a particular child. What it can do is say which of four numbers the outcome depends on most, which is exactly what is needed to decide what the next trial should measure.

Facts and figures
scriptanalyses/duration-adjusted-value-menin/run.py
statusplanned
result summaryNot yet run. No result should be quoted from this node.
cost tierlow
computelaptop-scale; Python with lifelines and a simple discrete-event loop
timeline2-3 weeks including Kaplan-Meier digitisation
statistical planReport coefficients with confidence intervals rather than p-values given single-digit degrees of freedom; pre-specify the cohort inclusion rule and harmonised conversion definition; report the list of excluded cohorts and the reason for each; run the simulation with probabilistic sensitivity analysis over all parameters simultaneously, not only one-way.
honest limitationThe meta-regression is ecological and will be underpowered. Reporting it as if it established a causal ordering between CR rate and conversion rate would be wrong, and the write-up must lead with the sensitivity ranking from the simulation instead.
inputs
[
  "Published menin-inhibitor cohort outcomes: PMID 39121437, PMID 40997296, PMID 41351513, plus later cohorts at run date",
  "Interfant-06 conversion and outcome parameters, PMID 31283407",
  "Frontline blinatumomab infant consolidation parameters, PMID 37099340",
  "Digitised Kaplan-Meier curves where restricted mean survival time must be recomputed"
]
outputs
[
  "cohort meta-regression coefficients for CR/CRh and conversion rate on overall survival, with confidence intervals and a count of cohorts not reporting conversion",
  "discrete-event model restricted mean survival time at 24 months, separately for infant and adult parameter sets",
  "tornado plots ranking sensitivity to CR rate, duration of response, conversion rate and post-consolidation cure fraction",
  "a labelled secondary analysis weighting infant post-transplant years for long-term burden, presented as a range"
]
tests hypotheses
[
  "hyp-menin-value-dominated-by-bridge-rate"
]
How this could bite us back
  • A model is only as good as its duration-of-response parameter, and the best available figure comes from 69-year-olds with NPM1-mutant AML. Applying it to infants is the weakest link and must be shown as a sensitivity range spanning at least 2 to 12 months.
Still open
  1. Do menin-inhibitor trials report conversion to consolidation consistently enough to harmonise, or does the analysis reduce to documenting that they do not?
  2. What post-consolidation cure fraction should be assumed for an infant receiving a T-cell engager rather than a transplant, given a single 30-patient study?
Every source behind this entry (5)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    AUGMENT-101 phase 2 reported CR+CRh 22.8% and overall response 63.2% in 57 efficacy-evaluable patients with relapsed or refractory KMT2A-rearranged acute leukemia, from 94 treated patients with median age 37 years.

    Menin Inhibition With Revumenib for KMT2A-Rearranged Relapsed or Refractory Acute Leukemia (AUGMENT-101) · Issa GC, Aldoss I, Thirman MJ, et al. · Journal of Clinical Oncology 2024
    PMID 39121437DOI
    What this evidence cannot show

    Duration of response and transplant conversion are not in the abstract and must be extracted from the full text or the label.

    Evidence that argues against this (1)
    • This is one phase 2 cohort, and the figure most often quoted from it is the least favourable framing of a heavily pretreated adult population: median age 37, no prespecified survival endpoint in the registered protocol, and a gap between 63.2% overall response and 22.8% CR+CRh that consists largely of responses without count recovery. Whether CR+CRh or overall response is the right input depends on what the model is for, and the beat uses both without saying which. Journal of Clinical Oncology 2025
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    KOMET-001 phase 2 reported CR/CRh 22%, overall response 33% and median duration of response 4.6 months in 92 patients with relapsed or refractory NPM1-mutant AML at median age 69.

    Ziftomenib in Relapsed or Refractory NPM1-Mutated AML · Wang ES, Issa GC, Erba HP, et al. · Journal of Clinical Oncology 2025
    PMID 40997296DOI
    What this evidence cannot show

    Different genotype and a much older population than the infant question of interest; used as a duration parameter with an explicit transferability caveat in the model.

    Evidence that argues against this (1)
    • The duration parameter is regimen-dependent, so a model parameterised from monotherapy will not generalise to how the class is actually being used. Median duration of CR/CRh with revumenib plus decitabine/cedazuridine plus venetoclax was 10.5 months in relapsed/refractory AML, against 4.6 months for ziftomenib monotherapy. The model should carry duration as a regimen-indexed parameter with a range spanning both, and its conclusions should be reported separately for monotherapy and combination sequencing; a single point estimate taken from monotherapy would understate the class by roughly a factor of two. Issa GC et al. 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    After menin-inhibitor failure, median overall survival from the start of next therapy was 4.4 months in 84 adults, with 15.4 months for those achieving CR/CRi versus 3.4 months otherwise and 40% receiving supportive care only.

    Outcomes of relapsed or refractory acute myeloid leukemia after menin inhibition failure · Multicentre retrospective consortium · Blood Advances 2026
    PMID 41351513DOI
    What this evidence cannot show

    Provides the post-failure survival parameter; retrospective and selected.

    Evidence that argues against this (1)
    • THE MODEL'S CENTRAL PARAMETER IS BIASED AS USUALLY ESTIMATED. The '15.4 months if CR/CRi versus 3.4 months otherwise' contrast, and any conversion-to-transplant probability derived from observational cohorts, are exposures acquired only by surviving long enough to acquire them. Suissa shows this misclassified or excluded immortal time biases the rate ratio in favour of the exposed group in proportion to the length of the immortal period, and that the bias is larger under a decreasing hazard — the hazard shape documented for this population. A decision model that takes these numbers at face value will systematically overstate the value of strategies that raise the consolidation rate. The analysis must either use landmark-conditioned estimates or state the direction and rough magnitude of the residual bias in its sensitivity analysis. Suissa S 2008
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Frontline blinatumomab added to Interfant-06 chemotherapy in 30 KMT2A-rearranged infants gave 2-year disease-free survival 81.6% versus 49.4% in historical controls, providing a parameterisable non-transplant consolidation branch for the infant arm of the model.

    Blinatumomab Added to Chemotherapy in Infant Lymphoblastic Leukemia · van der Sluis IM, de Lorenzo P, Kotecha RS, et al. · New England Journal of Medicine 2023
    PMID 37099340DOI
    What this evidence cannot show

    n=30, single-arm, historical controls; the parameter should be entered with a wide uncertainty range. **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)
    • Thirty infants against a historical comparator, and the comparator was SELECTED from Interfant-06 rather than randomised. Every survival figure this beat draws from that study — 81.6% two-year DFS, 83.3% four-year DFS, 93.3% four-year OS — rests on the same 30 children, so quoting several of them across entries gives one small single-arm study the appearance of several. With 30 patients the confidence interval on any of those figures spans a range that overlaps the control estimate. Calling it a "parameterisable non-transplant consolidation branch" also commits the model to a branch whose transplant-avoidance benefit was never the study's endpoint. Leukemia 2021
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In Interfant-06 only 46% of high-risk infants with an intention to transplant actually received one, because many had early events, giving an empirical conversion-failure rate for the model's baseline arm.

    PMID 31283407DOI
    What this evidence cannot show

    Pre-menin-inhibitor era; used as the historical baseline conversion rate, not as a current estimate.

    Evidence that argues against this (1)
    • The Interfant-06 conversion-failure rate is being used as the model's empirical anchor, but it is specific to an infant chemotherapy protocol with a fixed transplant landmark and does not transfer to the relapsed/refractory adult menin-inhibitor setting the model also covers: there, approximately 30-40% of responders proceeded to allogeneic transplantation. Two very different conversion regimes are being represented by one parameter. The analysis should stratify. Alati C et al. 2026
Connected to