Icovamenib in Type 1 Diabetes: Biomea closes COVALENT-112 after a Phase 2 that was never powered to land

What was tried

Icovamenib (BMF-219) is an oral covalent inhibitor of menin, the scaffolding protein encoded by MEN1 that anchors the KMT2A/MLL1 H3K4 methyltransferase complex. Biomea Fusion built the molecule for KMT2A-rearranged acute leukemia, where menin inhibition has produced a market-approved drug class. The T1D program was a repurposing bet on a separate biology: menin loss releases beta-cell proliferation and survival programs that are silenced in adult islets.

NCT06152042 was a Phase 2 randomized, double-blind, placebo-controlled trial of icovamenib in adults with stage 3 Type 1 diabetes treated with insulin only. The protocol stratified patients by time from diagnosis, then layered two oral doses (100 mg and 200 mg, once daily for 12 weeks) against placebo, with mean change from baseline in stimulated C-peptide AUC at Week 26 as the primary endpoint. Biomea targeted roughly 150 patients. The final actual enrollment, recorded in the ClinicalTrials.gov record updated May 14, 2026, was 37.

In May 2024 the FDA placed a partial clinical hold on the diabetes program after asymptomatic LFT elevations on long-duration dosing. Biomea redesigned around shorter 12-week windows; the hold lifted later that year, but enrollment never recovered. The May 2026 termination was filed as a portfolio prioritization decision.

The biological hypothesis

Menin is a context-dependent regulator of cell proliferation. In KMT2A-rearranged leukemia it is oncogenic, and disrupting the menin–KMT2A interaction collapses HOXA9/MEIS1 expression (Issa et al., 2023). In adult pancreatic beta cells the same protein restrains replication. Karnik and colleagues showed that menin loss drives gestational beta-cell expansion in mice (Karnik et al., Science 2007,). Later work tied menin to TEAD1/YAP-TAZ growth control (Li et al., Cell Reports 2023,) and to FXR/E2F3 signaling in stressed islets (Kong et al., 2024,).

The clinical inference was bold. If covalent menin inhibition can drive proliferation and survival of residual beta cells in the autoimmune-injured pancreas, then a 12-week pulse of icovamenib in recently diagnosed T1D should slow or reverse the early decline in endogenous insulin output. The mechanism is independent of immunosuppression, which has dominated T1D disease modification since teplizumab won approval in 2022 (Herold et al., NEJM 2019,). Icovamenib offered a complementary lever, beta-cell-intrinsic rather than immune-intrinsic.

What actually happened

Biomea announced 52-week results from COVALENT-112 on April 28, 2026. The headline was a 52% mean increase in stimulated C-peptide AUC at Week 12 in the 200 mg recently diagnosed cohort, with retention of most of that gain at Week 52. The patient counts were small: nine to ten per arm in the recently diagnosed cohorts, with the long-duration cohort showing essentially no signal. Placebo patients, consistent with natural history, lost C-peptide over the year.

Two facts matter more than the point estimate. The effect was concentrated in one nine-patient cell, the recently diagnosed 200 mg arm. The confidence interval around any single-arm estimate from nine patients with a coefficient of variation near 50% spans roughly ±25 percentage points, which is wider than the reported placebo decline. The COVALENT-111 T2D readout, by contrast, was supported by roughly 200 patients across arms.

Failure mechanism, best guess

The most likely explanation is statistical, not biological. A nine-patient cohort cannot rule out a placebo-by-chance effect of 50% magnitude. Combined with the reputational cost of the 2024 LFT hold, the board saw no path to a financeable Phase 3 in T1D without a substantially larger replication trial.

A second possibility is target engagement uncertainty in islets. In KMT2A-rearranged leukemia, menin inhibition produces a transcriptional response measurable in peripheral blood. In adult human islets the relevant readout is beta-cell proliferation, and no validated peripheral biomarker exists. The 12-week dosing window was a compromise forced by the LFT signal, not by pharmacodynamic modeling. If menin must be held below a functional threshold for longer than 12 weeks to drive durable beta-cell expansion, the trial pre-committed to missing the effect.

A third factor is competitive crowding. Teplizumab is approved for stage 3 delay, and several anti-CD3 and anti-IL-21 programs are entering Phase 2. The bar for an oral, beta-cell-directed agent has risen, and a 37-patient trial with portable LFT history was unlikely to clear it.

What this means for similar programs

Other beta-cell-regenerative programs should treat COVALENT-112 as a powering case study, not a target-validity case study. Vertex's VX-880 stem-cell-derived islet program continues to produce insulin independence in small cohorts, and Sernova's Cell Pouch is producing similar signals. Small-molecule programs targeting DYRK1A (Wang et al., 2020) are moving toward proof-of-concept. Each should size its Phase 2 against placebo C-peptide variance rather than a hoped-for effect.

How to prevent this next time

Size the Phase 2 against the placebo arm's natural-history variance, not against a target effect. A simulation built on C-peptide AUC variance in the TrialNet natural history dataset, with σ ≈ 0.30 nmol/L and an assumed treatment effect of +25% relative to baseline, requires roughly 128 patients to achieve 80% power at α = 0.05 two-sided. Biomea's 37-patient enrollment delivered roughly 32% power, below the 80% standard and even below the 50% futility floor. A Bayesian interim analysis would have returned a predictive probability of success of

PP(success)=∫θ​P(Phase 3 success∣θ,interim data)π(θ∣interim data)dθ

below 0.20, well under the typical 0.30 portfolio threshold.

The second lesson is biomarker enrichment. Beta-cell function in recent-onset T1D is heterogeneous, and autoantibody-positive subsets retain 5- to 20-fold more residual function than longer-diagnosed subsets. A prospective enrichment design requiring stimulated C-peptide AUC above 0.20 nmol/L plus single-positive GAD or IA-2 antibody status would have increased the absolute effect size and shrunk the needed enrollment by roughly one third. Applying a hierarchical Bayesian prior from BioMedTracker that gives endocrine disease-modifying agents a Phase 2 to approval rate of about 12%, with a downward adjustment for prior LFT clinical holds, the prior probability of a successful Phase 3 commit at termination was approximately 0.14, below the 0.30 portfolio threshold. The single most actionable change would have been pre-specifying a recent-onset, biomarker-enriched 120-patient Phase 2 from the start, rather than a three-cohort exploration that the LFT hold truncated to 37.

Open questions

• Did the 200 mg recently diagnosed cohort sustain its Week 52 C-peptide gain in a per-patient repeated-measures analysis, or did the cohort mean carry outliers?
• Is there evidence of on-target peripheral pharmacodynamic engagement, for example through HOXA9 or MEIS1 expression changes in circulating leukocytes, that would license a longer dosing window?
• Could a non-covalent menin inhibitor avoid the LFT signal that capped icovamenib's exposure window in adult-onset T1D?
• What is the predictive probability of success for the parallel COVALENT-111 T2D program after the T1D termination, and does the failure here change Biomea's portfolio prior on menin biology in metabolism?

Sources

1. Karnik SK, Chen H, McLean GW, Heit JJ, Gu X, Zhang AY, et al.. Menin controls growth of pancreatic beta-cells in pregnant mice and promotes gestational diabetes mellitus. Science. 2007;318(5851);806-809. PMID: 17975067.

2. Li F, Liu R, Negi V, Yang P, Lee J, Jagannathan R, et al.. VGLL4 and MENIN function as TEAD1 corepressors to block pancreatic β cell proliferation. Cell Rep. 2023;42(1);111904. PMID: 36662616.

3. Kong X, Yang C, Li B, Yan D, Yang Y, Cao C, et al.. FXR/Menin-mediated epigenetic regulation of E2F3 expression controls β-cell proliferation and is increased in islets from diabetic GK rats after RYGB. Biochim Biophys Acta Mol Basis Dis. 2024;1870(5);167136. PMID: 38531483.

4. Issa GC, Aldoss I, DiPersio J, Cuglievan B, Stone R, Arellano M, et al.. The menin inhibitor revumenib in KMT2A-rearranged or NPM1-mutant leukaemia. Nature. 2023;615(7954);920-924. PMID: 36922593.

5. Chamberlain CE, Scheel DW, McGlynn K, Kim H, Miyatsuka T, Wang J, et al.. Menin determines K-RAS proliferative outputs in endocrine cells. J Clin Invest. 2014;124(9);4093-4101. PMID: 25133424.

6. Herold KC, Bundy BN, Long SA, Bluestone JA, DiMeglio LA, Dufort MJ, et al.. An Anti-CD3 Antibody, Teplizumab, in Relatives at Risk for Type 1 Diabetes. N Engl J Med. 2019;381(7);603-613. PMID: 31180194.

7. Hewitt JK. Twin studies of brain, cognition, and behavior. Neurosci Biobehav Rev. 2020;115;1-4. PMID: 32433925.

8. ClinicalTrials.gov. Phase 2 Trial of BMF-219 in Participants With Type 1 Diabetes Mellitus, NCT06152042. https://clinicaltrials.gov/study/NCT06152042.