Re-thinking the current paradigm for clinical immunogenicity assessment: an update from the discussion in the European Bioanalysis Forum

Abstract

Immunogenicity regulatory guidance and industry recommendations have evolved over the last two decades since unexpected immune reactions were first reported with erythropoietin. Since then, the guidelines and practices for immunogenicity have stemmed from a reaction to a high-risk molecule causing significant clinical impact. Similar thinking is often applied to all biotherapeutic drugs, even when a well-defined risk assessment suggests otherwise. In recent years, the current testing paradigm for immunogenicity has been challenged with more informative approaches being proposed. In a Focus Workshop held by the European Bioanalysis Forum in September 2023, the current immunogenicity testing paradigm was challenged based on the experience and learning of 20+ years of immunogenicity strategies. The workshop recommendations proposed a new paradigm, challenging the value of multiple tiers depending on the immunogenicity risk assessment based on context of use and moving toward treating immunogenicity as a pharmacodynamic biomarker for the drug. Such rethinking ultimately results in the appropriate and efficient focusing of resources on immunogenicity testing strategies that benefit patients most, moving to a new paradigm where implementation of appropriate and truly informative immunogenicity testing strategies, depending on the context-of-use, become the norm .

Keywords: :

• biomarker
• context-of-use
• new paradigm
• three-tier immunogenicity testing

1. Background

The potential to cause an unwanted immune response following the administration of a biotherapeutic, termed as ‘immunogenicity’, is a known potential risk that is assessed, monitored and characterized during drug development to support biotherapeutic drug submissions. Additionally, the inclusion of immunogenicity information in the prescribing information or drug label is described by the Food and Drug Administration (FDA) as a mechanism to “to enable health care practitioners to easily access, understand, and use this information to inform prescribing decisions and patient management, and to help enable safe and effective use of applicable products” [1].

The advent of such assessments arose following several reports of pure red-cell aplasia (PRCA) after treatment of patients suffering from anemia and chronic renal failure with human recombinant erythropoietin (EPO) [2–4], despite the drug having been used clinically without significant impact for several years. In 2002, 13 cases of PRCA over a 3-year period in patients with chronic kidney disease treated with EPO were described [5]. The authors found an immunological reason for the anemia in these patients, which was the formation of antibodies against the dosed recombinant EPO [5] and more cases were reported in subsequent years. While there are multiple theories on the specific cause(s) of the PRCA in the EPO example, it is now well understood that unwanted immune responses following administration of a biotherapeutic can be caused by an inter-play of multiple factors such as the characteristics, molecular complexity and manufacturing method(s) of the specific drug molecule, disease and patient factors and the administration route, dosing duration and frequency [6]. The industry looked at ways to develop strategies and assays to detect and characterize these unwanted immune responses and drew on knowledge in the vaccine space where wanted immune responses had been routinely assessed. However, it should be noted that immune responses to vaccines are expected and desired, and many magnitudes higher than most of the immune responses elicited from biotherapeutic drugs. Though titer was a starting point, the current paradigm evolved partly based on caution and a lack of experience at that time; a screening tier was conceived to remove negative samples and thus avoid labor-intensive titer assessments in all samples, yet due to concerns of missing responses, false positive rates were incorporated with cut points. Additionally, a confirmatory assay was introduced as a mechanism to remove intentionally false positive response brought through from the screening assay. This led us to the three-tiered immunogenicity paradigm.

Regulatory agencies provided guidance for the approaches and methods required to evaluate immunogenicity over the coming years, which have shaped the practices we employ today: the Food and Drug Administration (FDA) [7], the European Medicines Agency (EMA) [8], the National Medical Products Administration (NMPA) [9] and more recently a draft guideline has been issued for public comment by the Pharmaceuticals and Medical Devices Agency (PMDA) [10]. While such guidance has been a cornerstone of immunogenicity testing as understanding and knowledge grew, the regulations were generated in response to what are deemed as high-risk molecules with significant clinical impact and in a time when a conservative and stepwise approach was warranted. Change in regulatory guidance takes time, and new data will naturally lag behind what industry, clinicians and patients are experiencing. In recent years, industry has questioned such topics as: whether the three-tiered testing paradigm is still fit-for-purpose, the utility of testing at all stages of drug development, the replacement of titer assays with signal to noise evaluation and therefore overall, determining the best value-added approach for patients. Additionally, the biotherapeutic field is expanding quickly with newer modalities, progressing beyond the more traditional protein therapeutics such as monoclonal antibodies, and there is much discussion in the field on whether we continue to apply the ideas and strategies that have been previously developed for traditional biotherapeutics, to current new(er) modalities, biosimilars and those yet to be imagined.

In response to such questions, the European Bioanalysis Forum (EBF) built upon past EBF discussions and EBF workshops on immunogenicity and held a Focus Workshop (FW) in Malaga, Spain in September 2023 entitled “Challenging the Current Paradigm for ADA testing” [11]. This FW was attended by delegates from pharmaceutical, biotech and contract research organizations and included an invited speaker from the Center for Drug Evaluation and Research (CDER), FDA. Strategic presentations and case studies were discussed along with roundtable sessions that formed the following EBF recommendations discussed herein.

2. Context of use

The Context of Use (CoU) discussion has gathered momentum particularly in the biomarker field with a number of recent publications [12,13], focusing on understanding not just the technical bioanalysis requirements, but how to design and implement the right assay(s) of appropriate quality to generate the correct data for informed decision-making, documenting the CoU decisions and ensuring timely and effective stakeholder communication. Such thinking should not be reserved solely for biomarker assays; in fact, every assay has a CoU in that it should be appropriate for its intended purpose.

The three-tiered immunogenicity testing paradigm that is currently described by the regulatory agencies - tier 1 (screening), tier 2 (confirmation), tier 3 (characterization) -was created from a position of caution to generate a stepwise approach whereby each assay at each tier has a different intended purpose and therefore a different CoU. However, we know that this stepwise approach was born from a vaccine CoU (wanted and large immune responses utilizing titer assays), which looking back is inherently flawed for unwanted immune responses. When CoU for immunogenicity is considered, immunogenicity assays should really be regarded as biomarker assays. The BEST (Biomarkers, EndpointS and other Tools) definition of a biomarker states “A defined characteristic that is measured as an indicator of normal biological processes, pathogenic processes, or biological responses to exposure or intervention, including therapeutic interventions” [14]. Clearly, this describes immunogenicity, whereby the biological response to the administration of a biotherapeutic is detected, monitored and may be further characterized. If this thinking is extrapolated further, the three-tier testing paradigm, the utility of the assay cut point and how data are displayed and interpreted can be challenged [15]. By plotting signal/noise (S/N) data from the screening assay alone and plotting in the same way as biomarker data, fuller, richer datasets, produced much earlier make it possible to provide more meaningful data interpretation than current practices.

3. The impact on resources of immunogenicity testing

The impact of the current immunogenicity testing paradigm on time and resources should not be underestimated. Unfortunately, as an industry, possibly through fear of regulatory push-back or internal company pressures, we do not tend to stop and reflect on what activities are value added for meaningful data interpretation and ultimately the best return on investment for the patient. As bioanalytical scientists we often feel comfortable in defined processes and “tick-box thinking”. However, just because we can measure something doesn't always mean that we should. Rather the value and importance of the data needs to be considered. Drug development is a long and costly process and resources should be directed to areas that ensure safe and efficacious medicines are reaching patients in need, who have been possibly waiting years for a therapy. Additionally, trial conduct should consider patient-centric approaches, ensuring that only appropriate samples and volumes are collected. However, our comfort with “tick-box thinking” and the misapplication or overuse of the three-tiered testing paradigm has created an environment where immunogenicity assays are highly sensitive, often with low cut points that are measuring immunogenicity signals with little or no relevance. As a result, there are multiple examples of marketed drugs that report high incidence of immune response detection that do not correlate with clinical impact. Additionally, excessive characterization activities such as domain specificity testing or applying neutralizing antibody assays (nAb) as default for all programs even when nAb can be detected by other means such as functional pharmacokinetic (PK) assays or robust pharmacodynamic (PD) markers are commonplace. The practice of analyzing samples in the final assay plate in duplicate has become default, even when the first step in a bridging assay is performed in a single well and many companies still perform testing on placebo samples. Furthermore, guidance that was generated for clinical testing has been applied by industry to nonclinical studies even in the absence of impact of changes in exposure or PD readouts, sometimes using multiple immunogenicity tiers. It is well-known and understood that human proteins administered to animals will likely generate an immune response as they will be seen as foreign, and responses observed in animals do not translate to the clinic [16]. Such scope creep should be mitigated by guidance such as ICH S6(R1) [17] and the EBF recommendation [18] yet the practice continues. All such procedures result in a heavy burden of testing, consuming time, money and resources often without added value. We should serve patients more effectively by repurposing efforts into those areas that truly have impact; patients are gifting their samples and it is our responsibility to ensure we only take samples that we use and that provide meaningful assessments.

4. EBF 2023 FW

During the FW, there were presentations on the theme of challenging the three-tiered approach. The presentations can be found here: https://e-b-f.eu/fw202302-slides. Additionally, 6 topics were discussed among the delegates in a roundtable format to help shape the EBF recommendation on how immunogenicity assessment should evolve based on 20+ years of experience and are discussed herein:

▪	three-tier testing paradigm
▪	Signal to noise (S/N) as an alternative for titer
▪	Characterization testing
▪	Singlicate analysis
▪	Drug tolerance
▪	Measurement of placebo samples in clinical testing

5. Three-tier Testing Paradigm

The three-tier testing paradigm is accepted practice, described in regulatory guidance for clinical immunogenicity assessment (Figure 1). However, the application of all three tiers is often stage dependent in that for early programs, and depending on the risk assessment [8,19,20] of the molecule, screening, confirmatory and titer assays may be applied and nAb assays for further characterization of the response may only be deployed for pivotal Phase 3 studies. This stepwise approach is conservative in that samples are ‘screened’ for potential positivity with an inherent 5% false positive rate (FPR) built into the assays via the cut point. All positives are then taken into a confirmatory assay to determine the specificity of the response against the drug molecule (by the addition of spiked drug) using a 1% FPR. These assays are non-orthogonal; they are based on the same assay format, often generating a moderate to strong positive correlation between the two assays. Therefore, the deployment of the confirmatory assay rarely reduces the number of screened positive results by any significant amount, and instead removes borderline positive signals hovering around the cut point and have the potential to contaminate final data sets with false positives that dilute interpretation of the impact of ADA. Kubiak et al. 2013 [21] demonstrated through case studies that similar results could be achieved with the removal of the confirmatory tier and deploying just the screening assay with a 1% FPR. Additionally, the titer tier is the same assay format as the screening tier, the only difference is that a positive sample is serially diluted down to the cut point and below, to evaluate the magnitude of the positivity. Thus, a sample is tested three-times in the same assay.

Figure 1. Tiered Approach for Clinical Immunogenicity Assessment.

Adapted from EMA Guidelines [8].

The need for a confirmatory assay can be assessed in validation using a specificity plot whereby by the Tier 1 results (expressed as Signal/Noise) are plotted on the x axis against the Tier 2 results (expressed as % inhibition) on the y axis. Data can be visually evaluated along with the calculation of a correlation coefficient (r) to statistically measure the strength of the linear relationship between the 2 variables.

Following the round table discussions, the EBF recommendations regarding the three-tier testing paradigm are as follows:

▪	Inclusion of the confirmatory assessment should be performed based upon risk of the molecule rather than a mandatory requirement. The omission of the confirmatory tier is considered low risk; should the data be required, it can be performed at a later date due to the known and well documented stability of antibodies in serum and plasma.
▪	Omission of the confirmatory tier can be justified through data such as specificity curves shown in assay validation and/or early phase clinical studies.
▪	If the confirmatory assay is omitted, the screening assay should use a 1% FPR (rather than a 5% FPR).
▪	Should a confirmatory assay be included in early phase clinical studies, a screening assay with a 1% FPR alongside the confirmatory assay should be considered, to determine if the data shows that the confirmatory assay can be excluded.

6. Signal to noise as an alternative for titer

The titer assay used to determine the magnitude of the response utilizes the same format as the screening assay but titrates the positive sample via serial dilution. These assays are often prone to variability as the approach relies on reporting a value based on the lower plateau of the assay range where small changes in signal can have a significant impact on the reported value. Over recent years, alternative approaches to titer have been proposed such as the assessment of signal to noise (S/N) [22–25] and several case studies were presented at the FW. However, there have been some barriers to adoption, for example:

▪	Concerns that prescribing physicians may not understand S/N data despite titer data not being used for dosing decisions.
▪	Appetite for more data and case studies.
▪	Wanting others to go first and receive successful submissions.
▪	Comfort in established processes and inertia to make a change.
▪	Fear of management scrutiny should regulatory questions be received.
▪	Concerns that hook effect in the assay could impact the data.
▪	Late in the drug development process and harder to change approach.

Though some companies may need more data to feel comfortable and are performing their own internal evaluations, the EBF recommendations on S/N are:

▪	The S/N approach is justified as shown by numerous presentations and publications and each assay should be evaluated for the applicability of S/N.
▪	Should the assay be acceptable for S/N, teams should plan to discuss and justify the approach with Health Authorities and in IND/CTA submission documents as part of the proposed immunogenicity strategy.
▪	To build confidence with stakeholders and regulators, teams should adopt S/N in Phase I studies and consistently apply through drug development.
▪	A S/N value >x can be used instead of titers >y to describe clinically meaningful responses on the drug label.
▪	Samples should be retained in suitable storage conditions; titer analysis can then be performed at a later date, if required by Health Authorities.

7. Characterization testing

For biotherapeutics that are bi- or tri-specific in nature, possess multiple domains, or are fusions or conjugated molecules, further characterization of the immune response is often performed by companies to assess which part of the molecule is triggering the response. This is mostly performed for internal decision-making, rather than a regulatory requirement, whereby the data may be useful for understanding potential liabilities in next generation molecules or to de-risk a platform. Regulatory interest is usually reserved for high-risk molecules, considering the likelihood of generating an immune response and the impact and significance of any clinical consequence. Historically, the industry has routinely implemented nAb for low(er) risk biotherapeutics, only deploying them at later stage studies such as Phase 3, while Phase I nAb assays should only be reserved for those truly high-risk molecules. An FDA analysis, presented during the Focus Workshop, of approved Biologics License Applications submitted to FDA from January 2019 to December 2022, showed that 5.4% of the approvals had successfully used titer and PK/PD data as an alternative approach to nAb assays and 21.7% had a successful approval with no nAb data at all [26].

The EBF recommendations for characterization testing are:

▪	The inclusion of additional characterization assessments must be linked to a risk-based justification for the study or bring future benefit and safety for patients.
▪	Alternate, less resource-intensive approaches should be considered and implemented where possible for higher benefit for patients.
▪	Should characterization assays be deemed necessary, early development and characterization of the reagents should be considered due to the complexity of the assays.
▪	S/N data can be used instead of titer for justifying alternative approaches to nAb assays.

8. Singlicate analysis

Singlicate analysis, whereby a sample is tested in one well rather than the often default two wells in a plate, is becoming more common for PK assays with the issuance of ICH M10 [27] and industry papers [28,29]. However, there has been some resistance from the community for adoption for immunogenicity assays even though regulatory guidance does not state that a sample must be analyzed in more than one replicate. The industry feels the need to adopt duplicate analysis and we may do this three-times with essentially the same assay format used for screening, confirmatory and titer. Furthermore, in the bridging assay format where labelled reagents are incubated with the sample or the positive control, this is usually performed in a single well before the single sample is aliquoted into two wells of the final assay testing plate, i.e., a technical and not biological replicate. Scientifically, this practice is not value adding and serves only to test the pipetting skills of the analyst rather than true precision of an assay.

The EBF recommendations for singlicate analysis for immunogenicity assays are as follows:

▪	Singlicate analysis is acceptable for immunogenicity assays and duplicate analysis is not a regulatory requirement.
▪	Significant operational benefits can be gained by singlicate without impacting data quality especially when combined with other approaches such as S/N or omission of the confirmatory tier.
▪	Singlicate analysis supports green and sustainability considerations, and patient centric approaches; less sample volume is needed, less storage space is required for banked samples or samples awaiting analysis thus resulting in a reduction in CO2, a reduction in single use plastics and critical reagents.
▪	Immunogenicity data are not evaluated in isolation and form part of integrated datasets that assess clinical impact in conjunction with PK, PD and safety evaluation.

9. Drug tolerance

Administered drug present in study samples at certain concentrations has the potential to interfere in immunogenicity assays, potentially causing false negative results. As a result, significant efforts are often made to ensure that assays can tolerate sufficiently high levels of drug, possibly employing additional assay steps, or by increasing the minimal required dilution in the method, to meet the regulatory expectations [30]. The EMA guidance has set a high bar in this regard stating: “demonstrate that the tolerance of the assay to the therapeutic exceeds the levels of the therapeutic protein in the samples for ADA testing”. However, EMA does acknowledge that this may not always be feasible “due to technical limitations” and states “If this occurs, the best possible assay should be employed and the approach taken should be properly justified”.

During validation, it is often typical to evaluate the impact of multiple levels of drug on ADA concentrations even though the assessment is performed with a surrogate ADA positive control. Therefore, the drug tolerance determined in validation may not be reflective of the actual drug tolerance observed in study samples and can only be viewed as an estimate [31]. Additionally, many new assays are more sensitive compared with historical assays, with cut points that are at the limits of the analytical detection capable by the instrument. Given the sensitivity of current assay formats, the level of surrogate ADA tested in validation should be relevant and the results should be considered in context to the risk of the molecule and the potential clinical consequence(s). During the FW, there was a presentation that demonstrated a case where drug tolerant assays confounded the effect of ADA on efficacy [32,33] and side effects caused by ADA were mostly related to high ADA magnitude.

Drug tolerance within the study needs to be understood and contextualized alongside other available data; often PK is the first indicator of an ADA response, and the immunogenicity data (usually generated much later than the PK data) are often just confirmation of information already understood from the PK data. Additionally, the impact of potentially false negative samples in relation to clinically relevant responses, and the risk assessment of the biotherapeutic, should be considered. Missing a low response is unlikely to be clinically meaningful for low to medium risk biotherapeutics but would have more significance for drugs with non-redundant endogenous counterparts.

The EBF recommendations for drug tolerance are as follows:

▪	The risk assessment and the potential clinical consequences should drive the need and the extent of the drug tolerance assessments performed in validation.
▪	Whenever possible, thoughtfully select sampling time points, including washout samples, depending on the (known or predicted) PK of the drug candidate and choose appropriate drug levels for testing in development and validation, depending on your study.
▪	Choose an appropriate target for drug tolerance depending on your study, considering Ctrough rather than Cmax.
▪	When selecting drug concentrations for testing in validation, the relevance to the study(ies) should be considered and drug tolerance only needs to be achieved at levels of drug expected in samples that will need to be tested to understand the impact of the presence of ADA.
▪	Testing drug tolerance at 100 ng/ml of PC is recommended for validation. If suitable drug tolerance is not achieved at 100 ng/ml, then 250 ng/ml and 500 ng/ml PC levels may be included; other levels should be risk- and study-related. Testing at the LPC level is not recommended. HPC level is not required.
▪	For most molecules with low immunogenicity risk, assessment of drug tolerance in the screening assay is deemed sufficient. Depending on the risk and consequences of immunogenicity, assessment of drug tolerance in the confirmatory tier might be necessary, but drug tolerance is not needed to be assessed in the titer assay.
▪	One validation run is sufficient for drug tolerance parameters.
▪	It is recommended to engage with the health authorities early in a drug development program, including when the desired level of drug tolerance is not achieved.

Given the depth of discussion, EBF is writing a more extensive paper on the topic of strategies to assess appropriate drug tolerance.

10. Measurement of placebo samples in clinical testing

Often it is common practice to omit the evaluation of placebo samples in pharmacokinetic assessments without justification, yet when it comes to immunogenicity, some companies routinely analyze placebo samples. This can increase cost and timelines without added value.

The drivers for routine testing placebo samples can be:

▪	Testing labs may not be unblinded and therefore whether samples are active or placebo is unknown.
▪	Challenges in receiving randomization codes to know which samples should not contain drug.
▪	Checking that the assay is performing as expected.
▪	Stakeholder pressure.

The EBF recommendations on placebo testing are as follows:

▪	Placebo samples should only be analyzed when there is a justifiable cause.
▪	If placebo samples are analyzed, only selected timepoints in early studies should be tested.

11. Conclusion

The current immunogenicity testing paradigm was created in response to adverse events seen with a high-risk molecule, drawing on existing experience from the vaccine world. However, assessment of wanted immunogenicity has a different CoU to the evaluation, interpretation and impact of unwanted immune responses. Knowledge and experience have grown and evolved over the last two decades which has prompted discussions within the community that question whether the established practices still add value or whether new and scientifically valid approaches should be adopted rather than using guidance that was originally written based on industry experience with high-risk molecules. “Tick-box thinking” should be eliminated and replaced with strong scientific rationale; just because something can be measured, it does not mean that it adds extra value. It should be remembered that not all drug programs are the same and neither are the risks and therefore potential clinical consequences. Regulatory guidance may not be applicable for all scenarios, and the EBF recommendations discussed herein can form the basis of more meaningful and scientific approaches for immunogenicity testing strategies.

12. Future perspective

Going forward, a new paradigm where immunogenicity assessment is considered as a biomarker of the drug is gaining traction from the discussions at the Focus Workshop and in continuation within the EBF community. As follow-up from the discussion at our meeting, Lauren Stevenson has connected with and engaged CDER and CBER immunogenicity reviewers. The FDA regulators have shown significant interest and are eager to see more data sets using the simplification of the three-tiered approach into a one-tier approach, i.e., evaluating signal to noise (S/N) as an alternative approach to titer as a means to measure magnitude of immunogenicity responses to biologic therapeutics. Additionally, the EBF plans to join this effort and a team is formed to generate more data in support of this one-tier approach. When such a mindset is adopted, the initial testing paradigm of screening, confirmatory and titer shifts toward plotting signal to noise data from only the screening tier. The screening assay would not require a cut point, removing the concerns over false negatives. Instead, all generated data are plotted and reported. This affords much richer datasets that show the totality of the response in a single plot. The advantages to such an approach are numerous; faster generation of data without the potential to miss responses, the potential for real-time analysis and faster correlation of responses. When considering the regulatory review, this would streamline tables, figures and listings and have the potential for more efficient and faster review of the data by the agencies. This can only benefit patients and their access to life-changing medicines if we implement the best scientific approach rather than continuing with current industry practices because that is the way it has always been done.

Article highlights

• The EBF is gathering data in support of recommending a new paradigm for immunogenicity testing, challenging the value of multiple tiers and by including the principles of context-use moving towards treating immunogenicity as a pharmacodynamic biomarker for the drug.

Financial disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the recommendation. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Writing disclosure

No writing assistance was utilized in the production of this manuscript.

Disclaimer

The views and conclusions presented in this paper are those of the European Bioanalysis Forum (EBF) and do not necessarily reflect the representative affiliation or individual company's position of the authors on the subject.

Acknowledgments

The authors wish to thank J Pedras-Vasconcelos for his valued contribution to the Focus Workshop.

Competing interests disclosure

The authors have no competing interests or relevant affiliations with any organization or entity with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, stock ownership or options and expert testimony.