Human immune challenge to assess proof-of-mechanism in drug development

Background: High attrition rates in immunomodulatory drug development, particularly in Phase II clinical trials, are largely driven by failure to demonstrate efficacy at this key development stage. First-principles and observational evidence indicate that establishing proof of mechanism (PoM) early is essential to support robust go/no-go decisions. Human immune challenge (HIC) models provide an experimental platform to elicit mechanism-relevant pharmacodynamic (PD) biomarkers in healthy volunteers, thereby bridging the translational gap between preclinical evidence and Phase I proof of concept (PoC). However, existing HIC models—including those employing the widely used challenge antigen keyhole limpet haemocyanin (KLH)—are characterised by substantial methodological heterogeneity, limited standardisation, and narrowly restricted outcome measures. In parallel, despite rapid growth in the number of druggable immunological targets and clinical-stage investigational medicinal products (IMPs), there remains a lack of validated HIC paradigms capable of eliciting many target pathways in healthy volunteers, precluding mechanism-relevant PD interrogation for a large proportion of immunomodulatory IMPs in Phase I. Collectively, these limitations constrain the feasibility and decision-making value of HIC-enabled PoM studies in contemporary drug development.This thesis tests the hypothesis that systematic optimisation of challenge agent, dose, adjuvant, tissue, and assessment timing enables HIC models to generate robust, mechanism-specific PD readouts in healthy volunteers. It further hypothesises that formal integration of these readouts within a Bayesian decision-theoretic framework yields more informative and economically optimal go/no-go decisions than conventional Phase I development strategies that either lack mechanism-relevant PD assessment or incorporate such evidence informally.Methods: To address these aims, a multi-modal experimental and analytical research programme was conducted:1. An observational study of modified vaccinia Ankara (MVA) vaccination to characterise immunogenicity and assess its potential as a viral HIC model relevant to PD evaluation of immunomodulatory agents targeting innate viral signalling, T-helper 1 (Th1), and CD8+ T-cell pathways.2. A systematic review of 46 studies employing KLH challenge in drug development to identify design limitations and opportunities for model optimisation.3. Two randomised, single-blind HIC studies (KLH1 and KLH2) designed to optimise the KLH model, including evaluation of adjuvants (Alhydrogel vs. Montanide), Bayesian modelling of the intradermal KLH rechallenge dose–response relationship using Emax models, and characterisation of the temporal evolution of cutaneous immune responses via multiparameter flow cytometry.4. A pilot study evaluating contrast-enhanced ultrasound (CEUS) to improve identification of primary draining lymph nodes—a site of increasing interest for invasive sampling to enable tissue-based readouts of HIC and PD activity.5. A Bayesian decision-theoretic simulation examining how HIC-derived PD data can be formally integrated with economic utility to optimise go/no-go decision thresholds.Results: The MVA-BN study demonstrated the feasibility of MVA-BN as a viral HIC agent, subject to further optimisation. MVA-BN elicited robust Th1-biased and CD8+ T-cell responses; however, substantial inter-individual variability was observed in serological and cellular responses measured in blood. These findings indicate a need for further characterisation of statistical operating characteristics, dose–response relationships, and endpoint selection, particularly with respect to immune responses at the site of intradermal administration. In the KLH optimisation studies, Montanide elicited significantly greater systemic T-cell responses (IFNγ and IL-4) than Alhydrogel, despite comparable serological effects, with both adjuvants demonstrating greater immunogenicity than unadjuvanted subunit KLH. Bayesian modelling identified 10 μg as the optimal intradermal rechallenge dose, balancing response consistency with expected modulation-sensitivity. Temporal analyses revealed a shift in the cutaneous immune infiltrate from a mixed myeloid (classical monocyte)–lymphoid phenotype at 48 hours to a Th1-polarised response with enrichment of intermediate and non-classical monocytes at later timepoints (Days 5 and 14). The CEUS study was terminated early for futility due to failure to reliably identify target lymph node enhancement, demonstrating the utility of adaptive experimental designs in efficiently evaluating feasibility in experimental medicine studies. Finally, the decision-theoretic simulation showed that conventional frequentist significance thresholds are likely suboptimal from an economic perspective, supporting the use of utility-calibrated probabilistic decision rules to guide go/no-go decisions.Conclusions: This thesis delivers a standardised and optimised protocol for the KLH HIC model and provides PoC evidence for MVA-BN as a promising viral HIC agent capable of eliciting innate and adaptive antiviral immune pathways. It demonstrates that assessment timepoints must be tailored to the biomarker and mechanism of interest, with early timepoints (e.g., 48 hours) optimally capturing innate and myeloid responses and later timepoints (Days 5–14) required to interrogate adaptive T-cell modulation and local tissue adaptation in KLH HIC. By integrating optimised experimental platforms with rigorous Bayesian decision-theoretic analysis, this work advances the feasibility and decision-making value of HIC-enabled Phase I PoM assessment in immunomodulatory drug development.