Refined ADA testing strategy to overcome sRANKL interference in denosumab/FKS518 immunogenicity assessment

ABSTRACT

Background

Fresenius Kabi developed FKS518, a fully human monoclonal antibody biosimilar to denosumab, that inhibits osteoclast activation by targeting the receptor activator of nuclear factor kappa-Β ligand (RANKL). RANKL exists as a trimeric soluble protein (sRANKL) with high affinity for denosumab. Post-dose, sRANKL levels can rise due to drug–target complex accumulation, creating a risk of interference in bridging anti-drug antibodies (ADA) assays, particularly after acid dissociation.

Methods

We evaluated sRANKL interference in the ADA bridging assay and, to competitively block sRANKL, we introduced a specificity tier by adding osteoprotegerin (OPG). This approach enabled reanalysis of previously ADA-positive samples to confirm whether signals represented true ADA responses or artifacts caused by sRANKL interference.

Results

Acid dissociation significantly exacerbated target interference, resulting in ADA positivity rates of ~96–98% in clinical studies. Introducing a specificity tier corrected incidence to ≤3.9%. OPG incorporation did not change the minimum required dilution (MRD) of the assay and did not affect signals for negative/positive control, confirming assay integrity.

Discussion

These findings underscore the importance of early interference assessment and mitigation. A multi-tiered strategy, encompassing screening, confirmatory, and specificity tiers, provided a robust solution applicable to programs facing similar challenges.

Plain Language Summary

This study identified and solved a problem with the laboratory test for measuring immune responses to denosumab or its biosimilar, FKS518, in people treated with these medicines in clinical studies. Sometimes, after treatment, the body can make antibodies against the drug (called anti-drug antibodies or ADAs). Detecting these antibodies is important because they can affect how well the medicine works and its safety.

The initial test showed that too many people had ADAs, but this was mostly because a protein called sRANKL, which increases after treatment, was interfering and causing false positives results. To fix this, an extra step was added to the test using a natural protein called osteoprotegerin (OPG), which blocks sRANKL. This made it possible to differentiate between real ADA responses and false results caused by interference. This new approach showed that the actual number of people with anti-drug antibodies was much lower and matched what is known from past studies.

These results show that it’s important to check for this kind of interference early when developing tests for measuring ADAs. Adding OPG to the test is a practical way to get reliable results for biosimilar drug development and approval, ensuring that patients and doctors have reliable information about the safety of these medicines.

GRAPHICAL ABSTRACT

For the assessment of ADAs to denosumab/ FKS518, a four-tiered testing strategy was developed, comprising an initial screening assay, followed by confirmatory assay, specificity assay and titration assay. For the specificity and titration tiers, the original ADA method was modified to include osteoprotegerin (OPG) to sequester sRANKL and mitigate the target interference.

1. Introduction

1.1. Denosumab overview and biosimilar development

Denosumab (originator’s brand name Prolia®, Xgeva®, Amgen) is a fully human IgG2 monoclonal antibody (mAb), which targets the receptor activator of nuclear factor kappa-Β ligand (RANKL), an essential factor for the formation, activation, and survival of osteoclasts.

The mechanism of action of denosumab mimics that of the natural decoy receptor of RANKL, osteoprotegerin (OPG), by binding to RANKL and preventing it from activating its receptor RANK, found on the surface of osteoclasts and their precursors RANKL production is increased in conditions associated with depleted estrogen levels, such as menopause or hormone ablation, but also in solid tumors with osseous metastasis [1].

Fresenius Kabi has developed FKS518 as a denosumab biosimilar to Prolia and Xgeva. In accordance with regulatory requirements for biosimilars, the clinical development program aimed to demonstrate that there are no clinically meaningful differences between the biosimilar and its reference product in terms of pharmacokinetics (PK), pharmacodynamics (PD), safety, efficacy, and immunogenicity.

1.2. ADA assay challenges and soluble RANKL (sRANKL) interference

The administration of therapeutic proteins has the potential to cause unwanted immune responses in humans, due to the formation of anti-drug antibodies (ADA), that could potentially impact patient safety and drug efficacy. Thus, ADA determination is a critical component for the development of a therapeutic biological drug as well as that of a biosimilar. Well-designed and specific ADA immunoassays are crucial for performing thorough comparative immunogenicity assessments between reference products and biosimilars. The contemporary testing paradigm for detecting and characterizing ADAs involves a multi-tiered strategy encompassing screening, confirmatory and titer assessment. This strategy is widely accepted and recommended by regulatory agencies.

However, the presence of soluble drug targets can lead to erroneous results in the ADA measurements for therapeutic antibodies. The interference of circulating target on the ADA assay can present as either false-positive or theoretically false-negative ADA results, depending on the form of the drug target (whether it is “free” or “bound”), the format of the assay (e.g., with or without acid dissociation), target characteristics (monomer or multimer), disease-specific factors, the concentration of soluble drug target in the sample, and the affinity and avidity of the overall ADA response. In addition, it has been observed that for certain biological drugs, like therapeutic monoclonal antibodies, target levels are low and non-interfering at baseline but may increase upon treatment, due to the formation and accumulation of target-drug complexes [2]. False positive ADA results due to target interference have been reported for patients treated with mepolizumab [3], fulranumab [4], ofatumumab [5]. Therefore, it is important to evaluate the interference on ADA detection by the drug’s intended targets, particularly in the presence of high levels of soluble multimeric targets and when utilizing bridging immunoassays that incorporate an acid dissociation step.

Denosumab is uniquely susceptible to target interference in the ADA assay due to the intrinsic properties of its target. RANKL is a type II transmembrane protein that can be cleaved by matrix metalloproteinases to release a soluble, biologically active product (sRANKL). sRANKL naturally exists in trimeric form under physiologic conditions [6]. In blood samples from untreated patients, trimeric sRANKL most likely circulates complexed to OPG [7]. Denosumab shows high affinity for both sRANKL and membrane-bound RANKL, favoring persistent drug–target complexes in circulation. These multimeric complexes can act as bridging agents in immunoassays, particularly after acid dissociation steps that liberate sRANKL from complexes. When sRANKL concentrations are elevated post-treatment, its multivalency enables re-binding to labeled denosumab during assay incubation, generating artificial bridges and false-positive ADA signals.

The potential for interference in the ADA assay was evaluated during the validation of the initial method, by spiking increasing amounts of sRANKL, in the absence or presence of the polyclonal positive control, at low and high concentration (LPC and HPC). It was demonstrated that target interference is not observed in the assay in the presence of up to 500 pg/mL of sRANKL. As reported in various studies from literature that aimed to quantify sRANKL in different populations [8–11], the sRANKL levels applied in the target interference assessment during assay validation were well in excess of the physiological levels or of the levels observed in women with postmenopausal osteoporosis (PMO). Therefore, the assay was deemed fit for purpose and applied in sample analysis.

Nevertheless, the ADA incidences initially observed in the Phase 1 healthy volunteer study (FKS518-001) and Phase 3 study in women with PMO (FKS518-002) were comparable between treatment arms, but much higher than that observed in recent studies of other biosimilar candidates [12–15], which are expected to similarly rely on more advanced techniques, supposedly bearing higher sensitivity. These results did not raise any concern about differential immunogenicity between FKS518 and US-Prolia itself, which was supported by the comprehensive analytical similarity data and overall comparable safety profiles across clinical studies, including the general lack of adverse events potentially tied to immunogenicity (such as hypersensitivity reactions). Nonetheless, the unexpectedly high incidence triggered an investigation. It was hypothesized that the high overall incidence of ADA positive samples was not only due to a highly sensitive method, but to sRANKL levels exceeding 500 pg/mL, that may have contributed to false positive ADA signals.

The suspicion was supported by the observation of repetitive peaks of ADA prevalence in the FKS518-002 study (at week 12, 40 and 64) (Supplementary Figure S1B), which happened at 12 weeks post each dosing (week 0, 26 and 52) (Supplementary Figure S1A).

1.3. Resolving sRANKL interference and revised ADA testing strategy

The investigation confirmed that sRANKL interfered with ADA detection in both screening and confirmatory tiers, leading to false positive signals and overreporting of ADA incidence in both arms and in both studies. It was demonstrated that addition of OPG abrogates the signal induced by spiked sRANKL, while retaining sensitivity, specificity and precision of the assay, allowing unequivocal detection of positive ADA samples.

Two potential approaches were evaluated at this point to address sRANKL interference: complete ADA assay revalidation, followed by reanalysis of all clinical samples, requiring significant time and resources, or implementation of an OPG-based specificity tier to confirm ADA-positive signals and eliminate sRANKL-related false positives. Both approaches are scientifically sound and ensure robust data; however, the second option was selected as the most operationally feasible. A denosumab ADA specificity tier using OPG was developed and validated. This approach was chosen because clinical samples had already been analyzed using the original assay. The additional tier allowed to limit analysis to only previously positive samples to verify whether the detected signals represented true ADA responses or were artifacts caused by sRANKL interference.

True ADA incidence was re-determined and a comparison of the immunogenicity results in the FKS518-001 and FKS518-002 clinical studies, by using the initial or the revised assay, is presented in this paper, together with a recommendation on how to approach assay development to minimize target interference.

These findings highlight the importance of a careful examination of drug target interference in ADA methods to ensure reliable immunogenicity assessments.

This work extends the observations reported by Miller et al., 2025 [16], in their publication “Inhibition of RANKL is critical for accurate assessment of anti-drug antibody incidence to denosumab in clinical studies”, which highlighted the potential for sRANKL to interfere with ADA detection in bridging immunoassays. Both groups recognize that multimeric sRANKL can form nonspecific bridges between labeled drug reagents, mimicking ADA signals and leading to false-positive results. While Miller et al. used an anti-RANKL antibody to mitigate this interference, we took a different approach, introducing a specificity tier using OPG, an endogenous decoy receptor for RANKL. This additional tier was applied to reassess clinical samples that initially tested positive. The results demonstrated that incorporating the specificity tier substantially reduced ADA incidence, confirming that the previously high rates were due to sRANKL interference.

Our strategy of adding the specificity tier preserved assay sensitivity, precision, and drug tolerance and enabled the generation of robust and reliable immunogenicity data, supporting the immunogenicity assessment.

2. Materials & methods

2.1. Materials & reagents

Two different anti-denosumab antibodies have been used as positive control (PC) during validation: 1) Polyclonal rabbit anti-denosumab (Prolia) antibody, affinity purified from immunized rabbit sera at Biogenes GmbH. 2) Monoclonal human anti-denosumab (Prolia) antibody (BioRad). The polyclonal anti-denosumab antibody was used to evaluate and report assay characteristics because it better resembles the spectrum of possible immune responses in comparison to a monoclonal antibody. The monoclonal anti-denosumab antibody was selected as system suitability control to monitor in-study sample analysis due to a limited availability of the polyclonal antibody. The negative control (NC) was prepared by pooling commercial human serum samples (BioIVT) that had a response below the cut point in the initial ADA assay.

Biotinylated- FKS518 and Sulfo-tag-FKS518 were derived from FKS518 and were produced by PPD Laboratory Services (Virginia, USA). Streptavidin coated plates, read buffer, and all other Meso Scale Discovery (MSD) consumables were sourced from MSD. Recombinant Human RANKL was sourced from Cell Science Systems or EMD Millipore and Recombinant Human OPG was sourced from BioLegend.

2.2. Methods

2.2.1. Bridging ECLIA to detect ADAs against denosumab

A validated stepwise electrochemiluminescent (ECL) bridging assay format, incorporating an acid-dissociation pre-treatment step was developed and validated by PPD Laboratory Services, part of Thermo Fisher’s Laboratory, based on MSD ECL platform. For the specificity tier, the original ADA method was modified to include OPG to sequester sRANKL and overcome the target interference.

2.2.2. Initial ADA testing strategy

For this assay, an MSD-Streptavidin (MSD-SA) plate was first blocked, and then coated with 0.250 µg/mL biotin-FKS518 for 1 hr ±5 min. Samples, positive controls (PCs), and NC were pre-diluted in assay buffer 1:5. To release any denosumab bound to ADA, pre-diluted serum samples (1:5) were acidified by further diluting 1:5 in 300 mM acetic acid and incubated at room temperature (RT) for 40 min ±5 min. Acid-treated serum samples were then neutralized by diluting 1:2 in neutralization buffer (0.280 M Tris-HCl). The final minimum required dilution was 1:50.

Neutralized samples were then incubated for 1 hr ±5 min on the blocked MSD-SA plate to allow any ADA present in the human serum to bind to the biotin-FKS518. A detection solution containing 2.00 µg/mL Sulfo-TAG FKS518 (detection reagent) was then added for an additional 1 hr ±5 min to bridge biotin FKS518-ADA complexes to Sulfo-TAG FKS518. The plate was then washed, and MSD GOLD Read Buffer was added to the plate. Plates were read on an MSD Sector S 600 microplate reader, and the signal was directly proportional to the amount of ADA present in the human serum. Samples that tested positive in the screening assay were further assessed in a confirmation assay.

A competitive inhibition with excess denosumab was used for the confirmatory assay. It was based on the use of excess FKS518 in a competitive binding format to demonstrate the specificity of the binding interactions in the antibody/labeled drug complex. The mixtures were incubated and tested as described previously in the screening assay procedure.

The assay was validated in accordance with regulatory guidelines, demonstrating acceptable performance across all parameters, as summarized in Supplementary Table S1. System suitability criteria and failure handling procedures were applied in accordance with current regulatory guidance.

2.2.3. Investigation of drug target interference by sRANKL in the ADA method

2.2.3.1. Assess the impact of increasing sRANKL levels on ADA measurements, with different assay conditions

In order to verify our hypothesis regarding the accumulation of sRANKL bound to denosumab/FKS518 at elevated levels and its subsequent release upon acid pre-treatment, we tested a wider range of sRANKL concentrations (up to 20,000 pg/mL, from initially up to 500 pg/mL during the method validation) and examined the potential impact of these higher concentrations on target interference in the screening and confirmatory tiers in undiluted normal human serum. The samples were also spiked with relevant FKS518 concentrations (0.0, 0.8, 2.0, and 6.0 µg/mL) to evaluate the impact of denosumab/FKS518 concentrations on the target tolerated levels. These samples were first incubated at room temperature to ensure complex formation between sRANKL and FKS518, underwent one freeze-thaw cycle prior to analysis to better represent study samples conditions, and then tested in the bridging ADA assay, with or without acid dissociation. The experiments were performed in duplicate by two analysts on three different days.

2.2.3.2. Evaluate target interference in the presence of preformed complexes (ADA- FKS518; sRANKL-FKS518), that mimic clinical samples conditions

Mock samples containing varying amounts of ADA and/or FKS518 and/or sRANKL in normal human serum were prepared sera at concentrations intended to mimic expected clinical study samples. Starting from this experiment, and for all the subsequent ones, the maximum level of sRANKL tested was fixed at 5000 pg/mL, corresponding to the maximum signal level observed during sample analysis of FKS518 clinical studies. The experiments were performed in duplicate by two analysts on two different days.

2.2.3.3. Mitigation of drug target interference with addition of OPG

The assay was optimized with mock samples by testing increasing concentrations of OPG (0–4 000 ng/mL), OPG was added at the neutralization step to block any free sRANKL or released from its bound state (denosumab-sRANKL) upon acid dissociation. The addition of OPG in the neutralization buffer did not change the MRD. In a partial validation, sensitivity, LPC, precision, drug tolerance, and selectivity were confirmed as OPG does not affect the signal of the NC or LPC (refer to supplementary Table S1).

2.2.4. Two-step specificity assay

To generate accurate immunogenicity data, the testing strategy was adapted to incorporate, after the screening and confirmatory assay (original method, Tier I and II), an ADA specificity tier (Tier III), to identify positive signals due to ADA, while eliminating false positive signals due to sRANKL. Since this interference led to reporting of false positive results only, the results for samples that tested ADA negative in the original method remained valid and only the confirmed ADA positive samples from the original analysis were further tested in the ADA specificity tier (Graphical Abstract). The specificity tier is constituted of a screening and subsequent confirmatory assay, in the presence of decoy receptor for RANKL (OPG). The samples confirmed positive in the specificity tier were further titrated in the presence of OPG (Tier IV) to characterize the magnitude of the ADA response.

In addition to the NC and ADA positive controls (PCs), a sRANKL control (5 000 pg/mL), with and without OPG, was introduced to monitor the signal from sRANKL interference in the assay and its mitigation in the presence of OPG.

2.2.5. Assays cut points

Following industry practices, during initial validation the screening cut point (SCP) was calculated to allow for 5% false positives rates (FPR), confirmation cut point (CCP) to allow for 1% FPR and titer cut point (TCP) to allow for 0.1% FPR [17]. A balance design utilizing drug naive matrix samples from 50 individuals generating 300 data points was used during assay validation to generate screening, confirmatory and titer cut point factor values. Calculation of FPR in baseline samples from studies FKS518-001 and FKS518-002 confirmed the suitability of the validation cut points for samples analysis, with observed screening FPRs of 2–11%.

During partial validation of the specificity tier, 48 drug-naïve clinical matrix samples from Study FKS518-002 were used to reestablish the cut points in the presence of OPG with similar FPR of 5%, 1% and 0.1% for the screening, confirmatory, and titration tiers, respectively. The analysis of a random set of 50 pre-dose samples from study FKS518-001 led to a false positive rate of 2%, which confirmed the suitability of these cut points for the analysis of the FKS518-001 samples as well.

Refer to supplementary Table S1 for a summary of screening, confirmatory, titer cut point values for initial ADA assay (screening and confirmatory) and for the specificity tier.

3. Results & discussion

3.1. sRANKL interference assessment and its inhibition in the presence of denosumab/FKS518

Reported physiological RANKL levels in drug naïve subjects are approximately between 20 and 140 pg/ml [8–11]. sRANKL interference in the screening and confirmatory assay was examined during the investigation for target interference by spiking increasing concentrations of sRANKL (up to 20,000 pg/mL) in undiluted normal human serum. As shown in Figure 1, target interference was observed starting from 313 pg/mL sRANKL in the confirmatory assay. The discrepancy with the original validation, where interference was not observed at levels up to 500 pg/mL, can be explained by method variability and different dilution factors.

Figure 1.

sRANKL detection in the validated ADA assay, with acid dissociation.

sRANKL interference is evaluated in the screening and confirmatory assay by spiking increasing concentrations of sRANKL up to 20,000 pg/ml, in undiluted normal human serum. Red dotted line: SCP, Green dotted line: CCP.

Abbreviations: SCP = Screening cut point; CCP = Confirmatory cut point; RLU = Relative light units; sRANKL = Soluble Receptor activator of nuclear factor kappa-Β ligand.

The samples were also spiked with relevant FKS518 concentrations to determine the impact of denosumab/FKS518 concentrations on the target tolerated levels. The FKS518 concentrations selected represent the range from 0 to the expected Cmax at 6.0 µg/mL, including 2 µg/mL, which was observed at week 40 and week 64 during the clinical study FKS518-002, and corresponds to the timepoints at which the highest ADA incidence was observed (Supplementary Figure S1, panel A and B).

The results have shown that interference by sRANKL is reduced in the presence of increasing drug concentrations. In particular, in the presence of 0.8 to 2.0 µg/mL of FKS518, sRANKL interference was observed at levels above 625 pg/mL, while in the presence of FKS518 concentrations higher than 6.0 µg/mL, sRANKL interference was detectable only at levels above 2500 pg/ml (Figure 2).

Figure 2.

sRANKL interference in the ADA assay with acid dissociation in the presence of relevant FKS518 concentrations.

In the method with acid pretreatment, sRANKL interference is evaluated in the confirmatory assay by spiking increasing concentrations of sRANKL up to 20,000 pg/ml, in undiluted normal human serum, in the presence of relevant concentrations of FKS518. FKS518 was added to sRANKL spiked samples at 0.0, 0.8, 2.0, 6.0 µg/mL, and incubated for 2 h at room temperature to ensure complexes formation, between sRANKL and FKS518. The polyclonal positive control was used to monitor consistency in assay performances.

Black dotted line: Samples were frozen prior to analysis to better represent study samples conditions.

Abbreviations: CCP = Confirmatory cut point; sRANKL = Soluble Receptor activator of nuclear factor kappa-Β ligand; PC = Positive control.

These results clearly demonstrate that the false-positive sRANKL signal in the screening ADA assay cannot be differentiated from a specific result using competitive inhibition with FKS518 in the confirmatory assay.

3.2. Impact of acid dissociation on tolerated target levels

Acid dissociation is a routinely employed technique in ligand binding assays to dissociate ADA and target complexes, improving or reducing drug interference [18,19]. However, in some cases, the acidic conditions can also aggravate target interference by disrupting the drug–target complex and by releasing accumulated target, thereby increasing the concentration of free targets. To verify this hypothesis, we spiked negative control samples with varying concentrations of sRANKL (up to 20,000 pg/mL), and these samples were evaluated for target interference in the presence and absence of acid dissociation.

The results have shown that, in the presence of 0.8 to 6.0 µg/mL of FKS518, concentrations of sRANKL up to 20,000 pg/mL did not cause false positive results when acid dissociation was not employed, with the only exception of sRANKL at 20,000 pg/mL, in the presence of 0.8 µg/mL of drug (Figure 3). On the other hand, false positive signals were observed when the samples were pre-treated with acid, in the presence of concentration of sRANKL as low as 313 pg/mL.

Figure 3.

sRANKL interference in the ADA assay without acid dissociation.

In the method without acid pretreatment, sRANKL interference was evaluated in the confirmatory assay by spiking increasing concentrations of sRANKL up to 20,000 pg/ml, in undiluted normal human serum, in the presence of relevant concentrations of FKS518. FKS518 was added to sRANKL spiked samples at 0.0, 0.8, 2.0, 6.0 µg/mL, and incubated for 2 h at room temperature to ensure complexes formation between sRANKL and FKS518. Samples were frozen prior to analysis to better represent study samples conditions.

Black dotted line: CCP determined during validation (assay with acid dissociation), shown as information only. The cut points (SCP and CCP) to be applied in the assay without acid dissociation were not calculated at this stage of assay development, therefore comparison should be made only based on % of inhibition data.

Abbreviations: CCP = Confirmatory cut point; sRANKL = Soluble Receptor activator of nuclear factor kappa-Β ligand.

Based on these results, we concluded that acid dissociation, which was employed to mitigate denosumab/FKS518 interference, also caused the release of denosumab/FKS518-bound sRANKL and produced false positive results (see Figures 2 and 3).

3.3. Target interference evaluated in the presence of preformed complexes (ADA- FKS518; sRANKL-FKS518), mimicking clinical samples conditions

When treated with denosumab/FKS518, patient sera may have different levels of denosumab/FKS518, ADA, sRANKL, and their immune complexes (ADA-drug, sRANKL-drug). To mimic clinical samples conditions, mock samples containing varying amounts of ADA and/or FKS518 and/or sRANKL in normal human serum were prepared. The inhibitory effect of FKS518 on the true ADA response as well as the interfering sRANKL response was examined.

In the presence of relevant concentrations of FKS518 and sRANKL concentrations ≥500 pg/mL, the samples which had signals below or close to the cut point have the potential to be wrongly classified as ADA positive because of sRANKL interference (Figure 4, panel a and b). The additive signal coming from sRANKL interference can change the ADA status from negative to positive, leading to inaccurate reporting of ADA incidence.

Figure 4.

Target interference tested with preformed complexes, mimicking clinical samples conditions.

To mimic samples conditions from patients treated with denosumab, sRANKL interference was evaluated in the confirmatory assay, in the method with acid pretreatment, by spiking increasing concentrations of sRANKL, in undiluted normal human serum, in the presence of relevant concentrations of FKS518 and positive control (A = Negative control; B = Low Positive Control; C = Mid Positive Control). FKS518 at 0.0, 0.8, 2.0, 6.0 µg/mL was added to the spiked samples with sRANKL and positive control. The maximum level of sRANKL tested was fixed at 5000 pg/mL, corresponding to the maximum signal level observed during sample analysis of FKS518 clinical studies. Samples were incubated for 2 h at room temperature to ensure complexes formation and were frozen prior to analysis to better represent study samples conditions.

Black dotted line: CCP.

Abbreviations: CCP = Confirmatory cut point; sRANKL = Soluble Receptor activator of nuclear factor kappa-Β ligand.

In contrast, at the mid-positive control (MPC = 100 ng/mL) level (Figure 4, panel c), which corresponds to the threshold of clinically relevant ADA [20,21], ADA responses are well above the cut point and any additive signal coming from sRANKL interference will not change the reported ADA status.

As shown in Supplementary Tables S2 and S3, across both studies (FKS518-001 and FKS518-002), the maximum mean signals of ADA-positive samples were consistently below the signals generated by the LPC (78 ng/mL). Only about 1.9–2.2% of clinical samples exceeded the LPC signal. Similarly, mean % inhibition values for study samples were substantially lower than those of the LPC, with only 2.5–3.3% of the samples exceeding the LPC % inhibition. These findings, together with evidence that observed signals are driven by sRANKL interference rather than true ADA responses, indicate that ADA levels are likely negligible and clinically irrelevant.

3.4. Mitigation of sRANKL interference with addition of OPG

The aim was to find a strategy to mitigate sRANKL interference without compromising drug tolerance (i.e., retaining the acid dissociation step). One approach employed to mitigate sRANKL interference in the ADA assay is the use of competitive binding of a molecule that recognizes the same epitope on sRANKL. OPG was identified as a potential candidate to mitigate target interference, since the epitope binding site of denosumab/FKS518 overlaps with the major binding sites of OPG on RANKL and OPG has high affinity to sRANKL [22]. Therefore, OPG could displace the binding of sRANKL to denosumab/FKS518, allowing unequivocal detection of positive ADA samples. It was proven that addition of OPG at the neutralization step, following acid dissociation, abrogates the signal induced by spiked sRANKL in mock samples (Figure 5).

Figure 5.

Mitigation of target interference by addition of OPG.

Samples spiked with sRANKL at 5000 pg/mL were pre-incubated with and without FKS518 to allow FKS518-sRANKL complexes formation. To mitigate target interference, preformed complexes were then incubated with a molar excess of OPG, added at the neutralization step, following acid dissociation. Samples were incubated for 2 h at RT to reach the equilibrium and then assessed in the screening (A) and confirmatory assay (B). Black dotted line: SCP (A) and CCP (B).

Abbreviations: SCP = Screening cut point; CCP = Confirmatory cut point; sRANKL = Soluble Receptor activator of nuclear factor kappa-Β ligand; OPG = Osteoprotegerin; LPC = low positive control; MPC = Mid positive control.

Results have also shown that OPG does not interfere in the assay, since signal and % of inhibition of LPC and MPC, in presence or absence of drug, were similar with or without addition of OPG (Figure 5, panels a and b), suggesting that the assay characteristics (sensitivity, drug tolerance and potentially CP) might not be substantially affected by the addition of OPG. Therefore, a partial validation was sufficient to confirm the cut points, sensitivity, LPC, precision, drug tolerance, and selectivity (refer to supplementary Table S1).

Subsequently, to test the above-mentioned hypothesis, the modified ADA assay, with addition of OPG at 60 ng/mL, was used on a selected subset of a subset of 50 anonymized samples from the study FKS518-002, originally identified as ADA positive. The samples were selected by the bioanalytical laboratory before unblinding the database, to cover a wide range of measured drug concentrations and ADA signals magnitude and included samples from both US-Prolia and FKS518 arms. In addition, all samples with the highest titer levels were included (4 samples with titers = 800). It was proven that, when re-tested with addition of OPG, all the previously positive samples returned signals below the cut point (data not shown).

Therefore, it appeared likely that the presence of elevated sRANKL levels during sample analysis was responsible for the false positive signal in the assay, by creating a target-mediated bridge with the biotin and ruthenium-labeled drug forms, leading to most of the samples being wrongly reported as ADA positive.

3.5. Two-step specificity assay

Consequently, the immunogenicity testing strategy was adapted by introducing a two-step Specificity Assay (Tier III) to confirm the specificity of the “potential positive” ADAs detected in the screening and confirmatory assay (Tier I and Tier II). Results of assay validation showed similar drug tolerance and sensitivity, and other assay parameters, between the initial and the modified ADA method.

The final immunogenicity testing scheme and interpretation of results for FKS518 clinical studies is illustrated in the Graphical Abstract.

3.6. Clinical sample analysis

In the comparative PK study FKS518-001 and efficacy and safety study FKS518-002, samples analysis with the initial ADA assay (Tier I and II) resulted in nearly all subjects being overall ADA positive (i.e., having at least one positive ADA result after dosing): 96.6% ADA positive subjects in FKS518-001 study and 95.5% ADA positive PMO patients in FKS518-002 study. The proportion of ADA positive patients was similar between FKS518 and US-Prolia, the ADA responses were transient, and the ADA titers were low. These unusually high ADA positive incidences were not accompanied by any alteration in PK/PD data, increased frequency or severity of adverse events, or loss of efficacy (Table 1).

Table 1.

Overall ADA incidence in study FKS518-001 and FKS518-002 (Safety Analysis Set) with initial ADA method and after implementation of a specificity assay (Tier III).

	ADA incidence % (n/N*)
	Initial ADA method (Tier I and II)			After specificity assay (Tier III)
	FKS518	US-Prolia /FKS518	US-Prolia	FKS518	US-Prolia /FKS518	US-Prolia
FKS518-001 (Week 0 to 40)	96.2 (101/105)	N/A	97.1 (102/105)	0 (0/105)	N/A	0 (0/105)
FKS518-002 (Week 0 to 78)	96.0 (263/274)	98.4 (122/124)	92.1 (140/152)	1.8 (5/274)	1.6 (2/124)	3.9 (6/152)

Abbreviations: ADA = Antidrug Antibody; n = Number of subjects with Positive Status; N* = Number of Subjects with an ADA valid Result; N = Number of subjects in the Safety Analysis Set. Note: Overall is determined across all time points except Baseline (pre-dose). The ADA status is defined as positive if the subject has at least 1 positive post-dose result in any time during this period.

The potential positive ADA samples from Tier II were reassessed in the ADA specificity Tier III and a dramatic reduction of the positive rates was observed: all samples from the comparative PK study FKS518-001, which were previously classified as ADA positive, were found to be negative. For the efficacy and safety FKS518-002 study, the ADA incidence was found to be low (≤3.9%), and comparable between treatment arms. In addition, the immunogenicity profile was not affected by the transition at week 52 (Table 2).

Table 2.

ADA incidence by time point in study FKS518-002 (Safety Analysis Set) with initial ADA method and after implementation of a specificity assay (Tier III).

	ADA incidence in FKS518-002 study % (n/N*)
	Initial ADA method (Tier I and II)			After specificity assay (Tier III)
	FKS518 (N = 277)	US-Prolia /FKS518 (N = 124)	US-Prolia (N = 152)	FKS518 (N = 277)	US-Prolia /FKS518 (N = 124)	US-Prolia (N = 152)
Overall	96.0 (263/274)	98.4 (122/124)	92.1 (140/152)	1.8 (5/274)	1.6 (2/124)	3.9 (6/152)
Week 0	1.1 (3/276)	0	3.3 (5/152)	0.4 (1/276)	0	0
Week 2	2.2 (6/267)	2.5 (3/121)	1.3 (2/149)	0	0	0
Week 4	6.0 (16/268)	7.4 (9/122)	8.2 (12/146)	0	0	0
Week 8	34.7 (93/268)	35.0 (43/123)	38.7 (55/142)	0	0	0.7 (1/142)
Week 12	65.3 (175/268)	70.7 (87/123)	68.3 (97/142)	0.7 (2/268)	0.8 (1/123)	1.4 (2/142)
Week 26	39.8 (106/266)	33.3 (41/123)	38.2 (52/136)	0	0	0
Week 32	22.6 (57/252)	24.4 (29/119)	18.9 (25/132)	0	0	0
Week 40	74.2 (187/252)	81.8 (99/121)	83.7 (108/129)	0.4 (1/252)	0	0
Week 52	46.1 (117/254)	42.7 (53/124)	43.0 (55/128)	0	0	1.6 (2/128)
Week 64	72.1 (173/240)	74.6 (91/122)	73.0 (89/122)	0.8 (1/240)	0.8 (1/122)	0.8 (1/122)
Week 78	47.3 (116/245)	46.3 (56/121)	36.9 (45/122)	0	0	0.8 (1/122)

Abbreviations: ADA = Antidrug Antibody; n = Number of subjects with Positive Status; N* = Number of Subjects with an ADA valid Result; N = Number of subjects in the Safety Analysis Set.

Note: Overall is determined across all time points except Baseline (pre-dose). The ADA status is defined as positive if the subject has at least 1 positive post-dose result in any time during this period.

This low incidence is consistent with what has been historically reported in the literature regarding the immunogenicity of denosumab, as well as that described in the US Prescribing Information (USPI) and Summary of Product Characteristics (SmPC) (Prolia USPI, 2025 [23]; Xgeva USPI, 2025 [24]; Prolia SmPC, 2025 [25]; Xgeva SmPC, 2025 [26].

Overall, these results indicated that the majority of samples detected in the initial ADA assay (TierI/II) were false positives due to sRANKL interference.

4. Conclusions

Comparative immunogenicity evaluation is essential for biosimilar development and regulatory approval. Generating reliable results requires a well-designed screening assay, followed by a selective confirmation assay, and a rigorous assessment of potential target interference, particularly for therapeutic antibodies with soluble multimeric targets. Importantly, target interference assessment should employ target concentrations far exceeding those reported in the literature for the relevant disease indication (at least 100-fold), accounting for post-treatment increases driven by the formation and accumulation of target-drug complexes.

The initial immunogenicity assessment of FKS518, a biosimilar to denosumab (Prolia®/Xgeva®), revealed unexpectedly high ADA incidence rates in both Phase 1 and Phase 3 clinical studies. These elevated rates were not associated with clinical consequences, such as altered pharmacokinetics, increased adverse events, or reduced efficacy, and were comparable between treatment arms. Further investigation revealed that the elevated ADA signals were due to interference from sRANKL, a multimeric target capable of forming nonspecific bridges in the bridging immunoassay format, particularly following acid dissociation.

To address this issue, two options were considered: complete ADA assay revalidation and reanalysis of all samples or implementation of an additional specificity tier and analysis of only previously positive ADA samples in the additional specificity tier. The latter was selected as scientifically robust and the most operationally feasible solution. A revised ADA testing strategy was developed, incorporating an OPG-based specificity tier, to block sRANKL and confirm true ADA-positive signals. This approach successfully mitigated interference while preserving assay sensitivity and drug tolerance, without requiring full revalidation of the ADA assay. Reanalysis of clinical samples using the refined method resulted in a dramatic reduction in ADA incidence, aligning with historical denosumab data and meeting regulatory expectations for denosumab immunogenicity.

This strategy offers a biologically relevant and effective alternative to previously reported mitigation approaches, such as anti-RANKL antibody use (Miller et al. [16]). By leveraging OPG, a natural decoy receptor with high affinity for RANKL, the method minimizes risk of unintended assay interactions, maintains bridging assay integrity, and provides a practical solution.

While effective in this context, this solution is not intended as an universal approach; for future programs, it is recommended that potential target interference be assessed early in assay development using target concentrations substantially above physiological levels (at least 100-fold), considering possible post-treatment accumulation and formation of target–drug complexes. Mitigation strategies should ideally be integrated directly into the screening assay rather than applied as an additional specificity tier. Furthermore, we advise that acid dissociation pretreatment should not be universally applied to improve drug tolerance in ADA assays without evaluating its bioanalytical risks versus benefits. Where applicable, a similar mitigation approach and testing strategy may be adapted to overcome soluble target interference and ensure robust immunogenicity assessment in biosimilar development.

Funding Statement

All work described in this paper was funded by Fresenius Kabi SwissBioSim GmbH.

Article highlights

• Following completion of the sample analysis from Phase 1 and Phase 3 denosumab/FKS518 clinical studies, the ADA incidence was found to be significantly higher compared to the originator historical data and recent studies.

• Dimeric or multimeric forms of soluble drug targets, such as sRANKL, can interfere with ADA assays by bridging the ADA assay reagents, potentially resulting in false positive results.

• Treatment with therapeutic monoclonal antibodies, like denosumab/FKS518, may increase target levels due to target upregulation and/or release from target/drug complexes during the acid-dissociation step.

• An investigation confirmed the suspected interference of denosumab target (sRANKL) in the ADA assay, leading to false positive signals and overreporting of ADA incidence in both treatment groups in both clinical studies.

• The mitigation of soluble target interference is essential in order to obtain reliable ADA assay results, which are essential for the accurate assessment of comparative immunogenicity for biosimilars.

• By introducing the RANKL decoy receptor, OPG, into the procedure, target interference was successfully mitigated, thereby improving the assay’s specificity and accuracy.

• This alternative approach, with the addition of OPG, was validated and incorporated as a specificity tier in the immunogenicity testing strategy.

• Clinical study samples were reanalyzed using the new specificity tier, resulting in very low true ADA incidence, confirming the low immunogenicity profile of denosumab/FKS518.

Author contributions

Anna Lucia Buccarello: conceptualization, data analysis and interpretation, project administration, writing – original draft.

Adriano L.S de Souza: project administration, conceptualization, data analysis and interpretation, writing – review & editing.

Martin Ullmann: conceptualization, data analysis and interpretation, supervision, writing – review & editing.

Corinne Petit-Frere: conceptualization, data analysis and interpretation, project administration, supervision, writing – review & editing.

Disclosure statement

All authors are employees of Fresenius Kabi SwissBioSim GmbH and have no other 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 manuscript apart from those disclosed.

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

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Ethical declaration

The authors state that they have obtained appropriate Institutional Review Board (IRB) approval and have followed the principles outlined in the Declaration of Helsinki for clinical studies. In addition, written informed consent has been obtained from all participants involved prior to their inclusion in the study.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/17576180.2025.2607080