BGB-45035, an oral IRAK4 degrader, was terminated in rheumatoid arthritis for insufficient benefit

What was tried

BeiGene ran NCT07100938, a Phase 2 randomized, double-blind, placebo-controlled study of BGB-45035 in adults with moderate to severe active rheumatoid arthritis. BGB-45035 is an orally administered IRAK4 degrader. Patients were randomized to BGB-45035 or placebo, with participants and investigators masked. The primary endpoint was change from baseline in the Disease Activity Score using C-reactive protein, DAS28-CRP, at week 12. Actual enrollment was 49 participants and the actual primary completion date was May 12, 2026 (ClinicalTrials.gov NCT07100938). The trial was terminated. The posted reason was that the sponsor concluded patients were unlikely to derive sufficient benefit from BGB-45035 to support continued development given an unfavored benefit to risk profile (ClinicalTrials.gov NCT07100938).

The biological hypothesis

IRAK4 sits directly downstream of most Toll-like receptors and the interleukin-1 receptor family. After ligand binding, MyD88 nucleates a signalosome in which IRAK4 acts first, both as a kinase and as a scaffold, to activate IRAK1 and IRAK2 and drive NF-kB and MAP kinase signaling. The pathway feeds production of IL-6, TNF, and other cytokines central to synovial inflammation. Genetic and pharmacologic work supports the node. Kinase-inactive IRAK4 knock-in mice show reduced inflammatory responses, and IRAK4 inhibitors lowered inflammatory readouts in rheumatoid arthritis models (Umar 2021). The degrader logic goes one step further than enzyme inhibition. Removing the whole protein eliminates the scaffold function that a kinase inhibitor leaves intact, which in principle should suppress signaling more completely. The same rationale drove KT-474, an IRAK4 degrader that lowered target protein and inflammatory markers in early studies (Ackerman 2023, Zheng 2024).

What actually happened

The study stopped before delivering a positive controlled efficacy readout, and no results were posted at termination (ClinicalTrials.gov NCT07100938). The sponsor framed the decision around benefit to risk rather than a safety halt. The pattern is familiar for this target. The IRAK4 kinase inhibitor PF-06650833, zimlovisertib, blocked inflammatory signaling and showed activity in rheumatoid arthritis, yet its clinical effect sat below what established agents deliver (Winkler 2021). A contemporaneous review of investigational IRAK4 inhibitors reached a similar reading, that target engagement was achievable but differentiation from existing therapy was not clear (Wiese 2020). Open Targets scores the IRAK4 to rheumatoid arthritis association at 0.0992 overall, with the supporting evidence concentrated in literature (0.210) and clinical data (0.153) rather than human genetics (Open Targets Platform).

Failure mechanism, best guess

The most consistent explanation is redundancy rather than insufficient target engagement. Rheumatoid arthritis synovitis is sustained by several parallel cytokine circuits, and blocking the TLR and IL-1R arm through IRAK4 leaves IL-6, TNF, JAK-dependent, and T cell driven signaling largely in place. A degrader can drive IRAK4 protein close to absent and still produce only a partial clinical effect if the disease does not depend heavily on that single node. The 49 patient size also matters. A study this small was built to detect a clear effect, not a marginal one, and a benefit that lands inside the range already covered by approved drugs would not justify a larger program. This reads as efficacy limited by pathway redundancy, consistent with the modest controlled results seen for IRAK4 kinase inhibition in the same indication (Winkler 2021).

How to prevent this next time

Two quantitative tools would have sharpened the decision earlier. First, an explicit power calculation against a clinically relevant bar. With roughly 24 patients per arm, a two sample comparison at alpha 0.05 has about 41 percent power to detect a standardized DAS28-CRP difference of 0.5 and about 55 percent power at 0.6 (computed by normal approximation). A study powered to rule a drug in but not to rule it out leaves the benefit to risk question unresolved, so pre-registering the minimum effect worth pursuing converts an ambiguous stop into a clean one. Second, a Bayesian posterior probability of success that updates a prior built from the IRAK4 class. The class prior should be skeptical, because the Phase 2 to Phase 3 transition rate across indications is near 28.9 percent (Wong 2019) and prior IRAK4 inhibitor data in rheumatoid arthritis were modest (Winkler 2021). Formally,

PP(success)=∫θ​P(final success∣θ,interim data)⋅π(θ∣interim data)dθ

Pairing a defined minimum effect with a class-anchored posterior would have set a clear continuation threshold before the first patient enrolled. The single highest leverage change would have been ...

What this means for similar programs

The IRAK4 degrader thesis is not refuted by one stop, but the burden of proof shifts. Programs in rheumatoid arthritis should plan to show separation from approved agents, not just placebo, because the bar is set by what patients already receive. The cleaner opportunities for IRAK4 degradation may lie in diseases where the TLR and IL-1R axis is closer to the dominant driver, such as MyD88 dependent conditions, rather than a redundant indication like rheumatoid arthritis. The contrast with KT-474, which advanced in hidradenitis suppurativa and atopic dermatitis (Ackerman 2023), reinforces the point that indication choice, not molecule quality, may decide these programs.

Open questions

Did BGB-45035 achieve deep IRAK4 degradation in patients, and if so what was the dose response on DAS28-CRP? Was the benefit to risk concern driven by weak efficacy alone or by a tolerability finding not captured in the public record? Would an enriched population, selected for high TLR pathway activity, have produced a different signal? The answers depend on data BeiGene has not released.

Sources

1. ClinicalTrials.gov, NCT07100938, BGB-45035 Phase 2 in rheumatoid arthritis, enrollment 49, terminated, primary completion 2026-05-12. Open Targets Platform, IRAK4 to rheumatoid arthritis association 0.0992 (literature 0.210, clinical 0.153). Winkler A, et al. Arthritis Rheumatol 2021,Wiese MD, et al. Expert Opin Investig Drugs 2020,Ackerman L, et al. Nat Med 2023,Zheng X, et al. J Med Chem 2024,Umar S, et al. Life Sci 2021,Wong CH, Siah KW, Lo AW. Biostatistics 2019,FAERS queried for BGB-45035, no records (investigational compound). Power values computed by normal approximation. Available from: https://clinicaltrials.gov/study/NCT07100938.