Bulk material was purchased from the UK Blood Service in the form of plasma, which had been prepared by centrifugation of whole blood collected into CPD adenine anticoagulant. A second centrifugation step was performed to remove all cellular material and the plasma rapidly frozen at −70°C. Individual donations were tested at source and found to be negative for HBsAg, anti-HIV-1 and HIV-2 antibodies, and anti-HCV. The material was prepared for filling by thawing immediately prior to use at 37°C and then pooled. Glycine and HEPES were added at a final concentration of 1% w/v and 40 mM, respectively. To avoid activation of FXI by cold activation or contact with glass, plastic vessels were used and the plasma was maintained at room temperature after thawing and throughout the duration of the fill. The preparation was filled into siliconized glass ampoules and freeze-dried over a 5-day cycle in accordance with recommendations for the preparations of International Standards (3, 4). The 6,000 ampoules of 15/180 available have the following physical characteristics: mean fill mass of 1.0094 g (cv = 0.3%), mean dry weight of 0.0928 g (cv = 0.2%), mean residual moisture of 0.605% (cv = 13.7%), and mean oxygen head space of 0.23% (cv = 31.4%). The activation status of this preparation was investigated by the non-activated partial thromboplastin time, which is known to be sensitive to activated coagulation factors such as FXIa. The long mean clotting time of 300 s (n = 9; SD ± 2 s) for 15/180 indicates the sample to be relatively unactivated. Long-term stability of 15/180 has been assured by results from accelerated degradation study where activity of the preparations that have been stored at elevated temperature was compared with activity of the ampoules that have been stored at ultra-low temperature (−150°C). The predicted loss of FXI: C and FXI:Ag per year when the ampoules are stored at −20°C was 0.0025 and 0.039%, respectively.

The participants were provided with the 1^st International Standard for FXI, Plasma (NIBSC code, 04/142), coded S and the candidate material (NIBSC code, 15/180), as coded duplicate samples A and B. In addition, local plasma pools were required to assess the relationship between the IU and local plasma unit for FXI:C and to value assign the antigen (FXI:Ag) value. The participants were asked to collect fresh local plasma pools in the study for use both fresh, coded P1, and subsequently frozen, coded P2 during the study. NIBSC in-house studies on both FXI functional activity and FXI antigen have shown that there is no significant difference in results when using fresh plasma pools compared to the same pool of plasma used after freezing; therefore, if participants were unable to collect fresh plasma pools, then frozen plasma could be used as an alternative. Across the pools used in the study (fresh and frozen) for either functional activity or antigen value, plasma from more than 20,000 donors was used.

Of the 29 laboratories from 11 countries (Austria, Canada, Croatia, Denmark, France, Germany, Italy, The Netherlands, Spain, UK, USA) that took part in the FXI:C assignment, there were 8 clinical laboratories, 9 therapeutics producers, 6 diagnostics manufacturers, 5 regulatory laboratories, and 1 research laboratory. Eleven laboratories (2 clinical laboratories, 1 diagnostics manufacturer, 6 plasma therapeutics producers and 2 regulatory laboratories) from 8 countries (Austria, Canada, Denmark, France, Germany, Spain, UK, USA) carried out assays for FXI:Ag.

For FXI:C, participants were asked to assay coded duplicates A and B against sample S, the 1^st IS for FXI Plasma. Since the FXI antigen is a new analyte, the unitage of the candidate standard was assigned relative to normal plasma pools. Participants were requested to perform their in-house methods for functional and antigen measurements. All assays for FXI:C were one-stage clotting assays using activated partial thromboplastin time (APTT). In total, 13 different APTT reagents used across 14 different instrument platforms and a variety of deficient plasmas were employed by the participants and details of the reagents and deficient plasma used are indicated in Tables 1 and 2. For antigen assays, five different kits or paired antibody sets were used. The kits employed and the sources of local plasma pools used as the standard in the assays are listed in Tables 3 and 4, respectively.

For both FXI:C and FXI:Ag, four independent assays for each analyte were requested. A study protocol was provided, giving guidelines on randomization of samples and sufficient replication to enable statistical analysis. Raw data were returned to the NIBSC for parallel line analysis (5) with CombiStats software (CombiStats™, Version 5.0, Council of Europe). A log₁₀ transformation of the assay response was used for the analysis of the FXI antigen assays. No transformation was necessary for the FXI functional assays. All mean potencies were calculated as unweighted geometric mean (GM). Variability between assays and laboratories was expressed using geometric coefficients of variation (GCV = {10^s − 1} × 100% where s is the SD of the log₁₀ transformed potency estimates). Grubbs’ Test (6) was applied to the log-transformed laboratory mean estimates in order to detect any significant outliers. Comparisons between methods were made by appropriate t-tests of log-transformed laboratory mean estimates. One laboratory performed single point estimates only for FXI:C, hence was excluded from the overall analysis. Where a laboratory had performed more than one set of assays using a different APTT reagent or coagulometer, these data were treated independently.

Participants were asked to collect two plasma pools (P1 and P2) for use both as fresh and frozen pools in the assays. Where participants were unable to collect fresh plasma pools, it was requested that two different batches of local frozen plasma pools were used. As there is no current International Standard for FXI antigen, the units of FXI antigen in 1 ml of plasma pool was taken to be one, and the potencies of A and B calculated relative to the laboratories’ local plasma pool. Overall, the intra-laboratory variations ranged between 0.5 and 18%, with most laboratories having a GCV of less than 10% (Table 5). The variation of these assays would be expected to be higher than that of the FXI functional assays by virtue of the fact that the calculations are relative to the local plasma pools, which themselves have inherent variability. Due to the low number of assays performed using fresh plasma pools (5), it was not possible to compare fresh and frozen plasma pools.

Since A and B were coded duplicates and paired t-test showed that there was no significant difference between the estimates for A and B (p = 0.198), the results were combined to give an overall result for AB against P. The overall GM was 0.78 IU/amp, with an acceptable inter-laboratory GCV of 10%. No statistical outliers were detected. The individual assay results from each lab and the laboratory GMs are shown in Table 9 and graphically in Figure 3.

Of the 12 laboratories that took part, 5 used an antigen kit from Affinity Biologicals. The GMs for AB against P of these laboratories were combined and resulted in a GM of 0.72 IU/amp (with 95% confidence limits of 0.70–0.75 IU/amp). This is slightly lower than the overall result for AB against P (0.78 IU/amp). The result for the other laboratories not using the Affinity Biologicals kit was 0.82 IU/amp, with 95% confidence limits of 0.75–0.90 IU/amp, which overlap the overall GM of AB vs. P, but not that of the laboratories using the Affinity Biologicals kit. A comparison of the two groupings was made using the unpaired two-tailed t-test and the difference was found to be significant (p < 0.05). The difference suggests that the Affinity Biologicals kit may yield lower results than other kits; however, this could also be a reflection of the different plasma pools used within each laboratory. Due to the small number of laboratories taking part in the study and the fact that the two groupings had 95% confidence limits that were extremely close to being overlapping, the consensus GM was taken as the antigen value for the candidate preparation. Any effect of minor method bias on the results is likely to be small, especially for measurement around the normal range. For measurements of abnormal samples, local qualification and validation of antigen method are advisable.

Based on the results of this study, the participants of the study and the Experts nominated by the Scientific and Standardization Committee (SSC) of the International Society on Thrombosis and Haemostasis (ISTH) agreed that the candidate preparation, NIBSC code: 15/180 is a suitable International Standard for FXI:C and FXI:Ag. In October 2016, the ECBS of the WHO established NIBSC code: 15/180 as the 2^nd International Standard for FXI, plasma with assigned values of 0.71 and 0.78 IU/amp for FXI functional activity (FXI:C) and FXI antigen (FXI:Ag) respectively.