Development and validation of a method for quantitation of CTx-1 for the assessment of biosimilarity of FKS518, a denosumab biosimilar

ABSTRACT

Fresenius Kabi has developed FKS518, a fully human monoclonal antibody biosimilar to denosumab. The clinical development program included two randomized comparative trials: a PK study in healthy volunteers and a safety and efficacy study in osteoporosis patients. The demonstration of similarity and equivalence between FKS518 and reference denosumab included the quantitation of the serum biomarker C-terminal cross-linking telopeptide of Type 1 collagen (CTx-1), a well-established marker of bone resorption. This paper details the development, validation, and analytical performance of the method for CTx-1 quantitation, emphasizing how the Context of Use (CoU) shaped the validation requirements. Application of the method in samples from the pharmacokinetic (PK) equivalence study allowed demonstration of pharmacodynamic (PD) similarity between FKS518 and Prolia. This publication also addresses the use of endogenous serum control (ESC) samples in monitoring assay performance. A retrospective analysis indicated that applying more flexible ranges and/or omitting ESC-based criteria for run acceptance would not have substantially changed the CTx-1 results. While recognizing the value of ESCs for stability and trending analysis and the importance of rigorous biomarker method development and validation in the assessment of biosimilarity, this paper fosters the discussion whether run acceptance based on tight ESC acceptance limits is always necessary.

Plain Language Summary

This study set out to create and prove a reliable lab test for measuring CTx-1, a marker that shows how quickly bone is breaking down in the body. Researchers used this test to compare FKS518, a new medicine similar to denosumab, with the original drugs Prolia®. Blood samples from people in clinical trials were checked using this method, following strict quality rules. The results showed that FKS518 and the original drugs gave almost the same CTx-1 results, meaning FKS518 works in a very similar way to Prolia®.

The paper also looked at using a special type of sample, called endogenous serum control (ESC), to double-check the accuracy of the test. Normally, there aren’t strict rules from health authorities about using ESC samples in these studies. The researchers checked whether being less strict with ESC rules would have changed the results. They found that loosening the rules would not have made a difference in the study’s outcome. This suggests that some extra quality checks may not always be needed. In summary, the study stresses the importance of having good methods to measure markers like CTx-1 and suggests that certain quality control steps could be made simpler without changing the results.

1. Introduction

Fresenius Kabi developed FKS518, a fully human monoclonal antibody of the IgG2 subtype and biosimilar to Prolia® (60 mg denosumab) and Xgeva® (120 mg denosumab). Denosumab’s target receptor activator of nuclear factor kappa-Β ligand (RANKL) mediates osteoclast maturation, activation and/or survival, processes that are exacerbated in osteoporosis patients. Denosumab is indicated to increase bone mass and strength in patients with osteoporosis or at high risk for fracture. Denosumab is also indicated to prevent osteolysis and skeletal-related events in patients with multiple myeloma and in patients with bone metastases from solid tumors. The clinical development of FKS518 consisted of a comparative pharmacokinetic (PK), pharmacodynamic (PD), safety, and immunogenicity study in healthy male subjects ([FKS518-001], Dryja et al., 2025) [1] and a comparative efficacy, PD, safety, and immunogenicity study in women with post-menopausal osteoporosis (PMO) ([FKS518-002, NCT04934072], Krecipro-Nizinska et al., 2025) [2]. Both studies used US-Prolia as the reference medicinal product. The current manuscript will focus on bioanalytical aspects and the necessary information to understand the assay Context of Use (CoU). For further details on trial design and conduct the reader is referred to the previously cited publications [1,2].

A method of quantitation of serum C-terminal cross-linking telopeptide of Type 1 collagen (CTx-1) [3] was applied to assess the PD response contributing to the demonstration of bioequivalence and similarity between FKS518 and Prolia.

Bone turnover markers comprise a series of protein or protein derivatives released during bone remodeling by osteoblasts or osteoclasts. On the C-terminal region of neighboring collagen molecules, lysyl residues participate in intermolecular cross-linking. Aspartyl residues are also present on this region and are subject to isomerization and racemization giving rise to the a and b isoforms. During collagen breakdown, these cross-links remain intact, and C-terminal fragments are released into the bloodstream [4,5]. CTx-1 has no described binding partners.

Procollagen type 1 N-terminal propeptide (P1NP) was also quantified as part of the FKS518 development program. Unlike CTx-1, which is marker of bone resorption, P1NP is a marker of bone formation [3] and was not applied as a co-primary endpoint in any study, following alignment with regulators. Given its different CoU, discussion about how the P1NP method was validated and applied is outside the scope of this manuscript. The reader is referred to other publications for a justification of the selection of CTx-1 and P1NP as relevant bone turnover biomarkers [6,7].

The present paper illustrates several aspects related to the development, validation, and sample analysis of the referred bioanalytical method and examines how the CoU guided the definition of the required extent of assessments. In addition, clinical results from the pivotal PK equivalence study demonstrate the biosimilarity between FKS518 and the reference denosumab and illustrate the method’s in-study performance. Finally, this publication explores the applicability of endogenous serum control samples in the monitoring of the reliability and consistency of biomarker assays and presents a discussion on the application of run acceptance criteria based on ESC ranges.

1.1. Industry approach to biomarker assay development & validation

Due to the lack of a biomarker-specific guidance from health authorities, validation of bioanalytical methods for the analysis of PD biomarkers has been historically carried out by extrapolation from guidances written for PK assays. The validation of bioanalytical methods for quantitation of biomarkers was extensively discussed at the 2015 Crystal City VI meeting. A publication summarizing the proceeding of this workshop [8] emphasizes that it is not only the nature of the analyte (biomarker versus drug) that should define the extent and rigor necessary for the validation and reporting of a bioanalytical method, but also its CoU. Assays seeking to produce evidence required to establish and confirm decision points, such as outcomes in clinical trials, are expected to undergo full validation to ensure that the level of assay characterization, its performance and reliability match the importance of its intended application [9]. The concept of CoU has been explored in more detail in a few subsequent publications as a critical determinant of the scope of development and validation assessments [10–12]. These considerations have guided the industry’s approach to satisfy the requirements for a biomarker method to be considered fit-for-purpose [8,9].

An extensive list of assessments and rigorously defined passing criteria were applied to the validation and CTx-1 sample analysis. This was consistent with the inclusion of the CTx response as a secondary endpoint in the comparative PK study (Dryja et al., 2025) [1] and as a coprimary endpoint in the Safety and Efficacy Study (Krecipro-Nizinska et al., 2025) [2].

1.2. CoU and critical aspects in development and validation

As a Context of Use (CoU) statement, the method of quantitation of CTx-1 was applied to assess the PD response contributing to the demonstration of bioequivalence and similarity between FKS518 and Prolia.

1.3. Assay sensitivity

Endogenous CTx-1 levels in healthy male subjects range from ~260–400 pg/mL on average [13,14] and ~510–750 pg/mL on average in women with PMO [14–16], which are the relevant populations for the assay validation and FKS518 clinical studies. In addition, following treatment with denosumab, major reductions in CTx-1 concentrations occur within 2 weeks up to 4–5 months in most patients, after which CTx-1 levels start to rise again, but without reaching predose concentrations, provided the approved frequency of administrations of 6 months is maintained [17]. The robust decrease and/or disappearance of CTx-1 from plasma underscores a transient interruption of biological processes mediated by RANKL, as a result of its neutralization by denosumab.

The baseline range expected in the healthy male and female PMO populations, the dispersion of those ranges, as well as their expected response to treatment with denosumab were carefully considered during method development. In this regard, the targeted sensitivity level was set to avoid missing baseline quantitations and to capture the recovery from the maximal effect following administration of FKS518 and the reference drug product. The desired method sensitivity was set to 50 pg/ml CTx-1, matching that of a method previously applied in clinical studies with the denosumab originator molecule [18].

1.4. Pre-analytical factors

The influence of the main intrinsic factors that could impact data variability is described below and was addressed via a carefully designed study protocol (See Section 2.3). Important circadian rhythm variability has been reported for CTx-1 concentrations in plasma, with an amplitude of 40–66% around the daily mean [19,20]. The circadian variability in CTx-1 measurements is reduced in fasting conditions. Other possible sources of variability such as menstrual cycle, seasonal period, exercise, and lifestyle have been described elsewhere [3,7,21].

The long-term analyte stability in the selected matrix was an important aspect of method validation since the two FKS518 studies were run almost in parallel. The sample analysis from the comparative efficacy and safety study was only initiated once the analysis of the samples from the comparative pharmacokinetic study was complete. Long-term storage seemed feasible based on a previous report where long-term stability is documented up 3 years at −20°C, −70°C, or −80°C [18,22].

Long term CTx-1 stability (at −20°C and at −80°C) has been shown to be comparable between serum and plasma. The stability of CTx-1 in promptly processed serum at benchtop (room temperature [RT]) and refrigerated (4°C) conditions is also satisfactory, provided that blood samples are processed within a few hours after sampling [7,20,23–25]. Serum was chosen as the matrix for FKS518 studies, as it had been previously utilized during the development of the originator [17,18,26–29].

1.5. Additional important considerations

A particular feature of biomarker assay development and validation is the use of ESC to monitor stability, kit lot-to-lot variability, and/or eventual drifting of the assay. Since the ESC measurements provide a means to monitor reproducibility over time in samples that are identical in nature to study samples, the incurred sample reanalysis, which are routinely employed for PK assays, becomes redundant and is not generally applied for biomarkers assays [30].

Standard solutions were prepared using components from the same kit lot during assay development, validation, and part of sample analysis. Due to impossibility of sourcing the same kit lot for completing long-term stability testing and analysis of samples from the pivotal comparative PK study reported in this manuscript, two additional kit lots were obtained and successfully bridged. The bridging experiments consisted of a comparison of the results obtained for the low mid and high QCs, as well as the applicable ESC levels, with the original and new kit lot.

In the absence of a universally adopted reference standard for CTx-1, the method here described falls into the relative quantitative category [10], and the term “accuracy” better refers to “relative accuracy” [31] or “Trueness” [11].

2. Methods

2.1. CTx-1 method development

Two different platforms were initially considered for CTx-1 analysis based on commercially available kits, namely enzyme-linked immunosorbent assay (ELISA) or electrochemiluminescence immunoassay (ECLIA), with the latter intended for use on Cobas® immunoassay analyzers. The CTx-1 serum CrossLaps® ELISA kit (AC-02F1, Immunodiagnostic Systems Ltd, UK) served as a starting point for the development of the CTx-1 method. This kit was chosen since it allowed, due to platform availability, for CTx-1 to be quantified in the same laboratory where other bioanalytical assays were run, including PK and immunogenicity assays.

The CTx-1 method is specific for homodimers of the amino acid sequence of EKAHD-β-GGR from type I collagen, where the aspartic acid residue (D) is β-isomerized [32]. The choice for a β-selective CTx-1 kit is important since the β:α isoform ratio increases with age [33] and b isoform is the most prevalent in adults, which makes it appropriate for the target population (PMO).

The kit uses the desalted urinary antigen of human origin as the calibration standard. The claimed detection limit is 0.020 ng/mL (20 pg/mL), and the proposed calibration range is 0.178–2.494 ng/mL.

In addition to the reagents described in the supplier’s package insert, the list of critical reagents included StreptaWellTM, high bind streptavidin-coated plates (Cat# 11989685001, Roche), Bulk CTx-1 Reference Standard (AC-0201 F FUJ, IDS), and Low and high ESCs prepared from pooled human serum. The bulk CTX-1 reference standard was used to limit variation due to different batches/preparation of the reference standard.

2.2. CTx-1 method validation

Validation was divided into two phases, namely core validation and non-core validation. During core relative accuracy and precision (A&P) experiments, runs are accepted based solely on the calibration curve performance, and an individual run may only be rejected for assignable cause or documented technical error. Relative A&P runs were conducted prior to any other validation experiments for minimum of ≥6 runs over ≥3 days by ≥2 analysts. The remaining validation experiments (non-core) were accepted based on calibration curve and quality control (QC) performance. Data from experiments that did not meet assay acceptance criteria were excluded from consideration in validation conclusions. The most important validation assessments are detailed below.

2.2.1. Quantitation range

The required sample volume for this method is 100 uL (50 uL/well, plated in duplicate). This assay was validated with the following calibrators: 50, 70, 98, 196, 392, 653, 875, and 1130 pg/mL. The QC levels tested were as follows: 50 (LLOQ), 70 (back up LLOQ), 247, 494, 988, and 1130 (ULOQ) pg/mL.

The calibrator concentration versus instrument signal curve was fitted following a four-parameter logistic regression at 1/response2 weighing.

The calibrator performance was calculated to determine core and non-core run acceptance during method validation. For core runs, the bias for ≥75% of calibrators (at least six non-zero calibrators) must be within ±20.0% difference from theoretical (DFT) (25.0% at LLOQ and ULOQ) and the coefficient of variation (CV) less than 20.0% between wells (25.0% at LLOQ and ULOQ). If LLOQ or ULOQ CAL were deactivated, the run was rejected and repeated. The blank response must be < LLOQ. For non-core runs ≥75% of calibrators (at least six non-zero calibrators) must be within ±20.0% DFT (25.0% at LLOQ and ULOQ). The reagent blank must be < LLOQ.

2.2.2. QC relative accuracy and precision

Relative A&P runs were evaluated by analyzing QCs prepared at 50.0, 70.0, 247, 494, 988, and 1130 pg/mL by fortifying the calibrator buffer with the reference standard. Intra- and inter-assay A&P were analyzed by Analysis of Variance (ANOVA) and the total error (TE) reported.

For core runs the %CV between duplicate responses must be ≤20.0% for QCs (≤25.0% for LLOQ and ULOQ) and relative accuracy within ±20.0% DFT for QCs (±25.0% for LLOQ and ULOQ). The %TE must be ≤30.0% for QCs (≤40.0% for LLOQ and ULOQ).

In non-core validation runs, QC A&P were calculated to determine run acceptance. The %CV between duplicate responses must be ≤20.0% (≤25.0% for LLOQ and ULOQ) and 2/3 of QCs and at least 50% for each level must be within ±20.0% of theoretical.

2.2.3. Use of endogenous serum controls

ESCs were prepared by pooling pre-screened human serum from commercial sources and used to assess long-term stability, selectivity, as well intra- and inter-assay precision. Two different levels (high and low) were applied to reflect the expected levels in study samples.

During core validation runs, the %CV between duplicate responses for ESCs must be ≤20.0%. There was no acceptance criterion based on relative ESC accuracy. The relative accuracy results for all acceptable validation runs were used to calculate acceptance ranges to be later applied in stability testing and to monitor assay performance during sample analysis. Prior to calculating ranges, any eventual outliers are removed with the Tukey’s outlier test. ESC ranges were obtained using either formula below (to include the greater of the two calculation limits results):

Range = Mean Concentration (pg/mL) ±2×SD

… or:

Range = Mean Concentration (pg/mL) ±20%;

where SD = standard deviation.

New ESCs qualification runs comprised analysis of each new level at n ≥ 3 across at least six runs conducted within 10 days. The concentrations obtained with the pair of ESC pools included in validation runs (ESC1 and ESC2) were deemed unsuitable. Then, a second ESC pair (ESC3 and ESC4) was prepared and qualified in eight runs that were closely timed (see Section 2.3.3 and 3.2.4). The variability of these runs might not capture the true assay variability, and derived ranges had the potential to cause an excessive and unnecessary number of failed runs. Applying the hybrid rule described above, which incorporates the ±20% limit in addition to the ±2×SD rule, helped to reduce this risk.

2.2.4. Biological variability

To evaluate biological variability, commercially sourced samples from 44 normal healthy individuals (22 males and 22 females) and from 21 disease state from women with PMO were tested in duplicate at n = 1. Data obtained from the biological variability evaluation was used to select individuals for further assessments including selectivity, hemolysis, lipemia, and parallelism.

2.2.5. Parallelism

Serum samples from five PMO subjects with high levels of CTx-1 (determined during the biological variability assessment) were selected for parallelism evaluation. Each sample was analyzed undiluted and serially diluted in assay buffer at 1.5-, 2-, and 3-fold dilutions in duplicate at n = 1, along with a dilution control prepared at the MQC level in assay buffer. Parallelism was declared if the overall precision (%C.V.) of the calculated concentrations (value ´ dilution factor) for all dilutions were ≤30%.

2.2.6. Hemolysis effect

Five individual samples with endogenous CTx-1 concentrations within the LLOQ and LQC range were used to prepare hemolyzed matrices by spiking at 5% v/v with fully hemolyzed whole blood. Baseline and hemolyzed matrices were assessed on the same plate. For acceptance, the %CV for the mean must be ≤20.0% and the %difference of mean hemolyzed sample value must be within ±20.0% from mean control baseline value.

2.2.7. Lipemic effect

The effect of lipemia on the quantitation of CTx-1 was evaluated by analyzing five individual human serum samples with endogenous CTx-1 concentrations between the LLOQ and LQC levels as non-lipemic (baseline) and lipemic via the addition of >300 mg/dL triglycerides. For acceptance, the %CV for the mean must be ≤20.0% and the %difference of mean hemolyzed sample value must be within ±20.0% from mean control baseline value.

2.2.8. Long-term stability

The effect of freezing the analyte in biological matrix and storing samples for the duration of the validation is evaluated by analyzing ESCs stored in a cryofreezer (−80°C). In addition, since this method was also applied in a multicenter confirmatory efficacy and safety study, it was also necessary to demonstrate stability at −25°C to cover the few days or weeks during which some study samples were stored at clinical sites, some of which were not equipped with −80°C freezers, prior to their shipping to the central laboratory. At each stability assessment timepoint, an ESC pair (low and high levels) was analyzed versus freshly prepared calibration standards and mid-level quality controls in each run.

For acceptance, at least three values must be available to calculate stability statistics for each level tested, the %CV for the mean must be ≤20.0%, and the ESC recovery must fall within the acceptance range determined from validation. The low- and high-level quality controls were stored at 2–8°C and used within expiration limits in each run.

2.2.9. Other validation assessments

In addition to the validation assessments described above, experiments were conducted to evaluate freeze/thaw (F/T) stability up to six F/T cycles, stability in thawed matrix (for 12 and 24 h), selectivity (using spiked matrix), exogenous interference (by RANKL and osteoprotegerin [OPG] at 1000 pg/mL), drug interference (at 6 mg/mL), and prozone or hook effect.

2.3. Pharmacodynamic assessments in healthy male volunteers (FKS518-001)

FKS518-001 was a double-blind, randomized, 2-arm, single-dose, parallel-group study to compare PK, PD, safety, tolerability, and immunogenicity of FKS518 (proposed denosumab biosimilar) with US-Prolia (denosumab) in healthy male subjects. The study had a total duration of up to 44 weeks, including a screening period of up to 4 weeks prior to the investigational product (IP) administration on Day 1, and a follow-up period of 40 weeks, consisting of 1 week of confinement in the clinic from Day 1 to Day 6 and 16 ambulatory visits up to Day 274.

The study was conducted with written and dated approval from the locally competent Independent Ethics Committee (Bioethics Committee of the District Medical Board in Warsaw), and in accordance with ICH Harmonised Guideline for Good Clinical Practice E6 (R2) [34], requirements for the conduct of clinical studies as provided in the EU Regulation 536/2014 [35], the general guidelines indicated in the Declaration of Helsinki [36], and all applicable regulatory requirements. All subjects provided written informed consent before any study-related activities were performed.

The sampling schedule in the healthy male volunteers’ study extended up to 40 weeks, longer than the dosing interval of 26 weeks approved for the reference product, with the purpose to better characterize the recovery phase of CTx-1 and therefore allow the detection of any eventual small differences in potency and/or exposure.

Blood samples for the bone turnover biomarker CTx-1 were collected at screening, predose, and 4 h postdose on Day 1, and on Days 2, 3, 15, 29, 57, 71, 85, 99, 113, 127, 183, 253, and 274. To limit inter- and intra-patient variability caused by the circadian influence on CTx level that is exacerbated by a non-fasting status (see Section 1.2.2), blood was harvested between 08:00 h and 12:00 h in the morning, after an overnight fast.

A few PD parameters were calculated for CTx-1 and included as secondary endpoints: (1) the percent change from baseline (%CfB) at all time points postdose; (2) the maximum %CfB (%CfBmax); (3) the area under the effect curve up to week 40 for %CfB (AUEC0-W40%CfB) in serum; and (4) the Net AUEC0-W40%CfB, for which any possible rebound effect where biomarker concentrations rise above baseline was subtracted from AUEC below the baseline. The exposure–response relationships of the %CfB versus denosumab concentrations were included as an exploratory objective.

2.4. CTx-1 sample harvest and analysis

Biomarker samples were appropriately handled so as to maintain their integrity at all steps in alignment with recommendations for sample processing as described elsewhere [7] and also in respect to the outcome of the validation.

Blood samples were processed to serum and stored at −80°C until analysis. Sample analysis for CTx-1 was performed by batching full patient profiles.

Each assay plate comprised a blank level consisting of the calibrator buffer, a set of calibration standards at the nominal concentrations of 50.0, 70.0, 98.0, 196, 392, 653, 875, and 1130 pg/mL, two sets of QCs (comprising low, mid, and high levels, each, with concentrations defined at each kit lot), one pair of ESCs (low and high) and up to 31 study samples. Of the two sets of QCs used, one set was plated preceding the study samples, and the other at the three last pairs of wells. All assay conditions and/or samples were plated in duplicate.

For run acceptance criteria during CTx-1 sample analysis, a minimum of six calibration standards must show precision and bias within ≤20% CV and 20% DFT, respectively. In addition, four of the six QCs and at least one QC at each level must show precision and bias ≤20% CV and ≤20% DFT, respectively.

Concerning ESCs, both precision ≤20.0% CV and an ESC result within the pre-defined range were applied as run acceptance criteria.

3. Results

3.1. Development

To obtain the desired sensitivity of 50 pg/mL, several improvements were made to the assay procedure described in the supplier’s package insert. The changes included the following: (a) substitution of the plates by Roche StreptaWellTM High Bind streptavidin-coated plates (cat# 11989685001); (b) inclusion of a blocking step prior to adding calibrators, QC, or samples; (c) extension of the number of calibrators from five to eight and adjustment of their range; (d) adjustment of antibody concentrations; (e) inclusion of a mid quality control (MQC) level; and (f) use of an ESC pair in each run.

3.2. Assay validation

The assay validation comprised 24 separate assay runs, 19 of which met the acceptance criteria and were performed over 15 analysis days. The most relevant assessments are detailed below.

3.3. Standard curve and assay range

The lower and upper limits of quantitation were 50 and 1130 pg/mL, respectively. From the 19 runs performed during validation, 98.7% of calibrators (150 of 152) were within ±20% DFT and 100% of the LLOQ and ULOQ calibrators (38 of 38) were within ±25% DFT. A composite standard curve showing the mean ±SD instrument responses from all 19 passing runs is available as the supplementary Figure S1.

3.4. Parallelism

Parallelism testing was performed using commercially sourced serum samples with relatively high CTx-1 concentrations, ranging from 377 to 869 pg/mL. All samples quantified within the assay range at all tested dilutions. The recovery percent difference from baseline (%DfB) was ≤20% in all cases, except for one sample (sample #5) tested at threefold dilution (%DfB = 28.1). Parallelism assessment was deemed acceptable as the %CV for the dilution-adjusted results was well below the pre-defined acceptance criterion (≤30% CV) with 4.98, 9.05, 7.39, 3.76, and 11.1% for samples #1–5, respectively. The parallelism results support absence of a matrix effect, and confirm both the minimum required dilution (MRD) and the LLOQ of 50 pg/mL for the method. The dilution-adjusted result (in pg/mL) and the dilution-adjusted %recovery versus the neat (diluted to the assays’ MRD only) sample are available in the supplementary Figure S2(A,B), respectively.

3.5. QC and ESC inter-assay relative accuracy and precision

An analysis of Variance (ANOVA) was conducted on the QC results from eight core A&P validation Runs. The mean bias, interbatch precision (%CV), and total error (|mean bias|+ interbatch precision) were calculated for each QC pool. The mean bias was ≤14.4% at all levels and the interbatch precision was 17.2% at the LLOQ (50 pg/mL) or ≤10.6% at remaining levels. The total error was 30.7% at the LLOQ or ≤22.4% at remaining levels.

As described in Table 1, the ESC pair ESC1/ESC2 was used during assay validation. The intra- and interbatch precision was calculated for the ESC1 and ESC2 levels based on eight runs during which the ESCs were tested in triplicate. The intrabatch precision was ≤5.72% and ≤3.59% and the interbatch precision was 3.59% and 2.25% for ESC1 and ESC2, respectively. During validation, the relative ESC accuracy was not calculated since no concentration had yet been assigned. Likewise, the total error was also not calculable (see Section 2.2.3).

Table 1.

Summary of ESCs applied in validation, stability, and sample analysis runs.

	ESC level	Theoretical Conc. (pg/mL)	Range (pg/mL)	Application
1st pair
	ESC1	77.2	50.5 – 104	Assay Validation
	ESC2	201	161 – 242	Assay Validation
2nd pair
	ESC3	141	107 – 175	LT stability, in-study run acceptance & performance (FKS518-001)
	ESC4	359	287 – 431	LT stability, in-study run acceptance & performance (FKS518-001)

Abbreviations: LT = long term. All ESC levels described here were prepared by pooling human serum from individual donors (male and female) from commercial sources.

3.6. Stability in frozen matrix

ESC3 & ESC4 were selected to assess LTS and monitor assay performance, as their concentrations better reflected the expected CTx baseline concentrations (Table 1).

The results for stability testing using ESC3 and ESC4 following storage for 70 days at −25°C are shown in Figure 1(A) and for 72, 236, 239, 464, and 698 days at −80°C in Figure 1(B). The data met the acceptance criteria and demonstrated CTx-1 stability in frozen matrix up to 70 days at −25°C and up to 698 days at −80°C.

Figure 1.

CTx-1 long-term stability in serum.

C-terminal cross-linking telopeptide of Type 1 collagen (CTx-1) stability testing in serum stored at -25°C and -80°C after storage at different durations. Circles represent the observed concentrations for each replicate during stability assessment, with the bar placed at the mean. The dotted lines matching the respective endogenous serum control ESC color represent the theoretical concentrations and shaded areas indicate acceptance ranges defined during validation/ESC qualification runs as summarized in Table 1. The dotted gray line indicates the lower limit of quantitation (LLOQ) for the assay.

In addition to the assessments described above, selectivity testing demonstrated the absence of matrix interference as 9 out of 10 samples from healthy donors and 10 out of 10 samples from disease-state (PMO) donors fortified with 226 pg/mL CTx-1 met the acceptance criteria. The absence of exogenous interference was demonstrated as there was no effect of RANKL or OPG up to 1000 pg/mL on the quantitation of CTx-1. Quantitation of CTx-1 was not affected in the presence of up to 6 µg/mL of drug. There was no effect from hemolysis up to 5% fully lysed whole blood and no effect from lipemia (>300 mg/dL triglycerides) on the quantitation of CTx-1. Moreover, analyte stability was confirmed for up to 24 h at room temperature and six freeze/thaw cycles.

3.7. In study assay performance

Sample analysis for the FKS518-001 study was performed in 243 physical runs over approximately 6 months. Samples were stored for a maximum of 581 days between collection and analysis, within the demonstrated analyte stability in frozen matrix. The overall pass rate was 87.7% (213 accepted and 30 rejected runs). The most common reason for assay run failure was rejection due to unacceptable ESCs (17 out of 30 failed runs). Among those, 3 runs failed due to both range and precision criteria and 14 runs failed due to range only. There were no runs that failed exclusively due to ESC precision. Other reasons for failure included unacceptable QCs (8 out of 30), unacceptable calibration standards (2 out of 30), unacceptable matrix blank (2 out of 30), and chemist error (1 out of 30).

The mean bias for the back-calculated calibrator concentrations was within ±4.67% and the precision measured as the percent coefficient of variation (%CV) at all levels was ≤7.07%. For QCs, the mean bias and precision at all levels were within ±10.5% and ≤7.64%, respectively. The total error for QCs was ≤16.3% across all levels.

In addition, the ESC concentration results from all passing runs from the FKS518-001 study, and runs that failed due to ESC-based criteria are shown in Figure 2. ESC acceptance ranges determined are indicated in shaded areas in the graph. The mean bias was 3.04% and 4.06% and the precision was 9.22% and 6.7% for ESC3 and ESC4, respectively. The total error was 12.3% for ESC3 and 10.8% for ESC4.

Figure 2.

ESC performance in the FKS518-001 study.

Endogenous serum control (ESC) performance in the FKS518-001 Study. A low (ESC3) and a high ESC (ESC4) level were applied to monitor assay performance during sample analysis. The ranges defined during validation are indicated as an aid to the visual evaluation of assay performance. The mean ESC concentrations (dotted lines) and acceptance ranges (shaded areas) were 141 [107–175] pg/mL for ESC3 (in yellow) and 359 [287–431] pg/mL for ESC4 (in blue). ESC results outside the acceptance criteria for relative accuracy and/or precision are circled in red. For failed runs, the corresponding ESC levels that passed their individual acceptance criteria are circled in green. The ESC4 results for the failed run #112 lay outside the scale and were not plotted (concentration = 873 pg/mL, %CV = 77%). CV = coefficient of variation.

In Figure 2, ESCs that quantified outside the acceptance range are indicated by red circles. In 11 out 17 runs rejected due to unacceptable ESCs, only one of the two ESC levels was out of the acceptance range, in which case the other level still testing within the defined limits is circled in green. In all cases, study samples included in the run were reassayed, as per run acceptance specified in Section 2.3.1.

Overall, the results also indicate that the assay performed well, and no apparent drift was observed in ESC recovery during sample analysis.

3.8. Results from CTx sample analysis in study FKS518-001

Table 2 summarizes the reported mean baseline CTx-1 serum concentrations from the FKS518-001 study. The mean baseline concentrations were similar between treatment groups at 545 and 597 pg/ml for the FKS518 or reference denosumab groups, respectively. Unlike the various samples from commercial sources that showed undetectable CTx-1 levels during validation, all study subjects had detectable CTx-1 levels in the samples collected at the Day 1 pre-dose timepoint. This difference may be attributed to the standardized and controlled sample collection procedures implemented at the clinical site. Inadequate processing time and/or storage of the commercial samples used during validation may also have had an impact, although this cannot be ascertained.

Table 2.

Serum CTx-1 baseline levels.

	Mean (SD) (pg/mL)	Median (pg/mL)	Min – Max range (pg/mL)
FKS518 (n = 105)	597 (228)	578	159–1310
Reference denosumab (n = 104)	545 (215)	528	152– 1140

Baseline CTx-1 levels in adult white males based on the results from the FKS518-001 study. The reported results are from the PD Analysis Set. The overall mean age and range were 38.8 [28–55] years.

Following SC administration of 60 mg FKS518 or reference denosumab, quantifiable reductions in CTx-1 serum concentrations were observed as early as 4 hours post dose (1st postdose sampling timepoint) in nearly all subjects. The overall shape of CTx-1% change from baseline (%CfB) versus time profile was similar between the FKS518 and the reference denosumab treatment groups (Figure 3). Robust drops in CTx-1 concentrations were maintained between Day 15 (Week 3) and Day 99 (Week 15) reaching average maximal inhibitions at Day 57 (Week 9) in the FKS518 group (95.91% inhibition) and day 85 (Week 13) in the reference denosumab group (94.91% inhibition). At around Day 127 (Week 19), the mean %CfB CTx-1 started to gradually increase for both treatments, without completely returning to baseline levels, on average, until the end of the study at 40 weeks postdose.

Figure 3.

Pharmacodynamic response as assessed by the %CfB CTx-1 in serum.

Time-course of the pharmacodynamic (PD) response to FKS518 and reference denosumab as assessed by the percent change from baseline (%CfB) in C-terminal cross-linking telopeptide of Type 1 collagen CTx-1 in the FKS518-001 study. The CTx-1%CfB mean ± standard deviation (SD) is shown for the PD Analysis set (FKS518 [n = 105], reference denosumab [n = 104]). The planned timepoints for PB biomarker sample collection were screening visit (not shown), Day 1 predose and postdose (+4 hours), Day 2, Day 3 (Week 1), Day 15 (Week 3), Day 29 (Week 5), Day 57 (Week 9), Day 71, (Week 11), Day 85 (Week 13), Day 99, (Week 15), Day 113 (Week 17), Day 127 (Week 19), Day 183 (Week 27), Day 253 (Week 37), and Day 274 (Week 40). For calculation of the %CfB, postdose CTx-1 values below the lower limit of quantitation (LLOQ) of the assay (50 pg/mL) were replaced by zero, leading to a corresponding %CfB = -100% at the respective subject timepoint. Reproduced with permission from Figure 2(B) from Dryja et al. [1].

The geometric means and the 95% confidence intervals (CIs) for various PD parameters were calculated for the PD dataset as shown in Table 3. The similar means and overlapping 95% CIs indicated similarity between FKS518 and reference denosumab in all cases.

Table 3.

Summary of CTx-1 pharmacodynamic parameters.

	Treatment group
	FKS518				Reference Denosumab
	Geo Mean	95% CI	n		Geo Mean	95% CI	n
CTx-1
%CfBmax (%)	97.73	96.85– 98.62	105		98.74	98.03– 99.46	104
AUEC0-W40 for %CfB (h∙%)	505,490	492,831– 518,475	103		491,154	475,798– 507,005	102
Net AUEC0-W40 for %CfB (h∙%)	503,561	490,314– 517,166	103		480,844	452.637– 510,808	103
AUEC0-W26 (pg∙h/mL)	2,199,655	2,028,587– 2,385,148	103		1,964,941	1,804,157– 2,140,053	103

Abbreviations: AUEC = area under the effect curve; CfB = change from baseline; CI = confidence interval; Geo Mean = geometric mean. Values are reported for the PD analysis set.

3.9. Impact of alternative ESC acceptance criteria on assay pass rate: a retrospective analysis

The benefit of adopting ESCs in addition to spiked QCs in different contexts, such as method validation, stability testing, assessment kit lot-to-lot consistency, and monitoring of performance, has been discussed in various publications (see discussion in Section 4.3). Nevertheless, one aspect that is often not precisely clarified and/or differently addressed among research groups is whether a run acceptance criterion should be set based on ESC ranges during sample analysis, and if so, what the appropriate range would be for a method applied in regulated analysis, such as the CTx-1 method discussed here.

ESC acceptance ranges displayed in Table 1 were obtained based on limits calculated from the mean ±2´SD or mean ±20%, whichever yielded the largest range. Since it has been noticed that some publications use different strategies, a retrospective analysis was performed to assess the impact of alternative formulas for ESC limits on the outcome of the CTx-1 quantitation.

The retrospective analysis was performed in three sequential steps: (1) assessment of the impact alternative ESC limits on the assay pass rate; (2) calculation of the relative percent difference (RPD) between the final sample analysis results and the results obtained during runs rejected due to ESC ranges; and (3) biomarker response profiles for a hypothetical scenario where sample analysis results runs were accepted in the absence of a criterion based on ESC ranges.

3.10. Impact alternative ESC limits on the assay pass rate

Alternative limits were calculated for comparative purposes for ESCs 1 to 4 using different formulas based on the mean concentration ±2×SD, ±3×SD, ±20% DFT, or ±30% DFT (Table 4). Formulas based on the %CV yielded relatively tighter limits for the two ESC levels closer to the LLOQ (ESC1 and ESC3), whereas the SD-based formulas yielded relatively tighter limits for the two higher ESC levels (ESC2 and ESC4).

Table 4.

Alternative limits for ESCs used in validation and sample analysis.

ESC Level	Theoretical concentration (pg/mL)	Lower to upper limit (pg/mL)
		Formula applied
		±2 × SD	±3 × SD	±20% DFT	±30% DFT
ESC1	77.2	50.5– 104	37.2– 117	61.7– 92.6	54.0– 100
ESC2	201	163– 239	144– 258	161– 242	141– 262
ESC3	141	107–175	88.9– 192	112– 169	98.4– 183
ESC4	359	303– 416	275– 444	287– 431	252– 467

Different formulas were applied to calculate alternative limits for ESC acceptance. The limits in bold were adopted as acceptance ranges following the rules specified in Section 2.2.

Next, the impact of alternative ESC limits on the assay pass rate was assessed. For this retrospective evaluation, all reasons for run rejection that are unrelated to ESCs (n = 13: e.g., unacceptable calibration standards, matrix blank, or QCs) were respected as per the original sample analysis. Concerning ESCs, there were three runs that failed due to insufficient precision, which automatically triggered a rejection regardless of the range applied. However, there were 14 runs that failed exclusively due to either one or both ESC levels testing outside the acceptable range. In those cases, the outcome was reassessed based on the alternative ranges described in Table 4 and summarized in Table 5.

Table 5.

Impact of different rules for defining ESC acceptance ranges on pass rate during sample analysis.

Rule applied to define ESC range	# Runs failed due to ESCa	Total # failed runsb	Passing runs	Pass rate (%)
Mean ±2 × SD for both ESCs	25	38	205	84.4
Mean ±20% for both ESCs	25	38	205	84.4
Mean ±3 × SD for both ESCs	7	20	223	91.8
Mean ±30% for both ESCs	9	22	221	90.9
Mean ±2 × SD or mean ± 20%, whichever is greater for each ESC level	17	30	213	87.7

The ESC ranges defined using alternative formulas shown in Table 4 were used to retrospectively assess their impact on run pass rate from a total of 243 runs performed during sample analysis of study FKS518-001.

^aESC-related reasons for run failure were either accuracy (result outside range) or precision (>20% CV).

^bIn addition to the ESC-related reasons, runs also failed due to unacceptable calibration standards, analyst error, unacceptable matrix blank, and/or unacceptable quality control samples.

The results show that applying either of the two more stringent ESC acceptance range criteria alone (“Mean ±2×SD” or “Mean ±20% for both ESCs”) would equally result in a pass rate of 84.4%. The use of more flexible acceptance ranges (Mean ±3×SD or the Mean ±30%) would lead to slightly increased pass rates of 91.8% and 90.9%, respectively. A fifth scenario, which corresponded to the one applied during method validation and sample analysis, consisted of combining the rules based on mean ±2×SD and mean ±20%, to whichever gives the larger range. The observed pass rate was 87.7%, which is an intermediary value among those obtained with stricter and more flexible rules. This scenario had been chosen as a commitment between avoiding failing too many runs and still trying to maintain sample results as reliable as possible. In summary, the implementation of these alternative rules for defining ESC ranges would lead to reasonable pass rates in all cases, with only small differences among different methods.

3.11. Assay reproducibility based on sample results from rejected runs

The relative percent difference (RPD) between the failed run result versus the reported result was calculated for all samples tested in the 17 runs that had to be rejected due to ESC criteria. The frequency results are presented in Figure 4 and show that the concentration of 92.3% of the samples that had to be reassayed due to ESC range fell within 30% of the final reported result. More specifically, in 55.2%, 26.2%, and 10.9% of these cases, the failed results fell within 10%; between 10% and ≤20%, and between 20% and ≤30% of the reported result, respectively.

Figure 4.

Histogram for the relative percent difference for sample results reassayed due to ESC range.

Histogram showing the relative percent difference (RPD) between final sample results and those from runs rejected for exceeding ESC acceptance limits. Among the 253 reassayed samples, 59 tested <50 pg/mL in both the failed and the passing run. Those were accounted for in the first bin (|RPD| <10). There were 47 other sample results that tested undetectable either in the failed run or the passing run, and the RPD could not be calculated. Those are not included in the histogram and had a median concentration of 58.8 pg/mL (95% CI of the mean 57.3–63.5), and an interquartile range of 55.1–68.55 pg/mL of CTx-1. The histogram displays the relative frequency of |RPD| for the 221 reassayed samples with calculable values.

3.12. Sample profiles in the absence of an ESC range-based acceptance criterion

Since reassayed sample results closely matched final reported values, the next step was to assess whether applying ESC limits for run acceptance influenced the CTx-1 evaluation outcome. Figure 5 compares the mean %CfB biomarker responses obtained with and without ESC acceptance criteria for 20 subjects who had at least one sample reassayed due to ESC-related run failure.

Figure 5.

Serum CTx-1 %CfB as reported vs disregarding ESC limits for run acceptance.

Comparison of the reported serum C-terminal cross-linking telopeptide of Type 1 collagen (CTx-1) result concentration expressed as the percent change from baseline (%CfB) versus hypothetical results disregarding endogenous serum control (ESC) limits for run acceptance from the comparative pharmacokinetic (PK) study. The plotted data represents the overall (unassigned to treatment) %CfB mean ±SD from subjects who had CTx samples reanalyzed due to ESC ranges (n = 20). The reported study results (i.e., obtained from runs meeting all acceptance criteria) are plotted in blue. In brown, the (first) sample result that was obtained during an assay run that failed exclusively due to an ESC-related reason was used to replace the reported result. Postdose visits on Day 1 (D1), D2, and D3 are omitted from the x-axis due to space limitations. The data were nudged along the X axis for clarity.

Using ESCs acceptance limits provided a slight reduction in data dispersion at a few timepoints in this subset of patients, but the effect diminished when averaging data from all patients, making its benefit negligible (not shown). Similar findings were observed for untransformed concentration data (pg/mL) (not shown). Briefly, no substantial differences in the mean CTx-1 %CfB were found with or without application of ESC limits.

In addition to their application in various assessments performed during method validation, testing of ESCs is unequivocally the most appropriate approach to demonstrate biomarker stability. Moreover, inclusion of an ESC pair in each plate during samples for trend analysis can contribute to a stronger confidence in study results and allow timely implementation of troubleshooting actions, when necessary. However, the retrospective analysis indicates that applying a strict run acceptance criterion based on ESC levels was marginal and would not change the outcome of the PD biomarker evaluation in the FKS518 development program. Further studies and retrospective analyses are encouraged, including input from various sponsors, to assess if these observations can be extrapolated to other assays.

4. Discussion

4.1. Outcome of assay validation & performance

The sensitivity was targeted and demonstrated at 50 pg/mL, similar to a method that had been used in previous denosumab studies and that applied an LLOQ of 49.0 pg/mL [18]. Achieving the desired sensitivity was one of the critical efforts of method development.

Assay validation included assessing biological variability in CTx-1 using commercially sourced serum samples. Mean CTx-1 levels for healthy individuals (128.5 pg/mL) and PMO women (248.0 pg/mL) were notably lower than literature values, with many PMO samples showing undetectable levels. These findings highlight challenges in evaluating biological variability with commercial samples.

Parallelism experiments were conducted using samples available during validation, all of which had concentrations within the assay range, so only up to threefold dilutions were tested to stay above the LLOQ. Parallelism was confirmed by precision ≤30% CV [37]. The recovery %DfB also remained consistently below 30%. These results verified the assay’s sensitivity, MRD, and matrix selectivity. The assay’s parallelism was successfully demonstrated. Of note, in the FKS518-001 study, only 0.54% of samples (18/3,333) exceeded the ULOQ on the first assay and required a reassay including a pre-dilution step.

Initial long-term stability experiments were conducted with an ESC pair (ESC1 and ESC2) whose acceptance ranges were characterized during assay validation. It was challenging to generate ESCs at the appropriate levels and at the required volumes using serum from commercial sources due to their low CTx levels. As a consequence, the range for ESC1 was too low and close to the LLOQ. Overall, the ranges for the first ESC pair (ESC1 and ESC2) were considered inappropriate to monitor assay performance during sample analysis. Additional commercial samples were acquired, and a new ESC pair (ESC3 and ESC4) was prepared, with theoretical concentrations more closely representing the range observed in clinical samples at baseline. Using this ESC pair, the stability of CTx-1 in serum samples was demonstrated up to 70 days at −25°C, and 698 days at −80°C, compatible with what is reported in literature [18].

Preparation costs for ESC pools are a key consideration in biomarker method validation and sample analysis. If multiple clinical studies are planned, obtaining optional consent for future use of patient samples can support ESC pool preparation [38]. Planning ahead allows leftover sample volumes and results from the initial study to guide ESC pool selection and ensure adequate amounts for monitoring assay performance in subsequent studies.

The results of the present method validation also demonstrated stability of CTx-1 in serum up to six freeze/thaw cycles and benchtop stability up to 24 hours. The data agree with previous studies that demonstrated stability of CTx-1 in serum samples for periods ranging from 6 hours up to 2 days at room temperature and between 22 hours up to several days at 2–8°C [18,23–25,32].

The assay development, validation, and sample analysis were designed to reach the appropriate sensitivity, to keep variability under control and to ensure reliable assay performance (e.g., through the use of ESCs). Overall, a comprehensive range of assessments were performed during method validation, justified by the use of a pharmacodynamic parameter derived from CTx-1 concentrations as a co-primary endpoint in the comparative efficacy and safety study. The method showed satisfactory results in all assessments performed during validation and is considered fit for purpose.

4.2. Clinical evaluation

The assay sensitivity was sufficient not only to detect an unequivocal decline in serum CTx-1 concentrations, but also a gradual return toward baseline values following the maximal response lasting up to around Day 113 post treatment, a point that was critical for detecting potential minor differences in exposure and/or potency. More specifically, a (partial) recovery in CTx-1 plasma levels could be observed toward the end of the study in nearly all subjects who completed the study, and only three subjects had undetectable CTx-1 levels at the last visit. These observations confirmed the adequacy of the LLOQ set at 50 pg/mL and corroborate the conclusion that the assay validation and performance during sample analysis was successful and met their requirements, in alignment with its CoU.

Baseline concentrations in both groups were in alignment with reports from literature. The baseline CTx levels observed for the healthy male population were 545 and 597 pg/mL on average for the FKS518-001 and reference denosumab groups, respectively. These results were slightly higher than those observed in the healthy male population, namely 400 (100–162) pg/mL for the geometric mean (95% confidence interval) as reported by Hu et al. 2013 [14] and 260 (182–366) pg/mL for the mean (interquartile range) as reported by Olmos et al. 2018 [39]. Both references required a fasting status for CTx sampling; therefore, the impact of the fasting status on the biological variability cannot account for the difference between FKS518-001 and the cited studies. One possible explanation could be the use of different assays, as has been pointed out by others that using different assays for CTx-1 may 1 May lead to poor agreement between results [40,41]. Another possible explanation could be, at least in part, the higher age range (50–92 years) of healthy volunteers from the cited study of Olmos et al. 2010 [39], compared to the subject population recruited for study FKS518-001 (28–55 years). A reference standard for CTx is currently lacking, and harmonization of commercial assays via the implementation of a reference standard preparation could facilitate comparison of results across studies.

Importantly, there was no evidence of a drift in ESC performance during the approximately 6 months spanning the sample analysis period. This observation grants stronger reliability on the obtained sample analysis results.

Concerning the response to treatment, the parameters derived from the quantitation of CTx-1 concurrently pointed to the similarity between FKS518 and the reference denosumab drug product based on all calculated PD parameters.

4.3. Use of ESCs to monitor assay performance

A review of the recent literature allows the conclusion that the use of ESCs to assess analyte stability and monitor assay performance – such as detecting analytical drift – is well established in the field of regulated bioanalysis. For instance, Jani et al. [42] recommend using endogenous quality controls (ESCs) to monitor assay performance as opposed to spiked QCs, with acceptance criteria based on observed precision (e.g., “4–6–30 rule”: at least four of six QCs within ±30% of nominal). Azadeh et al. [43] provide recommendations and best practices for preparation, qualification, and maintenance of lot-to-lot consistency for QCs applied in LBAs in general. Specifically regarding biomarker assays, the authors propose different ways to establish limits for (endogenous) QC performance for trending analysis, but these approaches are not discussed in the context of accepting or rejecting a run. Jentsch et al. [44] highlight the importance of assessing biomarker stability and assay performance in early method development and optimization. An acceptance criterion of ±25% bias is applied for stability testing using ESCs. The authors state that QC can be used in every run as an acceptance control while also providing a way to track plate shifts and assay drifts over time. However, no further recommendations were given as to how to set up acceptance ranges during sample analysis. Kar et al. [45] present three case studies to illustrate different approaches for the use of ESCs for trending analysis and stability determination. For this case study, which is most relevant to the present discussion, measurement of TGF-β1 in serum, the referred paper applies four modified Westgard rules for trending analysis. In addition, batch acceptance was set at ±30% of the nominal concentration for system suitability and stability assessment. The authors point to the exploratory nature of their work, which could imply that their recommendations may not necessarily apply to regulated analysis. Another relevant example was the study of Wang et al. [18] that demonstrated one-year CTx assay stability with authentic samples and used ESC data for trending analysis over 35 months. ESC performance results were plotted over time along with the representation of the mean ±2´SD. However, this range was not used to accept or reject a run. Rather, the acceptance criteria were based on the calibration curve (20% of the nominal value for at least 75% of the acceptable standard points) and on QCs (two-thirds within 20% nominal values).

Despite the recognition of the usefulness of ESCs in biomarker assays, some operational questions related to the implementation of ESC for monitoring performance may be handled differently across laboratories and sponsors. One such question is whether ESC-based performance monitoring should include predefined acceptance limits for run validity, or whether ESC trends should simply serve as qualitative alerts to prompt laboratory investigations and corrective actions when major deviations are observed. Even if a decision is made to implement quantitative acceptance criteria, a one size fits all statistical approach is not available. This is because unlike the predefined acceptance criteria established for small and large molecule PK assays, biomarker assay acceptance criteria are determined by the unique physiological characteristics of each individual biomarker [9]. Options range from fixed percentage thresholds (e.g., mean ±20%, ±25%, ±30%) to statistical control limits (e.g., ±2×SD, ±3×SD).

In the method described in the current paper, ESC ranges were applied for run acceptance and were calculated as a combination of two limits (i.e., mean ±2×SD or mean ±20%), whichever is greater for each ESC level. This led to a pass rate of 87.7% for in the CTx-1 method in the PK equivalence study. A retrospective analysis was performed to assess the impact of application of either more flexible or more stringent criteria for ESC acceptance ranges. Only minor differences in the pass rate would ensue, regardless of the criteria applied (range 84.4% to 91.8%). In addition, the vast majority of the sample results that failed due to ESC range fell within 30% of the final reported result.

If ESC run acceptance limits had not been applied, the study results would remain largely unchanged, except for small changes in data dispersion that would not impact overall conclusions. This supports the conclusion that, provided other acceptance criteria are met (e.g., those related to the blank, to the standard curve, spiked QCs A&P; and ESC and sample precision), the use of more flexible ESC limits for run acceptance could have been adequate for this particular method. This conclusion does not undermine the relevance of ESC trend analysis for enhancing data confidence, nor the acknowledged value of ESCs in assessing biomarker stability in regulated analyses.

5. Conclusion

In summary, a method for the quantitation of CTx-1 was developed and validated in agreement with industry standards defining a list of applicable assessments and passing criteria based on the assay’s CoU. The method met all of the desired attributes considered necessary to be applied as a supportive element in the assessment of biosimilarity between a biosimilar candidate and a reference drug, including desired sensitivity, adequate relative accuracy, precision and selectivity, sufficient freeze/thaw and benchtop and long-term stability, a lack of drug and exogenous interference, as well as no evident hemolysis or lipemia effects. All biomarker data generated with the validated method concurred with the biosimilarity between FKS518 and the reference denosumab.

The industry applies a fit-for-purpose approach in the validation of biomarker assays and sample analysis, with ESCs applied for assessing analyte stability and monitoring assay performance. While the use of ESCs for monitoring assay performance is becoming well established, some operational questions remain. Various approaches have been presented in literature on how to use ESC performance for trending analysis, with the possibility to set thresholds for runs acceptance based on the observed ESC performance to be used in addition to run acceptance criteria established for spiked QCs, calibrators, and blank. The present paper demonstrates an example of a biomarker method where, provided the samples are analyzed within the demonstrated stability period, and provided other acceptance criteria are respected, the use of narrow limits for run acceptance based on ESC performance brings negligible benefits versus not applying ESC performance for run acceptance. The disclosure of case studies like the present one can help determine whether these observations are method-specific or could be extrapolated to other biomarker assays in general.

Correction Statement

This article was originally published with errors, which have now been corrected in the online version. Please see Correction (http://dx.doi.org/10.1080/17576180.2026.2619351)

Funding Statement

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

Article highlights

• A CTx-1 quantitation method was developed by modifying a commercial kit. Assay development mainly aimed to increase sensitivity.

• The method was validated supported by a thorough set of assessments.

• During validation, acceptance ranges were defined for an ESC pair according to the formulas: Mean concentration (pg/mL) ±2×SD or ±20%, whichever was greater. These ranges were applied for assessing long-term stability.

• During samples analysis, relative ESC accuracy (based on the calculated limits) and precision (20% CV) were applied for run acceptance. The assay had a pass rate of 87.7%, with assay run failures mostly driven by unacceptable ESCs.

• All subjects in the FKS518-001 study had detectable CTx-1 concentrations at baseline, contrary to what was observed during biological variability testing with commercial samples during validation.

• The pharmacodynamic response was appropriately characterized, including the observation of a recovery phase starting around Day 127 postdose. The results contributed to the demonstration of similarity between FKS518 and reference denosumab.

• In a retrospective analysis, 92.3% of sample results that were rejected due to ESC range fell within 30% of the final reported value. This degree of reproducibility meets or exceeds the ISR standards commonly applied to PK assays.

• The response profiles for the reported results did not substantially differ from those obtained without applying ESC limits for run acceptance.

• The results exemplify a biomarker method for which, as long as other relevant criteria are met, imposing strict ESC-based run acceptance requirements is unnecessary.

Author contributions

Adriano L.S de Souza: project administration, data analysis and interpretation, writing – original draft.

Anna Lucia Buccarello: methodology, data analysis and interpretation, writing – review and editing.

Amandine Berthet: data analysis and interpretation, writing – review and editing.

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

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

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/17576180.2025.2607081