Series: What Early Human Data Tells Us
Early immune-targeting drug programmes often reach a deceptively simple question before they have any chance of demonstrating clinical efficacy: an inflammatory signal has moved, but what does that movement mean? Is it evidence of meaningful target engagement, a nonspecific pharmacodynamic effect, background variability, or simply an endpoint that is sensitive in a particular experimental setting?
That question reflects a persistent translational challenge. Animal models, cell systems, and ex vivo assays remain essential for understanding biology and selecting candidate mechanisms, but they cannot fully predict the timing, magnitude, or coordination of human immune responses. Many immunology programmes therefore enter early clinical development with a mechanistic hypothesis that is biologically plausible, but not yet sufficiently anchored in human pharmacology.
Controlled human challenge models help close that gap. In healthy participants, a defined and transient inflammatory response can be induced with tools such as lipopolysaccharide (LPS), keyhole limpet hemocyanin (KLH), or imiquimod. These models do not reproduce chronic inflammatory disease, nor are they intended to. Instead, they create a controlled human setting in which drug developers can ask whether a compound engages the intended pathway, which downstream readouts are affected, and whether the observed pattern is large and coherent enough to support the next development decision.
This article is the second in the series What Early Human Data Tells Us. The first article discussed why benchmarking inflammatory models matters. This article takes the next step: how early inflammatory signals should be interpreted once they are generated.
Challenge Models Are Mechanistic Tools, Not Disease Models
The distinction between a challenge model and a disease model is critical. A disease model implies an attempt to reproduce the full clinical condition, including its chronicity, tissue remodelling, patient heterogeneity, comorbidities, and prior treatment effects. A challenge model does not attempt to do that. Instead, it isolates a tractable component of immune biology and makes it measurable under controlled conditions.
LPS is well suited to probing acute innate TLR4-mediated inflammation, either systemically or locally in the skin. KLH can be used to evaluate adaptive, T-cell-dependent immune responses. Imiquimod can help characterise TLR7-driven skin inflammation. Each model asks a different mechanistic question, so selecting the right model is part of the translational strategy rather than an operational detail.
This framing prevents two common errors. The first is overclaiming: interpreting a challenge response as proof that the compound will work in patients. The second is underusing the model: treating it as a simple yes-or-no pharmacodynamic screen. The real value lies between those extremes. A well-selected challenge model can show whether a human pathway is engaged and whether the pattern of response is consistent with the compound's intended mechanism.
Why Benchmarks Change the Interpretation
A single inflammatory biomarker rarely speaks for itself. Cytokines are dynamic and time-dependent. Cell recruitment may peak later than soluble mediators. Skin perfusion, erythema, fever, heart rate, and cellular activation each reflect different levels of the inflammatory cascade. A visible or convenient endpoint may not be the endpoint most closely linked to the compound's mechanism.
Benchmarking gives these measurements a reference frame. When a marketed anti-inflammatory drug is tested in the same model, it shows what a known human pharmacological effect looks like across endpoints. When several compounds with different mechanisms are tested, the model begins to acquire a response map: which endpoints are sensitive, which are variable, and which readouts distinguish one mechanism from another. Beyond benchmarking against registered compounds, combining multiple complementary readouts within the same immune challenge model is equally strategic, as a multimodal characterization of the inflammatory response provides mechanistic insight into how a novel compound modulates the immune cascade and helps identify pharmacodynamic effects that may not be apparent from any single endpoint.
For drug developers, that map can separate an interesting experiment from a decision-enabling study. A novel compound may produce a modest cytokine shift that seems unimpressive in isolation but is meaningful when it resembles the response pattern of a benchmark drug. Conversely, a large movement in one biomarker may be less persuasive if it does not align with the expected pathway or is highly variable across previous studies.
Two Complementary Examples
One way to build this reference frame is back-translation. Instead of using a challenge model only to test a novel compound, approved immunomodulatory therapies can be studied under controlled challenge conditions. Because their clinical activity is already established, the question becomes: how does known therapeutic pharmacology appear in a human challenge model?
A recent study1 with marketed immunomodulators used this principle to evaluate how registered drugs modulate immune responses in humans. This was more than a positive-control exercise. Different therapeutic classes are expected to influence different parts of the cascade: broad inflammatory suppression, cytokine production, cellular recruitment, adaptive immune activation, or tissue-level readouts. Observing those differences in a controlled model helps define which endpoints are mechanistically informative and which are less sensitive for a given pathway.
That back-translation approach strengthens interpretation in two ways. First, it validates the model by showing that drugs with established clinical activity produce recognisable and biologically plausible pharmacodynamic patterns. Second, it generates new mechanistic insight into how those drugs behave in humans under controlled inflammatory stress.
The same logic can be applied prospectively to an investigational medicinal product. The study discussed in the first article in this series2 used intradermal and intravenous LPS challenges to evaluate p38 MAPK inhibition in healthy volunteers. Rather than repeating the full case study here, the key interpretive point is that the investigational compound produced a mixed but coherent pharmacodynamic pattern: reductions in pathway-relevant endpoints such as cytokine responses, immune-cell attraction, target-cell phosphorylation, and systemic vital-sign effects, while some local visible skin responses were less affected.
That pattern illustrates why benchmarking matters. Without a reference frame, unchanged erythema or perfusion might be read as a weak result. In a benchmarked multimodal model, it can instead help define where the compound is pharmacologically active and which endpoints are less informative for that mechanism. The conclusion is not that the compound has demonstrated patient benefit, but that the intended biology has been engaged in humans in a way that can inform dose, biomarker, and next-study choices.
From Target Engagement To Development Decisions
The strongest early immune studies are designed backwards from the decisions they need to support. Before the study begins, teams should ask what evidence would justify moving forward, what result would stop or redirect the programme, and which endpoints would be credible enough to support dose selection or patient-study design.
A benchmarked challenge model can support those decisions only if the model, dose range, sampling schedule, and biomarker panel are aligned with the compound's mechanism. For a drug expected to suppress acute innate cytokine release, an LPS model may be appropriate. For a compound targeting adaptive immune activation, KLH may provide a better test. For a therapy intended to modulate inflammatory skin pathways, an imiquimod or other skin challenge may be more informative.
This matters because patient studies are often noisy and slow. Disease activity fluctuates, prior therapies influence immune pathways, patient heterogeneity can obscure pharmacodynamic effects, and enrolment is often more difficult than expected. A challenge study cannot remove the need for patient trials, but it can make the step into those trials more rational by clarifying dose, endpoint selection, timing, and biological plausibility.
What These Data Tell Us
The main lesson for development teams is that early inflammatory signals should not be interpreted in isolation. A biomarker change becomes more meaningful when it is understood against a benchmarked human response map. That map helps distinguish true target engagement from assay noise, background variability, and endpoint-specific sensitivity.
The second lesson is that absence of movement in one endpoint is not always failure. In inflammatory biology, different readouts capture different levels of the cascade. A compound may affect cytokine production without visibly changing erythema, or modulate cellular recruitment without altering every systemic marker. Interpreting that pattern requires prior knowledge of how benchmark compounds behave in the same model.
The third lesson is that challenge models should be used with discipline. They should not be asked to prove clinical efficacy. Their role is to generate decision-quality human pharmacology: evidence that the intended pathway can be engaged, that the selected endpoints are fit for purpose, and that the next study can be designed with fewer assumptions.
As immune-targeted pipelines become more mechanism-specific, the need for this kind of human evidence will increase. Preclinical systems will remain essential, but they cannot fully substitute for controlled early clinical pharmacology. Benchmarked human challenge models provide the bridge, not by pretending to be diseases, but by making human immune modulation interpretable before large patient studies begin.
References
1. Assil, S., Buters, T. P., Hameeteman, P. W., Hallard, C., Treijtel, N., Niemeyer-Van der Kolk, T., de Kam, M. L., Florencia, E. F. I. I. I., Prens, E. P., van Doorn, M. B. A., Rissmann, R., Klarenbeek, N. B., Jansen, M. A. A., & Moerland, M. (2023). Oral prednisolone suppresses skin inflammation in a healthy volunteer imiquimod challenge model. Frontiers in immunology, 14, 1197650. https://doi.org/10.3389/fimmu.2023.1197650
2. de Bruin, D. T., Jansen, M. A. A., Pereira, D. R., van Schijndel, L., Klaassen, E. S., Otto, M. E., Skillington, J., Maguire, P., Tremble, L., Bell, A., Maher, L., Mihara, K., Sumeray, M., Gilroy, D. W., Klarenbeek, N. B., & Moerland, M. (2026). POLB 001, a p38 MAPK inhibitor, decreases local and systemic inflammatory responses following in vivo LPS administration in healthy volunteers: a randomised, double-blind, placebo-controlled study. Frontiers in immunology, 16, 1684307. https://doi.org/10.3389/fimmu.2025.1684307
