FAERS shows no endocrine signal for either menin inhibitor - and proves it could not have
also called FAERS menin endocrine · openFDA disproportionality menin inhibitor
We searched the FDA's public database of side-effect reports for every hormone- and bone-related problem we could think of, in every report naming a menin inhibitor. Not one of twenty terms appeared, while the two known side effects of the drug appeared hundreds of times more often than chance.
The same thing, for a scientist
Run on 2026-09-18 against the openFDA drug/event API (20,692,690 reports). Revumenib/REVUFORJ: 904 reports, of which **0** carried any of 20 prespecified endocrine or bone-mineral MedDRA preferred terms, against 9.68 expected in aggregate under a whole-database null. Ziftomenib/KOMZIFTI: 48 reports, 0 of 20 terms, 0.51 expected. The positive controls behaved exactly as they should: differentiation syndrome 43 observed against 0.031 expected for revumenib (roughly 1,400-fold) and QTc prolongation 44 against 1.55 (roughly 28-fold). So the pipeline works, the drug's known on-target toxicities are detected with enormous disproportionality, and the endocrine terms are entirely absent. The decisive interpretive point is that the revumenib **label** reports intact PTH raised in 34% of patients, and FAERS contains **zero** such reports - which demonstrates directly that spontaneous reporting does not capture asymptomatic laboratory endocrine abnormalities, and is exactly why FDA wrote that 505(k)(1) reporting would not be sufficient and required a clinical trial instead.
Result
Revumenib / REVUFORJ, 904 FAERS reports
| MedDRA preferred term | Observed | Expected under null |
|---|---|---|
| Hyperparathyroidism | 0 | 0.089 |
| Blood parathyroid hormone increased | 0 | 0.126 |
| Hypercalcaemia | 0 | 0.532 |
| Hypocalcaemia | 0 | 0.814 |
| Blood calcium increased | 0 | 0.379 |
| Hyperphosphataemia | 0 | 0.091 |
| Hypophosphataemia | 0 | 0.317 |
| Blood phosphorus increased | 0 | 0.118 |
| Blood alkaline phosphatase increased | 0 | 1.074 |
| Epiphyses premature fusion | 0 | 0.013 |
| Growth retardation | 0 | 0.128 |
| Hyperprolactinaemia | 0 | 0.280 |
| Hypothyroidism | 0 | 1.373 |
| Hyperglycaemia | 0 | 1.566 |
| Hypoglycaemia | 0 | 2.093 |
| Adrenal insufficiency | 0 | 0.516 |
| Neuroendocrine tumour | 0 | 0.061 |
| Parathyroid tumour benign | 0 | 0.032 |
| Pituitary tumour benign | 0 | 0.071 |
| Insulinoma | 0 | 0.008 |
| Total (20 terms) | 0 | 9.68 |
| Differentiation syndrome (control) | 43 | 0.031 |
| Electrocardiogram QT prolonged (control) | 44 | 1.551 |
Ziftomenib / KOMZIFTI, 48 reports: 0 of 20 terms, 0.51 expected in aggregate; differentiation syndrome 4 observed against 0.002 expected.
Script: analyses/faers-menin-inhibitor-endocrine/run.py. Results: analyses/faers-menin-inhibitor-endocrine/results.json 1.
What this licenses, stated precisely1 deeper
Three separate conclusions, with different strengths.
-
Clinically apparent endocrine disease is not being reported. Zero reports of hyperparathyroidism, hypercalcaemia, neuroendocrine tumour or growth retardation across 904 revumenib exposures in roughly two years post-approval. Per term the expectations are all below 2.1, so no single term is individually informative, but the aggregate deficit (0 against 9.68) is not what one would see if the drug were producing symptomatic endocrine events at any appreciable rate. This is real, if modest, evidence against the strong form of the hypothesis on current exposure durations.
-
The absence proves nothing about the laboratory signal. The label reports raised intact PTH in 34% of patients. FAERS has zero reports of it. A 34%-frequency finding generating zero spontaneous reports is a complete failure of the reporting channel for asymptomatic abnormalities, not evidence that the abnormality is absent. This is a measured demonstration of the point FDA made in writing when it required a clinical trial instead of relying on spontaneous reporting 2.
-
The method is sound, because the controls fire. Differentiation syndrome at roughly 1,400-fold over background and QTc at 28-fold show that FAERS detects this drug's on-target toxicities loudly. The endocrine silence is therefore a property of what gets reported, not of whether reports about this drug exist.
The honest summary: FAERS can rule out a loud, symptomatic, early endocrine catastrophe. It has done so. It cannot address biochemical drift, growth velocity, or anything with a latency longer than the drug has existed - and those are the outcomes the hypothesis is actually about.
Limitations that matter
- Confounded denominator. The expected counts use a whole-FAERS background whose case mix is nothing like relapsed acute leukaemia. Oncology reporting is dominated by serious and fatal events, which plausibly depresses reporting of mild laboratory abnormalities relative to background; part of the 0-versus-9.68 deficit is certainly this artefact rather than absence of events.
- Short observation. Revumenib was approved in November 2024. Two years of post-marketing exposure cannot address an endpoint modelled on a syndrome whose earliest human manifestation is at 9 years.
- Small ziftomenib denominator. 48 reports is too few for any inference at all.
- Product-name matching. Reports using a trial code or a misspelling are missed, and
OFF LABEL USEis the single most frequent term in the revumenib set (233), which indicates a reporting population skewed toward administrative rather than clinical events.
| script | analyses/faers-menin-inhibitor-endocrine/run.py |
|---|---|
| status | run |
| result summary | 0 of 20 endocrine/bone-mineral MedDRA terms in 904 revumenib reports (9.68 expected in aggregate) and 0 in 48 ziftomenib reports (0.51 expected); positive controls differentiation syndrome 43 vs 0.031 and QTc 44 vs 1.551 for revumenib. Rules out a loud symptomatic early endocrine effect; cannot address the 34% laboratory PTH signal, which generates zero spontaneous reports. |
| n reports all drugs | 20692690 |
| n reports revumenib | 904 |
| n reports ziftomenib | 48 |
| terms tested | 20 |
| terms observed | 0 |
| aggregate expected revumenib | 9.68 |
| aggregate expected ziftomenib | 0.51 |
| results file | analyses/faers-menin-inhibitor-endocrine/results.json |
| reproducible | true |
| requires api key | false |
inputs
[ "openFDA drug/event API (FAERS), queried 2026-09-18" ]
positive controls
{
"differentiation_syndrome": {
"observed": 43,
"expected": 0.031
},
"ecg_qt_prolonged": {
"observed": 44,
"expected": 1.551
}
}- This analysis is the clearest piece of evidence in the beat that points AGAINST the hypothesis, and it should be reported as such: no symptomatic endocrine event of any kind has been spontaneously reported for either approved menin inhibitor.
- It is also a demonstration of why that null is weak: a labelled 34%-frequency laboratory abnormality generated zero reports, so the channel is blind to exactly the endpoint in question.
- The aggregate deficit (0 observed against 9.68 expected) is partly an artefact of comparing an oncology reporting population against a whole-database background, and should not be quoted as though it were a protective effect.
- Does a disproportionality analysis restricted to an oncology comparator set - all reports for AML or ALL drugs rather than the whole database - change the expected counts enough to matter? That is a one-line change to the script and would remove the largest confounder.
- Will the picture change once post-marketing exposure extends beyond a few years and into maintenance use? This analysis should be re-run annually; the script is parameter-free.
- Is there a European (EudraVigilance) or Japanese (JADER) signal that FAERS misses? Both are publicly queryable and neither was searched here.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
Querying the openFDA drug/event API on 2026-09-18 across 20,692,690 FAERS reports returned 904 reports naming revumenib or REVUFORJ and 48 naming ziftomenib or KOMZIFTI; none of 20 prespecified endocrine and bone-mineral MedDRA preferred terms appeared in any of them, against aggregate expectations under a whole-database null of 9.68 and 0.51 respectively, while the positive controls differentiation syndrome (43 observed, 0.031 expected) and electrocardiogram QT prolonged (44 observed, 1.551 expected) were heavily over-represented for revumenib.
openFDA drug/event API query of FAERS for menin-inhibitor endocrine and bone-mineral adverse event terms · US Food and Drug Administration (data); analysis run for this graph, analyses/faers-menin-inhibitor-endocrine/run.py · openFDA FAERS, queried 2026-09-18 2026What this evidence cannot show
Run directly by this agent; the script, the raw counts and the results file are in the repository at analyses/faers-menin-inhibitor-endocrine/. FAERS is a spontaneous reporting system: counts are reports, not patients, with no denominator of exposed individuals, heavy reporting bias, duplicates, and no causality assessment. Expected counts use a whole-database background with a case mix unlike relapsed acute leukaemia, so they are indicative rather than a valid statistical null. The observation window is roughly two years post-approval for revumenib and less for ziftomenib.
Evidence that argues against this (1)
- The same label that this analysis finds no FAERS reports against states that intact parathyroid hormone increased in 34% of treated patients and lists hyperparathyroidism as a clinically relevant adverse reaction under Endocrine disorders. A 34%-frequency laboratory finding producing zero spontaneous reports establishes that the absence of FAERS reports is uninformative about laboratory-level endocrine perturbation. Syndax Pharmaceuticals / US FDA 2025
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
FDA determined that an analysis of spontaneous postmarketing adverse events reported under FDCA 505(k)(1) would not be sufficient to assess signals of serious risk including elevated parathyroid hormone for revumenib, that the active postmarket risk identification system under 505(k)(3) would also not suffice, and that only a clinical trial would be sufficient, and therefore imposed postmarketing requirement 4730-1.
NDA 218944 approval letter for REVUFORJ (revumenib), including postmarketing requirements 4730-1 to 4730-6 · US Food and Drug Administration, Center for Drug Evaluation and Research · Drugs@FDA, NDA 218944 approval letter 2024What this evidence cannot show
Read verbatim from the NDA 218944 approval-letter PDF on 2026-09-18. This analysis independently reproduces the empirical basis for that determination.
Evidence that argues against this (1)
- This is the finding that bounds the whole analysis and it belongs in its summary: the regulator concluded in advance that spontaneous postmarketing reports would be insufficient for this question. A disproportionality analysis of FAERS therefore cannot generate reassurance, only hypotheses, and its null result should not be cited as evidence of endocrine safety anywhere in this graph. FDA Drugs@FDA 2024
zero endocrine reports across 904 revumenib exposures - but the labelled 34% PTH finding also generated zero reports, so the channel is blind
independently reproduces the empirical basis for FDA's determination that spontaneous reporting would not suffice
904 FAERS reports; differentiation syndrome and QTc heavily over-represented, no endocrine term present
the pharmacovigilance arm of the test, already run