Hypothesis: the survival benefit of menin inhibition is governed by the rate at which it delivers patients to a curative consolidation, not by its remission rate — and the field is measuring the wrong thing

NPM1-mutant AML in older adults; the duration figure should not be assumed to transfer to KMT2A-rearranged infant ALL, where it is unmeasured.

• The 'short-duration instrument' premise is a property of monotherapy, not of the drug class. With decitabine/cedazuridine and venetoclax added, median duration of CR/CRh was 10.5 months overall in relapsed/refractory AML (not reached in the KMT2A-rearranged subset as reported), against the 4.6 months of ziftomenib monotherapy that anchors this entry. If duration more than doubles with combination partners, then reinvestment risk is not the dominant term in the drug's value and the argument for replacing CR/CRh with a conversion endpoint loses its premise. Caveat: cross-trial, different agents, different populations and different definitions of response duration; this is an order-of-magnitude observation, not a formal comparison. Issa GC et al. 2026

Retrospective and adult; the survival difference between those achieving CR and those not is exactly the kind of comparison that selection confounds.

• WHERE THE ANALOGY BREAKS, AND IT BITES THE PROPOSED ENDPOINT. Conversion to a curative-intent consolidation is not an exposure a patient has at randomisation; it is an exposure that can only be acquired by surviving, in remission, long enough to receive it. Suissa's analysis of immortal time bias shows that when an exposure is defined this way, the follow-up between cohort entry and the acquisition of exposure is time during which the outcome could not occur, and misclassifying or excluding it biases the rate ratio in favour of the exposed group in proportion to the duration of the immortal time — and the bias is *more* pronounced when the hazard function is decreasing, which is exactly the front-loaded hazard this graph documents in KMT2A-rearranged infant ALL. The '15.4 months versus 3.4 months' contrast in the cited cohort has the same structure. So the hypothesis's central regression — that between-cohort variation in conversion rate explains more OS variance than CR/CRh rate does — can come out 'true' as an artefact of how the covariate is constructed. Any test of this hypothesis must use a landmark or time-dependent-exposure analysis, and the proposal to adopt conversion-to-consolidation as a *primary trial endpoint* is the version of the idea most exposed to the bias, because it rewards regimens that keep fragile patients alive to the landmark rather than regimens that cure. Suissa S 2008

Chemotherapy era, before menin inhibitors and before frontline blinatumomab. It establishes that the bridge frequently fails, not that a menin inhibitor would fix it.

• The class was approved, and is being used, on CR/CRh despite a transplant conversion rate that is neither high nor the stated basis of benefit: approximately 30-40% of responders in the relapsed/refractory menin-inhibitor trials proceeded to allogeneic transplantation. If conversion were the dominant value term, a class delivering it in a minority of responders should not have produced the observed survival signal. This does not refute the hypothesis, but it removes the presumption that the field's use of CR/CRh is a naive error rather than a defensible proxy. Caveat: a review summarising across trials, with no per-trial conversion denominators given. Alati C et al. 2026

The trial reports response, not conversion to consolidation, in its abstract — which is the gap this hypothesis is about.

• 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. For a bridge-rate model the relevant quantity is the fraction reaching transplant in adequate remission, which neither CR+CRh nor overall response reports directly. Journal of Clinical Oncology 2025

Single-arm with historical controls, n=30, median follow-up 26.3 months. It supports the idea that the reinvestment vehicle need not be transplant, and proves nothing about menin inhibitors. **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.

• The infant blinatumomab result is read here as evidence that the consolidation route is the whole value. A narrative review of the same literature reads it the other way: the gain in KMT2A-rearranged paediatric B-ALL came from adding an immunotherapy that deepens and holds remission (2-year DFS 82% versus 49%) and from CAR T cells producing durable remissions in salvage — i.e. from the depth and persistence of the response itself, not from raising the fraction of patients bridged to transplant. Under the hypothesis's own accounting that is a coupon effect masquerading as reinvestment. Caveat: a review, and it does not decompose the effect statistically, which is precisely what `ana-duration-adjusted-value-menin-sequencing` would have to do. Ray A et al. 2025