INB03 in COVID-19 ARDS: Why a Selective Soluble TNF Strategy Did Not Separate From Placebo

What was tried

INmune Bio ran NCT04370236, a multicenter, randomized, quadruple-blind, placebo-controlled Phase 2/3 study of INB03 in hospitalized COVID-19 patients with pulmonary complications. INB03 is the development name for pegipanermin, also known as XPro1595, a dominant-negative TNF fusion protein that forms inactive heterotrimers with soluble TNF (sTNF) while sparing transmembrane TNF (tmTNF). Patients received up to two weekly 1 mg/kg subcutaneous doses on top of standard of care. The primary endpoint was a Cochran-Mantel-Haenszel analysis of disease progression at day 28. Enrollment began in October 2020 and the trial closed in 2022 after 79 of a much larger planned cohort, with the public termination reason listed simply as "Futility."

The biological hypothesis

TNF signals through two receptors with divergent biology. TNFR1 is broadly expressed, binds sTNF preferentially, and drives the apoptotic and proinflammatory branch implicated in cytokine release, endothelial activation, and acute lung injury. TNFR2 is more restricted, binds tmTNF preferentially, and supports regulatory T cell survival, neuronal survival, and tissue resolution. Pan-TNF blockers like infliximab and adalimumab neutralize both species and carry well-documented safety baggage including opportunistic infection and reactivated tuberculosis.

The XPro1595 thesis is that selective neutralization of sTNF can dampen TNFR1-driven inflammation without crippling TNFR2-driven repair. In animal models of neurodegeneration, multiple sclerosis, and Huntington's disease, peripheral XPro1595 modifies microglial phenotype and reduces neuroinflammatory signaling without the systemic immunosuppression of pan-TNF inhibition [1,3,4]. The COVID-19 program extrapolated this logic to the lung: in severe COVID-19 the macrophage-driven TNF/IL-6/IL-1 cascade is a major contributor to alveolar damage, and a sTNF-only blockade was framed as a way to break that loop while preserving alveolar repair signaling.

What actually happened

The trial closed early for futility at the prespecified interim. With only 79 patients randomized to a 1:1 design, the Cochran-Mantel-Haenszel comparison of day-28 disease progression did not separate INB03 from placebo to a degree that justified continuation. No safety signal was given as the stopping reason and no separate efficacy readout has been published in a peer-reviewed venue at the time of writing. The conduct of the study spanned the period during which dexamethasone became standard of care and IL-6 blockade (tocilizumab) and IL-1 blockade (anakinra) accumulated positive RCT data. The trial was running into a moving floor.

Failure mechanism, best guess

Three forces likely combined to produce a futile signal.

First, target selection. TNF is a minor node in the COVID-19 cytokine network compared to IL-6, IL-1, and the IFN/JAK axis. OpenTargets returns no TNF-COVID-19 association in the top 200 of TNF disease links, where the strong signals belong to psoriatic arthritis, rheumatoid arthritis, and inflammatory bowel disease (scores 0.4 to 0.7). The downstream cytokine RCTs that worked in COVID-19 hit IL-6 and IL-1, not TNF. Plasma TNF elevation in COVID-19 ARDS is real but modest compared to IL-6 elevation, and TNF is not the rate-limiting cytokine for alveolar injury once the cascade is active.

Second, dose and exposure. Up to two weekly subcutaneous doses at 1 mg/kg is a conservative regimen for a large protein in an acute inflammatory setting where sTNF is being generated continuously by activated macrophages. The pharmacodynamic question of whether INB03 achieved sustained sTNF neutralization in lung tissue across the relevant 7 to 14 day window was never answered with a public biomarker. Selective inhibition only works if free sTNF is meaningfully suppressed where it matters.

The resolution argument cuts both ways. By the time a hospitalized COVID-19 patient enters a trial, tmTNF/TNFR2 signaling on alveolar macrophages and Tregs may already be a desirable feature, not a backstop. Sparing it is not the same as enhancing it, and the active driver of progression at that point may be IL-6 amplification and microvascular thrombosis rather than residual sTNF.

Third, the endpoint. A composite day-28 disease progression endpoint in a population that improves quickly with steroids and oxygen leaves little room for separation, and a sample of 79 is severely underpowered to detect anything short of a large effect. The probability of detecting a clinically meaningful effect was low before the trial started.

How to prevent this next time

The Bayesian framing makes the problem explicit. The posterior probability of trial success is

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

A program should refuse to enter a Phase 2/3 in an acute, fast-moving indication until the prior on the relevant cytokine being a rate-limiting driver crosses a defensible threshold. Two changes would have raised that prior, or killed the program earlier and cheaper.

First, a biomarker-enrichment strategy. Stratify enrollment by baseline soluble TNF and TNFR1, and require evidence of sustained sTNF suppression in serum and ideally bronchoalveolar fluid within the first 72 hours of dosing. If pharmacodynamic engagement is not visible, the efficacy question is moot. A formal number-needed-to-screen calculation against the expected biomarker prevalence would have surfaced the feasibility problem in advance.

Second, competitive landscape red-teaming. By Q4 2020 it was foreseeable that dexamethasone, tocilizumab, baricitinib, and anakinra would compress the placebo arm and shrink the addressable signal. A red-team analysis would have priced the moving standard of care into the power calculation and pushed the program toward a narrower, biomarker-defined hyperinflammation subgroup or away from COVID-19 entirely toward neurodegeneration where the mechanistic rationale is stronger.

The single highest leverage change would have been a mandatory pharmacodynamic gate showing target engagement and sustained sTNF suppression in lung-relevant compartments before opening Phase 2/3 enrollment.

What this means for similar programs

Selective cytokine biologics succeed when the cytokine is a primary driver and when sparing the other arm of the receptor system delivers a clear therapeutic margin. COVID-19 ARDS satisfied neither condition. For sTNF-selective biologics the more defensible indication categories remain CNS programs (Alzheimer's, Parkinson's, treatment-resistant depression, ALS) where TNFR2-sparing biology matters and where the cytokine has direct mechanistic support in human genetics and lesion biology. INmune Bio's continued development of pegipanermin in Alzheimer's is the more defensible bet on the same molecule.

Open questions

Did INB03 achieve quantifiable sTNF suppression in any compartment during the trial. Was the futility signal driven by no separation in the placebo arm (standard of care saturation) or by absence of pharmacodynamic effect. Were any subgroups, such as patients with baseline elevated TNFR1, preserved as exploratory analyses. What was the actual day-28 progression rate in each arm.

Sources

1. [1] Maguire-Zeiss KA et al. Peripheral administration of the soluble TNF inhibitor XPro1595 modifies brain immune cell profiles. Neurobiology of Disease, 2017. [2] Hsu CW et al. Dominant negative biologics normalise the TNF-α induced angiogenic phenotype. BMC Immunology, 2023. [3] Hsu LJ et al. Lack of neuroprotection after systemic administration of the soluble TNF inhibitor XPro1595. Neurobiology of Disease, 2025. [4] Hsiao HY et al. Inhibition of soluble tumor necrosis factor is therapeutic in Huntington's disease. Human Molecular Genetics, 2014. [5] ClinicalTrials.gov. NCT04370236. Multicenter, Double-Blind, Randomized, Placebo-Controlled Trial of INB03 in Pulmonary Complications From COVID-19. INmune Bio. Terminated, last update 2026. [6] Open Targets Platform. TNF (ENSG00000232810) target dossier and disease associations. Available from: https://clinicaltrials.gov/study/NCT04370236.