A non-depleting anti-CD19 antibody in rheumatoid arthritis: the LY3541860 interim futility

What was tried

NCT06859294 was a Phase 2a, single-arm, open-label study run by Eli Lilly to test LY3541860 in adults with moderately to severely active rheumatoid arthritis (RA). LY3541860 is a non-B-cell-depleting anti-CD19 IgG4 monoclonal antibody, classed as a CD19 antigen modulator (AdisInsight drug profile, LY3541860). The trial enrolled 35 participants and used the disease activity score in 28 joints with high-sensitivity CRP (DAS28-CRP) as its primary measure (ClinicalTrials.gov, NCT06859294). Lilly terminated the study, recording that the interim analysis data failed to demonstrate the intended treatment effect (ClinicalTrials.gov, NCT06859294).

The biological hypothesis

CD19 is a pan-B-cell surface protein expressed from early B-cell development through the plasmablast stage, which makes it broader than CD20 because it persists on some antibody-secreting cells. That persistence is the stated reason to prefer CD19 over CD20 when the aim is to reach the plasmablast pool that sustains autoantibody production. The thesis behind a non-depleting anti-CD19 antibody is that engaging CD19, rather than killing the cell, can dial down B-cell receptor signaling and autoantibody output while preserving the protective B-cell repertoire. RA has a documented B-cell component, and anti-CD20 depletion with rituximab is established therapy, particularly in seropositive disease (Taha et al., Open Access Rheumatol 2017). Open Targets scores the CD19 to RA association at 0.13, again driven by literature evidence (0.91) with little human genetic support (Open Targets Platform). The molecular target is credible. The open question the trial actually tested was whether occupancy without depletion is enough.

What actually happened

The interim analysis did not show the intended effect on disease activity, and the study was terminated at 35 enrolled participants (ClinicalTrials.gov, NCT06859294). No results are posted, and there are no openFDA FAERS records for this investigational molecule. The protocol used single-group assignment with no randomization and no masking, and the primary completion date was recorded as November 2025 (ClinicalTrials.gov, NCT06859294). The design itself limits what can be concluded, because a single-arm, unmasked study with no concurrent control measures change from baseline in DAS28-CRP, an endpoint that moves substantially with background therapy and expectation effects. An interim that fails to clear the intended bar is a strong negative signal in a setting where active agents usually look clearly active.

Failure mechanism, best guess

The most likely reading is an efficacy failure tied to mechanism depth rather than target choice. The therapies with clear B-cell-driven benefit in autoimmunity remove or deeply deplete the compartment. Rituximab depletes CD20 B cells and is established in RA (Taha et al., 2017). CD19 CAR-T cells, which achieve deep tissue-level B-cell clearance, drove DORIS remission in 9 of 10 lupus patients in the CASTLE basket trial and produced responses across systemic sclerosis and myositis (Mueller et al., Nat Med 2026). Even obexelimab, a non-depleting bifunctional antibody that binds CD19 and Fc-gamma-RIIb, only brings the B-cell count to roughly 50 percent of baseline despite near-complete CD19 receptor occupancy (Wang et al., Adv Ther 2024). If partial suppression is what non-depleting CD19 engagement delivers, then established RA, with its entrenched autoreactive plasmablast and synovial B-cell populations, is a demanding first test. The mechanism may simply not reduce the pathogenic B-cell output enough to register on DAS28-CRP.

How to prevent this next time

A predictive-probability gate would have sized the bet better. Formalizing the interim as a Bayesian decision rule makes the stop explicit:

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

With a prior built from depleting-modality benchmarks and a mechanism that delivers only partial suppression, the predictive probability of a positive controlled Phase 2 is low, which is the right reason to stop early rather than expand.

The design also needed a control arm and an honest power calculation. A randomized Phase 2 powered to detect an ACR20 difference of 20 percentage points, for example 60 percent versus 40 percent, at 80 percent power and two-sided alpha 0.05 needs about 94 evaluable patients per arm by the two-proportion formula. A 35-patient single arm cannot separate a drug effect from background response at that resolution, so the study was underpowered to demonstrate efficacy even if the drug worked.

The third change is a pharmacodynamic gate tied to the mechanism. Because the open question was depletion depth, the program should have required a prespecified pharmacodynamic threshold, such as a defined reduction in circulating and ideally synovial B cells or plasmablasts at full receptor occupancy, before advancing to an efficacy readout. The obexelimab data showing a 50 percent B-cell nadir at near-complete occupancy were available as a benchmark (Wang et al., 2024).

The single highest leverage change would have been to require a prespecified B-cell pharmacodynamic threshold as a gate before any efficacy stage, rather than testing occupancy and clinical response together in one underpowered single-arm step.

What this means for similar programs

The CD19 and CD20 autoimmunity field now spans a clear depth gradient, from non-depleting engagement through antibody depletion to CAR-T reset, and the clinical signal strengthens as depletion deepens (Wang et al., 2024, Mueller et al., Nat Med 2026, Taha et al., 2017). Non-depleting B-cell programs should pick indications and endpoints where partial modulation is plausibly sufficient, or pair the mechanism with a biomarker that proves the pathogenic compartment actually contracts. Established seropositive RA, where deep depletion sets the bar, is a hard place to debut a partial mechanism. Single-arm Phase 2 designs compound the risk in any indication with a large nonspecific response.

Open questions

Would LY3541860 show activity in a B-cell-shallow indication, or in early treatment-naive RA where the autoreactive compartment is less entrenched? Did the program measure B-cell or plasmablast pharmacodynamics, and if so did occupancy translate into measurable depletion? And is there a dose or schedule at which non-depleting CD19 engagement crosses into a functionally depleting effect?

Sources

1. ClinicalTrials.gov. A study of LY3541860 in adults with moderately to severely active rheumatoid arthritis, NCT06859294. https://clinicaltrials.gov/study/NCT06859294.

2. AdisInsight (Springer). LY3541860 drug profile, non-B-cell-depleting anti-CD19 IgG4 monoclonal antibody. https://adisinsight.springer.com/drugs/800065461.

3. Wang X, et al. Pharmacokinetics, pharmacodynamics, bioavailability, and immunogenicity of obexelimab following subcutaneous administration. Adv Ther 2024, 42:813-829.

4. Mueller F, et al. CD19 CAR-T cells for treatment-refractory autoimmune diseases, the phase 1/2 CASTLE basket trial. Nat Med 2026, 32:1142-1151.

5. Taha R, et al. Systematic review of the role of rituximab in rheumatological disease. Open Access Rheumatol 2017, 9:201-214.

6. Open Targets Platform. CD19 (ENSG00000177455), rheumatoid arthritis (EFO_0000685) association. https://platform.opentargets.org/.