Accepted: 12 November 2018

Transcription

1 Received: 2 August 2018 Revised: 12 November 2018 Accepted: 12 November 2018 DOI: /hae ORIGINAL ARTICLE Laboratory science Accurate measurement of extended half life and unmodified factor VIII low levels with one stage FVIII assays is dependent on the matrix of calibration curves Dorien Van den Bossche 1 Jelle Toelen 1 Joke Schoeters 1 Isa Van Horenbeeck 1 Ingrid Vanlinthout 1 Mirjam Debasse 1 Kathelijne Peerlinck 2,3 Marc Jacquemin 2,1 1 Clinical Department of Laboratory Medicine, University Hospitals of Leuven, Leuven, Belgium 2 Department of Cardiovascular Sciences, Center for Molecular and Vascular Biology, University of Leuven, Leuven, Belgium 3 Vascular Medicine and Hemostasis, University Hospitals of Leuven, Leuven, Belgium Correspondence Marc Jacquemin, Department of Cardiovascular Sciences, Center for Molecular and Vascular Biology, KU Leuven, Leuven, Belgium. marc.jacquemin@kuleuven.be Funding information KU Leuven, Grant/Award Number: PF/10/014 Introduction: The monitoring of factor VIII (FVIII) replacement therapy relies on the accurate measurement of FVIII activity over a large concentration range. However, unexplained overestimation of low FVIII levels has recently been reported with extended half life recombinant FVIIIs. Aim: The objective of this study was to confirm previous publications indicating that the reagents used to generate the calibration curves determine the accuracy of the measurement of low FVIII levels. Methods: We generated FVIII calibration curves with FVIII deficient plasmas or a commercial diluent buffer. We then measured FVIII levels in FVIII deficient plasma spiked with plasma FVIII, a full length recombinant FVIII (Advate, Shire, Brussels, Belgium) and two extended half life recombinant FVIIIs (Elocta, Swedish Orphan Biovitrum, Woluwe Saint Lambert, Belgium and Afstyla, CSL Behring, Mechelen, Belgium). FVIII levels were also analyzed in spiked samples prediluted two and fourfold, either in diluent buffer or in FVIII deficient plasma to evaluate parallelism. Results: Coagulation times of calibration curves generated with diluent buffer were longer than those with FVIII deficient plasmas. This resulted in an overestimation of FVIII levels lower than 25 IU/dL in spiked samples and in detection of FVIII activity ( 1 IU/dL) in FVIII deficient plasma. Predilution of samples with diluent buffer rather than with FVIII deficient plasma also led to discordant results. Conclusion: Our data confirm that the generation of calibration curves by dilution in FVIII deficient plasma is crucial for the accurate measurement of low FVIII levels. When serial dilutions of samples are analyzed, predilution in FVIII deficient plasma is required to respect the parallelism criteria. These methods should be more generally implemented for coagulation instruments. KEYWORDS extended half life FVIII, factor VIII, FVIII one stage FVIII assay, haemophilia A, parallelism Haemophilia. 2019;25:e19 e26. wileyonlinelibrary.com/journal/hae 2018 John Wiley & Sons Ltd e19

2 e20 1 INTRODUCTION The monitoring of substitution therapy in patients with haemophilia A is dependent on the accurate measurement of factor VIII (FVIII) activity over a large concentration range. The evaluation of FVIII levels in the normal range is critical during surgical intervention, whereas the evaluation of FVIII levels in the low range is mainly important during prophylaxis. The availability of extended half life recombinant FVIII (EHLrFVIII) concentrates has made FVIII monitoring more complex. In certain cases, modifications in the rfviii molecules can induce unusual interactions with the assay reagents. For example, single chain rfviii (ScrFVIII, Afstyla, CSL Behring, Mechelen, Belgium) or PEGylated rfviii Bay (Bayer, Diegem, Belgium) are underestimated twofold and 10 fold, respectively, when measured using one stage assays activated with silica. 1,2 By contrast, PEGylated rfviii BAX 855 (Adynovate, Shire, Brussels, Belgium) and Fc fusion rfviii (rfviiifc, Elocta, Swedish Orphan Biovitrum, Woluwe Saint Lambert, Belgium) can be reliably measured with these reagents. 3-5 Therefore, assays and their reagents must be carefully selected in relation to the EHL rfviii being measured. For this reason, many studies were carried out with spiked samples with known FVIII levels. Despite the selection of suitable assays for the reliable measurement of EHL rfviii, international studies have reported an overestimation of the concentration of these molecules when the FVIII level was lower than 5 IU/dL. 1,4 In spite of using the recommended coagulation assays and correction factor for the measurement of Afstyla (ScrFVIII), Bowyer et al reported a two to fourfold overestimation of FVIII activity in FVIII deficient plasma spiked with 2 IU/dL of this product. By contrast, FVIII activity measured in samples containing about 100 IU/dL FVIII was within 77% 116% of the expected value with all coagulation assays. 1 Similarly, in a multicentre study, Bulla et al recorded a mean recovery of 127.7% of the expected FVIII activity in a sample spiked with 80 IU/dL PEGylated rfviii (Adynovate ). In a sample spiked with 3 IU/dL, however, the recovery of the same product reached a mean of 213.5%. 4 Interestingly, during this study calibration of the assay with a product specific standard improved the recovery in the normal range, but not in the low range. 4 A similar overestimation of low FVIII activity was observed in another multicentre study. 3 It is unlikely that this overestimation of low FVIII levels is solely caused by the poor reproducibility of the assays, because in this case a clear trend towards overestimation would not have been observed. Previous studies have indicated the use of diluent buffer instead of FVIII deficient plasma for the dilution of the reference material as a potential cause of overestimation in the low FVIII concentration range. 6,7 However, despite these publications, the preparation of calibration curves by diluting the reference material in FVIII deficient plasma is not always the standard in local laboratories. Moreover, this procedure is not systematically recommended by the companies that provide the automated coagulation instruments and coagulation reagents. We therefore studied the impact of the use of different reagents in the preparation of calibration curves on the measurement of different types of FVIII. 2 MATERIALS AND METHODS 2.1 One stage clotting assay and reagents The one stage FVIII clotting assay was performed on an ACL TOP500 coagulometer (Instrumentation Laboratory, Bedford, MA, USA) using HemosIL SynthASil aptt reagents (Instrumentation Laboratory) as activator and phospholipid source. 8 Citrated plasma samples were diluted in diluent buffer (Instrumentation Laboratory) prior to analysis. The diluted plasma (25 μl) was mixed with an equal volume of FVIII deficient plasma, followed by the addition of HemosIL SynthASil (50 μl). The mixture was incubated for 200 ± 20 seconds. In the last step, 50 μl of mol/l CaCl 2 was added. Commercial FVIII deficient plasmas were purchased from Werfen, Siemens (Healthcare, Beersel, Belgium) and Grifols (Barcelona, Spain; DG FVIII). The levels of all intrinsic clotting factors other than FVIII were within the normal range for the 3 FVIII deficient plasmas. VWF:Ag levels were 65.8, 1.7 and 57.7 IU/dL in Siemens, Werfen and Grifols FVIII deficient plasmas, respectively. 2.2 Calibration curves The assay was calibrated with an in house pooled normal plasma hereafter referred to as reference standard which was in turn calibrated against the HemosIL Calibration Plasma standard (Instrumentation Laboratory). The former standard is calibrated and traceable against the 5th World Health Organization (WHO) International Standard FVIII concentrate (02/150). The target value assigned to the in house reference standard was 89.6% FVIII. The reference standard was diluted in diluent buffer or in FVIIIdeficient plasma to generate an eight point calibration curve with the zero point being diluent buffer or FVIII deficient plasma alone. For each type of calibration curve, a low and a high range curve were prepared with samples containing 0 50 and IU/dL. FVIII levels lower than 5 IU/dL were determined using the low range calibration curve. All calibration points were determined in triplicates, with CVs 2.5%. The low and high range calibration curves were fitted with 3rd and 2nd order polynomials, respectively. The r 2 of the fitted calibration curves were Measurement of FVIII levels with a single dilution of the samples In one set of experiments, Grifols FVIII deficient plasma was spiked with IU/dL rfviii, Advate (Shire, Brussels, Belgium) or Elocta, based on the labelled potencies, or with IU/ dl plasma pool FVIII (Visucon, Affinity Biological Inc, Ancaster, Canada). FVIII levels in these samples were measured using Grifols FVIII deficient plasma as substrate in the assay. Siemens FVIII deficient plasma was spiked with IU/dL Advate or Elocta. FVIII levels in these samples were measured using either Grifols or Siemens FVIII deficient plasma as substrate in the assay. Siemens FVIII deficient plasma was also spiked with IU/dL Afstyla.

3 e21 valid if the sample dilution curves are parallel to the calibration curve. Parallelism was assessed by calculating the intra assay coefficients of variation (CVs) obtained with different dilutions of the samples. The samples and the calibration curves were considered parallel over the range of the assay when the intra assay CV was <20%. 10 Three automated dilution methods were compared. In the first method, the samples were prediluted two and fourfold with FVIIIdeficient plasma (Siemens) as recommended by Cinotti et al. 6 In the second method, the samples were prediluted with diluent buffer. These additional sample dilutions were already programmed by the manufacturer in the ACL TOP500. The samples were diluted twoand fourfold by mixing 50 and 25 μl samples with 50 and 75 μl of diluent material, respectively. The prediluted samples were then further diluted 10 fold in diluent buffer prior to analysis. Because predilution in FVIII deficient plasma requires large volumes of this reagent, we developed a protocol using 10% FVIII deficient plasma as a diluent, while maintaining the concentration of FVIII deficient plasma in the diluted sample identical to that in the mixtures used for the calibration of the assay. About 13 and 6 μl of samples diluted 10 fold in diluent buffer was mixed with 13 and 19 μl of diluent buffer supplemented with 10% FVIII deficient plasma (Siemens), respectively. This procedure required only 3.2 μl FVIII deficient plasma per sample instead of the 125 μl of this reagent required in the procedure programmed by the manufacturer. Siemens FVIII deficient plasma was used as substrate. FVIII deficient plasma (Siemens) was spiked with 100, 30, 10 and 3 IU/dL plasma pool FVIII (Visucon ). Samples with a concentration lower than 3 IU/dL FVIII were not analysed because the fourfold dilution would have brought the FVIII levels below the LOD of the assay (0.5 IU/dL). Values below 0.5 IU/ dl measured by the automated coagulometer were not taken into account to calculate the recovery. 2.5 Measurement of FVIII levels in commercial FVIII deficient plasma FIGURE 1 Generation of FVIII calibration curves with diluent buffer or FVIII deficient plasma. Eight point calibration curves were generated by diluting an in house pooled normal plasma standard with either diluent buffer or FVIII deficient plasma. A, Siemens; B, Werfen; C, Grifols FVIII deficient plasma. Data are shown as the mean of triplicates. The mean CV of the triplicates was 2% FVIII levels were then measured using Siemens FVIII deficient plasma. Before measurement, a single dilution (10 fold) of the samples was prepared by the ACL TOP. Each analysis was carried out in triplicate. 2.4 Parallelism testing To reduce the risk of reporting falsely low FVIII levels due to the presence of lupus anticoagulant, the measurement of multiple dilutions of each plasma sample is recommended. 9 The result is only The FVIII levels in the different commercial FVIII deficient plasmas were verified with a FVIII one stage assay using plasma from a patient with severe haemophilia A (inversion of intron 22) as substrate and calibrated with the reference standard. 3 RESULTS 3.1 Generation of calibration curves with diluent buffer or FVIII deficient plasma Calibration curves were generated by diluting a pool of plasma standard with either diluent buffer or the FVIII deficient plasma used as substrate plasma for the coagulation assay. Coagulation times recorded after dilution of the reference standard with diluent buffer were significantly longer than those recorded after dilution with FVIII deficient plasma for calibrator mixtures containing <25 IU/dL FVIII (paired t test, P 0.05; Figure 1). Importantly, the

4 e22 coagulation time recorded with diluent buffer was seconds longer than that measured with FVIII deficient plasmas (Figure 1). Additional experiments demonstrated that the shorter coagulation times recorded with FVIII deficient plasma in comparison to those measured with diluent buffer were unlikely to be caused by low residual FVIII concentrations in the commercial FVIII deficient plasmas. The FVIII levels measured in the three commercial FVIIIdeficient plasmas were <0.5 IU/dL. In addition, the coagulation times for the diluent buffer spiked with up to 0.3 U/dL plasma FVIII were not significantly shortened in comparison to the coagulation times with diluent buffer alone (data not shown). However, a possible interference of very low concentrations of FVIII in the FVIII deficient plasmas could not be totally excluded. 3.2 Overestimation of low FVIII levels with calibration curve generated in diluent buffer The differences between the calibration curves generated with diluent buffer and with FVIII deficient plasma suggested that the measurement of low FVIII concentrations could be significantly biased depending on the matrix of the calibration curves. We therefore measured FVIII levels in FVIII deficient plasma (Grifols) spiked with an FVIII plasma pool, an FL rfviii (Advate ) and an EHL rfviii (Elocta ), using calibration curves generated with either diluent buffer or FVIII deficient plasma. In these experiments, the samples were analysed after a 10 fold dilution rather than after multiple dilutions (the parallelism method) to avoid differences in the matrix of the assay. Factor VIII levels of between 70% 130% of the expected values were obtained with the calibration curve generated with FVIII deficient plasmas for all samples spiked with at least 0.8 IU/dL of any of the three types of FVIII (Tables 1-3). By contrast, the FVIII levels analysed with the calibration curve generated in diluent buffer were overestimated by 132% 600% in the samples containing <25 IU/dL FVIII. In addition, FVIII activity of about 1 IU/dL was measured in the control FVIII deficient plasma that had not been spiked with any FVIII when the calibration curve generated in diluent buffer was employed (Tables 1-3). These observations were not linked to the type of FVIII deficient plasma used in the experiments. Indeed, similar overestimations of FVIII low levels were also made using Siemens or Grifols FVIII deficient plasma as substrate when <25 IU/dL Advate or Elocta were spiked in Siemens FVIII deficient plasma (data not shown). Factor VIII levels were also measured in FVIII deficient plasma spiked with Afstyla. When using the calibration curve generated with FVIII deficient plasma and when applying the correction factor of 2 that is recommended by the manufacturer, the recovery of FVIII in samples spiked with IU/dL Afstyla ranged from 70% to 88%. By contrast, when using the calibration curve generated with diluent buffer, the recovery ranged from 105% with the sample spiked with 80 IU/dL to 353% with the sample spiked with 0.6 IU/dL Afstyla (Figure 2). 3.3 Parallelism Factor VIII deficient plasma samples, spiked with 100, 30, 10 and 3 IU/dL plasma FVIII, respectively, were analysed using three dilutions. FVIII levels lower than 0.5 IU/dL in the diluted samples were not taken into account for the analysis. As shown in Figure 3, a lack of parallelism (intra assay CVs 20%) was observed for samples containing 30 IU/dL FVIII when the dilution was carried out with diluent buffer. By contrast, by diluting the sample in FVIII deficient plasma or by using the method with diluent supplemented with 10% FVIII deficient plasma, the intra assay CVs of the spiked samples TABLE 1 Measurement of FVIII levels in FVIII deficient plasma spiked with plasma FVIII Calibrator in FVIII deficient plasma Calibrator in diluent buffer Expected recovery (IU/dL) Recovery (% of expected recovery) Recovery (% of expected recovery) ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± <0.5 NC 1.8 ± <0.5 NC 1.1 ± <0.5 NC 1 ± 0.1 >> Calibration curves were generated by diluting a pool of plasma standard with FVIII deficient plasma (Grifols), which was also used as substrate plasma for the coagulation assays, or with diluent buffer. Samples were prepared by diluting a plasma pool with FVIII deficient plasma. Results are expressed as the mean ± SD of FVIII:C in triplicates and as a % of the expected recovery. FVIII levels higher than 130% or lower than 70% of the expected recoveries are highlighted in bold. NC, not calculated. >>, the recovery is higher than expected but cannot be calculated because the divisor is 0.

5 e23 TABLE 2 Measurement of FVIII levels in FVIII deficient plasma spiked with Advate Calibrator in FVIII deficient plasma Calibrator in diluent buffer Expected recovery (IU/dL) Recovery (% of expected recovery) Recovery (% of expected recovery) ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± <0.5 NC 1.6 ± <0.5 NC 1.4 ± <0.5 NC 0.9 ± 0.3 >> Calibration curves were generated by diluting a pool of plasma standard with FVIII deficient plasma (Grifols), which was also used as substrate plasma for the coagulation assays, or with diluent buffer. Samples were prepared by spiking FVIII deficient plasma with Advate. Results are expressed as the mean ± SD of FVIII:C of three experiments carried out with triplicates and as a % of the expected recovery. FVIII levels higher than 130% or lower than 70% of the expected recoveries are highlighted in bold. NC, not calculated. >>, the recovery is higher than expected but cannot be calculated because the divisor is 0. TABLE 3 Measurement of FVIII levels in FVIII deficient plasma spiked with Elocta Calibrator in FVIII deficient plasma Calibrator in diluent buffer Expected recovery (IU/dL) Recovery (% of expected recovery) Recovery (% of expected recovery) ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± <0.5 NC 2.0 ± <0.5 NC 1.0 ± <0.5 NC 1.1 ± 0.3 >> Calibration curves were generated by diluting a pool of plasma standard with FVIII deficient plasma (Grifols), which was used as substrate plasma for the coagulation assays, or with diluent buffer. Samples were prepared by spiking FVIII deficient plasma with Elocta. Results are expressed as the mean ± SD of FVIII:C in triplicates and as a % of the expected recovery. FVIII levels higher than 130% or lower than 70% of the expected recoveries are highlighted in bold. NC, not calculated. >>, the recovery is higher than expected but cannot be calculated because the divisor is 0. were 11.1%. Accordingly, both methods provided an acceptable parallelism between the samples and the calibrator dilution curves. 4 DISCUSSION An important overestimation of low plasma FVIII, rfviii and EHLrFVIII levels has been reported in several studies. 1,3,4 Our data indicate that the generation of calibration curves by diluting the reference standard in diluent buffer rather than in FVIII deficient plasma can lead to a significant overestimation of low FVIII activities. This overestimation occurred with the EHL rfviiis Elocta and Afstyla, with unmodified rfviii Advate and with plasma FVIII, spiked at known concentrations in FVIII deficient plasma. In addition, a significant FVIII activity was also recorded in FVIII deficient plasma when using the calibration curve generated in diluent buffer.

6 e24 FIGURE 2 Measurement of FVIII levels in FVIII deficient plasma spiked with Afstyla. Calibration curves were generated by diluting a pool of plasma standard with FVIII deficient plasma (Siemens), which was also used as substrate plasma for the coagulation assays, or with diluent buffer. Samples were prepared by spiking FVIIIdeficient plasma with Afstyla. The FVIII levels were multiplied by a correction factor of 2 as recommended by the company. 1 Results are expressed as the mean ± SD of triplicates of the expected FVIII recovery Accordingly, this problem is likely relevant both for the monitoring of trough levels in patients treated with FVIII concentrates and for the diagnosis of moderate or severe haemophilia A. The cause of overestimation of low FVIII levels with calibration curves generated with diluent buffer instead of FVIII deficient plasma is that the coagulation time of diluent buffer used as a zero point of the calibration curve is longer than that of FVIII deficient plasma, as previously reported when measuring FVIII in plasma FVIII and plasma derived concentrates. 6,7 The reasons for coagulation times recorded with diluent buffer being longer than with FVIII deficient plasma have been explored by Frandsen et al. 7 These observations suggested that vitamin K dependent factors, FV and fibrinogen brought by the sample (aside from FVIII) contributed to accelerating the coagulation reaction. Interestingly, the difference between the two types of calibration was not due to the amount of VWF brought by the sample. Indeed, a similar difference was observed between calibration curves generated with diluent buffer and with FVIII deficient plasma without VWF or with FVIII deficient plasma containing normal levels of VWF. Dilution of the reference standards in FVIII deficient plasma rather than in diluent buffer ameliorates the measurement when a single dilution of the plasma sample is measured. When several dilutions of the samples are assayed, as recommended in the guidelines to evaluate the parallelism, the selection of the ideal assay becomes more complex. To overcome the dilution bias, Cinotti et al 6 suggested diluting the sample with FVIII deficient plasma to maintain consistent levels of coagulation factors in the sample, whether diluted or not. This procedure has the disadvantage of requiring the use of large volumes of FVIII deficient plasma. We therefore defined an alternative dilution method that also keeps the levels of coagulation factors identical in all dilutions of the samples, while using about 50 times less FVIII deficient plasma. This method was programmed on the automated coagulometer and provided accurate data. FIGURE 3 Influence of sample dilution method on parallelism. The experiments were carried out with four samples prepared by supplementing FVIII deficient plasma with different concentrations (100, 30, 10 and 3 U/dL) of a normal pool plasma. Three sample dilution methods were tested: A, Samples were undiluted or diluted two or fourfold in diluent buffer and then further diluted 10 fold in diluent buffer; B, Samples were undiluted or diluted two or fourfold in FVIII deficient plasma and then further diluted 10 fold in diluent buffer; C, Samples were diluted 10 fold in diluent buffer and then analysed without further dilution or with two or fourfold dilution in diluent buffer supplemented with 10% FVIII deficient plasma. All dilutions were carried out by the automated coagulometer. The mixtures generated by each dilution procedure were then added to the other reagents of the FVIII assay. The concentrations of FVIII in the diluted mixture were determined with a calibration curve generated with FVIII deficient plasma. Results were calculated by multiplying the FVIII levels determined in the sample, undiluted or diluted two or fourfold, by the dilution factor and were expressed as mean ± SD of triplicates. FVIII levels 0.5 IU/dL before mathematical correction by the dilution factor were not taken into account for the analysis of the data. The CV of the FVIII levels determined with the three dilutions of each spiked sample is indicated above the corresponding data. CV >20% identifies methods that do not correspond to the parallelism criteria 10

7 e25 One potential limitation of our study is that it was carried out with spiked samples rather than with patient plasma. It was necessary to use spiked samples as we required information about the expected FVIII recovery for each sample in order to evaluate the accuracy of the assays. Another limitation of our study is that the calibration curves were generated with only one reagent, SynthASil as activator and phospholipid source. Significant differences between reagents may occur, as Cinotti et al 6 have reported better results with SynthASil and Actin FS (Siemens Healthcare, Brussels, Belgium) than with Pathromtin SL (Siemens Healthcare), Actin (Siemens Healthcare) or Actin FSL (Siemens Healthcare). SynthASil, which is based on colloidal silica, was a reagent of choice for our study because it has been validated for the measurement of several of the new EHL rfviii, including Elocta and Adynovate. Further study is needed to determine the differences between calibration curves generated in FVIII deficient plasma and in diluent buffer when other reagents are used for the FVIII assay. We also observed shorter coagulation times with calibration curves generated with diluent buffer rather than with FVIII deficient plasma when Synthafax (ellagic acid, Instrumentaton Laboratory) was used in the FVIII assay. In these experiments, the coagulation times of the calibration points recorded with FVIII deficient plasma (Siemens) corresponded to that of diluent buffer spiked with 2.5 IU/dL FVIII (J. Toelen, unpublished data, September 2018). Importantly, as illustrated by our data using Afstyla, the use of calibration curves generated in FVIII deficient plasma rather than in diluent buffer is not expected to improve the measurement of FVIII levels by reducing the bias due to the modifications of the rfviii molecules aimed at prolonging the half life. Cinotti et al 6 also reported quantitative differences between FVIII assays carried out with calibration curves generated in different dilution buffers, although overestimations in the measurement of low FVIII levels were observed in all of them. We evaluated only the diluent buffer recommended by the manufacturer of the coagulation instrument. Accordingly, the range of differences between calibration curves generated with all commercially available FVIII deficient plasmas and diluent buffers cannot be predicted. Our study was carried out only with calibration curves generated with plasma as a calibrator rather than with product specific standards. The rationale to calibrate only with plasma was that for the two rfviiis used in this study, the use of a product specific standard was not recommended by the manufacturers. 3,5 In addition, a recent study had shown that the error in the measurement of low FVIII levels was only marginally reduced by the use of a product specific calibrator rather than plasma, indicating that other methodological aspects were responsible for the bias. 4 5 CONCLUSION Our data confirm that an important cause of overestimation of low FVIII levels is the generation of FVIII calibration curves in diluent buffer rather than in FVIII deficient plasma. Such calibration curves do not compensate for the coagulation factors other than the FVIII brought by the sample in the FVIII coagulation assay. This overestimation of FVIII activity occurs with plasma derived FVIII, rfviii and modified EHL rfviii. ACKNOWLEDGEMENTS This work was supported by the Programma Financiering KU Leuven (PF/10/014). The FVIII deficient plasma DG FVIII was kindly provided by Cristina Lebrero (Grifols). DISCLOSURES No competing financial interest or other conflict of interest exists for any of the authors. AUTHORS CONTRIBUTIONS D. Van den Bossche, J. Toelen, J. Schoeters, I. Van Horenbeeck, I. Vanlinthout, M. Debasse, R Lavend'homme, and M. Jacquemin were responsible for the in vitro studies. D Van den Bossche, M. Jacquemin, and K. Peerlinck contributed to the design of the study. All authors contributed to the writing of the manuscript. ORCID Marc Jacquemin REFERENCES 1. Bowyer A, Key N, Dalton D, Kitchen S, Makris M. The coagulation laboratory monitoring of Afstyla single chain FVIII concentrate. Haemophilia. 2017;23:e469 e Gu JM, Ramsey P, Evans V, et al. Evaluation of the activated partial thromboplastin time assay for clinical monitoring of PEGylated recombinant factor VIII (BAY ) for haemophilia A. Haemophilia. 2014;20: Turecek PL, Romeder Finger S, Apostol C, et al. A world wide survey and field study in clinical haemostasis laboratories to evaluate FVIII: C activity assay variability of ADYNOVATE and OBIZUR in comparison with ADVATE. Haemophilia. 2016;22: Bulla O, Poncet A, Alberio L, et al. Impact of a product specific reference standard for the measurement of a PEGylated rfviii activity: the Swiss Multicentre Field Study. Haemophilia. 2017;23:e335 e Sommer JM, Moore N, McGuffie Valentine B, et al. Comparative field study evaluating the activity of recombinant factor VIII Fc fusion protein in plasma samples at clinical haemostasis laboratories. Haemophilia. 2014;20: Cinotti S, Paladino E, Morfini M. Accuracy of FVIII: C assay by onestage method can be improved using hemophilic plasma as diluent. J Thromb Haemost. 2006;4: Frandsen H, Stender S, Scheibel E. An effect of predilution on VIII: C determination by the one stage assay. Thromb Res. 1987;48: Lattes S, Appert Flory A, Fischer F, Jambou D, Toulon P. Measurement of factor VIII activity using one stage clotting assay:

8 e26 a calibration curve has not to be systematically included in each run. Haemophilia. 2011;17: Ruinemans Koerts J, Peterse Stienissen I, Verbruggen B. Non parallelism in the one stage coagulation factor assay is a phenomenon of lupus anticoagulants and not of individual factor inhibitors. Thromb Haemost. 2010;104: Plikaytis BD, Holder PF, Pais LB, Maslanka SE, Gheesling LL, Carlone GM. Determination of parallelism and nonparallelism in bioassay dilution curves. J Clin Microbiol. 1994;32: How to cite this article: Van den Bossche D, Toelen J, Schoeters J, et al. Accurate measurement of extended half life and unmodified factor VIII low levels with one stage FVIII assays is dependent on the matrix of calibration curves. Haemophilia. 2019;25:e19 e26. hae.13656