Summary of the 1st Workshop
The goal of this 1½-day EUCRAF workshop was to illustrate the way in which specialist knowledge from different drug development disciplines may be effectively integrated to identify, evaluate and mitigate risks associated with undesirable immunogenicity of therapeutic proteins and peptides. Recognised as an increasingly important aspect of regulatory assessment of clinical benefit and risk for biopharmaceutical products, the challenge of blending the relevant competencies at pertinent stages of the development process remains a daunting task. Moreover, a front-end-loaded approach to minimise risks at the lead candidate selection stage needs to look forward to the ultimate clinical uses of the product, thereby prioritising the manner in which identified risks may be evaluated in pre-clinical and clinical studies to meet the expectations of regulatory agencies for the weight of evidence required to support investigational studies and eventual marketing authorisation.
Gabriele Dallmann (Study Director, EUCRAF) explained that the programme of the “Immunogenicity Integrated” workshop was exceptional in at least three respects: firstly, it featured highly knowledgeable experts from different disciplines – applied immunology, CMC, bioanalytical, non-clinical, PK, clinical and regulatory – in the immunogenicity field who are able to distil the most important messages in a manner that cross-refers to the respective stages of the biopharmaceutical development cycle; secondly, the subject material was orientated toward building, in an iterative fashion, an integrated presentation of relevant data package for submission to regulatory agencies; thirdly, the “how to” aspects of risk identification, evaluation and mitigation would be reinforced via an interactive case study that featured a “high risk” immune-modulatory therapeutic protein candidate that had been placed on “clinical hold” by a concerned regulatory agency: this provided a realistic test of how well the key learning points had been assimilated, in addition to illustrating the practical application of risk vs. benefit analysis in a multi-disciplinary setting. Each session included interactive panel discussions involving all of the participants – which allowed clarification of the elements to integrate.
Paul Chamberlain (NDA Advisory Board) set the scene by describing how the iterative approach to immunogenicity risk assessment could be aligned to successive stage gates in the product development cycle. There was an emphatic recommendation to initiate thinking, on a multi-disciplinary team basis, prior to selecting a candidate for development. Molecular features that could contribute intrinsic immunogenic stimuli should be considered in the context of the relevant biological system: this would enable identification of risk factors that might influence choice of lead candidate, as well as signal factors that might require focussed evaluation in non-clinical and clinical studies. Examples of the influence of choice of host cell substrate were provided, and the relationship to justification of strategy for process and product quality control explained by case examples. The scenario of biosimilar monoclonal antibodies was reviewed in the context of explaining the limitations of a one-dimensional reliance on anti-drug antibody monitoring, as opposed to a recommended approach to correlate measures of immune responses to PK, PD, efficacy and safety parameters respectively in order to understand the clinical impact of bioanalytical indices of the immune response. Submission of an integrated summary of immunogenicity in support of both clinical trial and marketing authorisation applications could prove critical to a successful outcome, particularly where overall clinical benefit and risk was equivocal; preparation of an outline version could be initiated very early in the development cycle to guide strategy for monitoring immunogenicity, and also for justifying product quality control strategy, during pre-clinical and clinical phases; the concluding part of the integrated summary of immunogenicity should link to the Risk Management Plan, including recommendations for product labelling. This integrated approach would help to regulatory reviewers to overcome the obstacle of a fragmented distribution of the pertinent data across different parts of the dossier.
Matthew Baker (Antitope) provided a mechanistic immunologic perspective that made clear distinctions between adaptive and innate immune responses, while illustrating the points of interaction between these. The role of T-cell-dependent mechanisms in driving clinically meaningful immune responses to therapeutic proteins was described, while recognising potential roles for T-cell independent immune stimulation. Despite the complexity, and our incomplete knowledge, of the immune response mechanisms, the relative immunogenic potential of therapeutic proteins can be explained, at least in part, by molecular features of the protein that are recognised by different elements of the host immune repertoire. Accordingly, knowledge of intrinsic immunogenic potential generated using in silico modelling coupled to in vitro and/or ex-vivo cellular assays, may contribute to important decisions regarding molecular design of a new therapeutic protein candidate. Results of recent studies of uptake by, and activation of, dendritic cells by product aggregates were described to illustrate the relationship of the product quality control dimension to immunogenicity-related risks. There is an expanding toolbox for evaluation of risk factors associated with intrinsic (e.g. T-cell epitopes) and extrinsic (e.g. product-related aggregates) immunogenic stimuli – these can be applied very early in the development cycle to minimise risks for later development. There remains a gap regarding availability of in vivo models for prediction of immunogenicity; although there is much ongoing development work in this area, transgenic / immune-deficient animals have frequently demonstrated inadequate sensitivity for reliable modelling human immune responses.
Arno Kromminga (IPM Biotech) described critical challenges and solutions for detection of anti-drug antibodies (ADA), most importantly: cut-point determination, drug interference and functionality of ADA. In emphasising that an optimal bioanalytical method does not exist, he indicated relative strengths and weaknesses of the different assay formats that may be applied according to the product-specific considerations as well as the clinical setting. Given the spectrum of severity of clinical consequences of IgE-mediated hypersensitivity reactions, there is a potential role of detection of drug-specific IgE in certain circumstances: follow-up investigation of relevant clinical manifestations; increased probability of hypersensitivity reactions due to oral, subcutaneous or inhalative administration; immunological or genetic predisposition to develop allergic reactions. From the regulatory perspective, neutralizing ADA’s are generally of high concern; depending on the primary amino acid sequence, treatment-emergent ADA’s against therapeutic monoclonal antibodies (mAb’s) may be directed to anti-idiotypic and/or anti-isotypic domains; and may be neutralizing, depending on the mechanism of action of the mAb. For these reasons, ADA monitoring allied to measurement of drug levels should be considered for all patients treated with biological products to enable a better understanding of longer-term impact of undesirable immunogenicity on treatment outcomes. However, great caution should be applied to avoid incorrect interpretation of the results of ADA testing: incidence of ADA detection does not imply clinical significance; ADA results should always be considered in the context of clinical indices of disease status and loss of efficacy; epidemiologic factors and bioanalytical bias must also be taken into account. A good example of the limitation imposed by low prevalence of a clinical outcome is the rarity of cases of Pure Red Cell Aplasia (PRCA) associated with the use of different recombinant erythropoietin products: even where diagnostic sensitivity is 100% and diagnostic specificity is 95%, the low prevalence of PRCA implies a very low (estimated at < 0.001%) positive predictive value of the detection of anti-EPO antibodies; thus, a positive anti-EPO antibody assay result would not necessarily correspond to a clinically overt case of PRCA.
Andrew Warren (Novartis) explained how his company had re-organized its bioanalytical resources to co-locate ADA, PD and PK assay specialists into a single operational unit that included pharmaco-toxicological development scientists. Based on the many case examples described, this strategy has been successful in achieving the goal of an integrated data presentation: most importantly, inter-dependent influences of endogenous factors, residual drug and ADA levels on the performance of the respective assays applied for the PK and immunogenicity analyses could be identified, thereby enabling a more valid interpretation of the data; this had increased the reliability of interpreting impact on safety and pharmacodynamic parameters. The integrated bioanalytical plan has the goals of characterising, during non-clinical and clinical development: in vivo exposure to drug and impact of immunogenicity; relationship between dose, frequency and pharmacological effect; link between exposure to drug and toxicity; and the toxicity caused by anti-drug immune responses. The extent of bioanalytical evaluation is adapted to the perceived risk level: testing for cross-reactive potential of pre-existing and treatment emergent antibodies to the drug product and to endogenous factors would be included if necessary; a test to estimate neutralizing potential of ADA’s in pre-clinical samples may be added if there are no effective PD markers. Different PK assay formats are configured to distinguish free “bioactive” drug from total drug, and validated for impact of ADA’s. Assays for soluble ligand PD markers would also be validated for impact of neutralizing and non-neutralizing ADA’s. In non-clinical studies, high levels of circulating drug may well interfere with ADA detection; even in the clinical sample setting, it may not be feasible to obtain wash-out samples to eliminate interference by residual drug. Distinction between transient vs. persistent ADA’s may be helpful; and patient-related heterogeneity must always be taken into account in setting the cut-point for the ADA assay. Careful comparison of the PK, PD and ADA assay read-outs may signal enhanced drug clearance due to treatment-emergent antibodies in association with loss of drug activity toward the PD marker. The situation where ADA’s are not detected, but in which there are unexpected PK and/or PD profiles in the same subjects, may indicate the need to question whether the ADA assay is sufficiently drug-tolerant and/or the assay cut-point is appropriate. Thus, there are many examples in which data integration can help us to understand the potential clinical impact of ADA responses, as well as identifying limitations with the assays themselves.
Birgit Reipert (Baxter BioScience) opened the second day of the workshop in reiterating how the adaptive and innate immune systems may interact to manifest in undesirable responses to administration of therapeutic proteins: co-stimulation of T- and B-lymphocytes results in the development of antibodies whose affinity and specificity adapt or evolve according to the nature of the stimulus; these responses may be influenced by innate immune mechanisms such as cytokine release, up-regulation of immune stimulatory ligands or receptors, recruitment of immune effector cells, targeted cytotoxicity and activation of the complement system. The utility of non-clinical modelling of these responses depends on the nature of the product and the availability of relevant comparators, since the output is a relative estimation of immunogenic potential – bearing in mind the limitations of non-clinical models to predict the incidence of adverse immune-mediated events in humans. Major shortcomings of conventional mouse models include the recognition of human proteins as foreign as well as inter-species incompatibility of antigen presentation and immune effector mechanisms. To some extent, the use of mice transgenic for both the human protein of interest and for a common human MHC Class II haplotype may enable an instructive evaluation of relative capacity to break immune tolerance in the context of presentation in association with human MHC Class II; indeed, this strategy was successful in the case of human factor VIII. In addition, in vitro modelling using human dendritic cells from different MHC Class II haplotypes, in combination with human factor VIII-specific T-cell hybridomas from either human MHC Class II-specific mice or from murine MHC Class II mice, was useful to demonstrate functional (measured as IL-2 release) responses to CD4+ T-cell epitopes for human Factor VIII. Future non-clinical evaluation could include identification (e.g. by LC-MS/MS) of peptide sequences presented in such antigen-specific model systems.
Jacques Descotes (Poison & Pharmacovigilance Centre and Lyon University) completed the speaker programme in an original and thought provoking style, bridging the non-clinical and clinical dimensions in explaining the role of immunogenicity in the wider context of immune-toxicology. He challenged operational notions of immunogenicity as representing any undesirable immune response to drug administration by encouraging a focus on antigen-specific immune responses. An important gap in the current evaluation of undesirable immunogenicity of therapeutic proteins, in contrast to vaccines or cancer immunotherapeutics, is that antigen-specific T-cell responses are rarely monitored. Immune-mediated hypersensitivity is a direct consequence of immunogenicity: features include antigen-specific immune recognition, involvement of immune memory, induction of humoral and cell-mediated immune responses, variable clinical manifestations depending on the underlying mechanism and, most importantly, prior contact to either the therapeutic protein or to other constituents of the drug product. Case examples were provided to illustrate common misinterpretation of AE’s such as erythema as “hypersensitivity” reactions; lack of suitably document cases of serum sickness caused by the modern generation of therapeutic protein products (as opposed to animal IgG preparations) was emphasised; as was the relative rarity of genuine IgE-mediated hypersensitivity reactions to therapeutic proteins. With the exception of cross-reactivity with pre-existing antibodies (due to prior treatment or environmental / dietary exposure), true anaphylaxis does not occur at the first administration; accordingly, case reports should be carefully reviewed to understand the temporal relationship of AE’s to the induction of an immune response against the therapeutic protein. The value of routine monitoring of autoimmune status is rather equivocal because the diagnostic specificity and sensitivity of candidate biomarkers such as anti-dsDNA is highly uncertain. Overall, on the basis of available clinical experience, immunogenicity of therapeutic proteins is a common finding, but rarely the cause of overt clinical consequences. This should not diminish the need for a diligent risk assessment and accurate diagnosis of clinical signs because immune-mediated adverse reactions, although occurring rarely, tend to be severe and treatment limiting.
Perhaps, the major “take-home-message” of this workshop was the importance of presenting an integrated summary of product-specific immunogenicity considerations in regulatory dossiers, both during clinical development and at the marketing authorisation application stage. This should seek to explain the strategy for evaluating immunogenicity in non-clinical and clinical studies in relation to the pertinent risks – including product quality-related factors – as well providing tabular data summaries and explicit linkage to the proposed risk mitigation plan.