A biological standard for measurement of blood coagulation factor VIII activity

Bull. Org. mond. SantJ 1971, 45, 337-351Bull. Wld Hlth Org. I A Biological Standard for Measurement of Blood Coagulation Factor VIII Activity D. R. BANGHAM,1 ROSEMARY BIGGS,2 MILICA BROZOVIK,' K. W. E. DENSON 2 & JOYCE L. SKEGG 1 Because of the instability of Factor VIII (antihaemophilic factor) in plasma in vitro, and since evidence has accumulated that the level of activity varies significantly between samples of "fresh normal plasma ", it is apparent that a stable biological standard is required. Two freeze-dried preparations ofFactor VIII, one plasma and one concentrate prepara- tion, were prepared and have been assayed in a collaborative trial by 20 laboratories in 9 countries. The laboratories used their own methods and altogether 248 assays were carried out. Comparison of materials of similar type gave precise estimates with all the methods used, but when materials dissimilar in type and in Factor VIII content were compared less precise results were obtained. Sources of variation, such as differences in method, differences between operators, and day-to-day variations, were analysed. Accelerated degradation tests, carried out on the freeze-dried plasma and concentrate preparations, showed that Factor VIII activity was sufficiently stable in the concentrate for it to be used as a long-term standard but that the Factor VIII activity was less stable in the plasma preparation. Factor VIII is one of the main components of the enzyme system involved in blood coagulation. Deficiency or defect of Factor VIII leads to inability of blood to clot. Haemophilia is the main clinical disease associated with a congenital deficiency of Factor VIII and in Von Willebrand's disease there is believed to be a deficiency of the precursor. For treatment, Factor VIII is given as fresh frozen plasma, or as a cryoprecipitate (Pool & Shannon, 1965; Verstraete et al., 1969), or as a " concentrate " prepared from fresh plasma by various procedures (Kekwick & Wolf, 1957; Blomback, 1958; Wagner et al., 1964; Pavlovsky et al., 1967; Johnson et al., 1967), or as preparations from bovine or porcine plasma. The instability of Factor VIII in vitro has long been 1 Division of Biological Standards, National Institute for Medical Research, London, England. recognized and estimates have hitherto been carried out in terms of " millilitres of fresh normal plasma ". When evidence accrued that the level of Factor VIII activity varied significantly between samples of fresh normal plasma, both between individuals and in an individual from time to time (Preston & Barr, 1964; Biggs & MacFarlane, 1966), it was apparent that a stable biological standard was required. Factor VIII assays are carried out for 3 general purposes: (1) for the diagnosis and study of disorders of clotting mechanisms, (2) for monitoring the treatment of patients, and (3) for the estimation of potency in the control of batches of concentrate and cryoprecipitate. This paper describes the collaborative work done to produce and characterize a preparation to serve as 2 Research Laboratory, Oxford Haemophilia Centre, Oxford, England. 2732 - 337 - 338 D. R. BANGHAM AND OTHERS an international standard for Factor VIII activity. 1 The monitoring of patients, both for diagnosis and during treatment, needs many assays over short periods, for which large supplies of a working stand- ard are required. The provision and calibration of a freeze-dried plasma standard for this purpose could be made by suitable organizations (e.g., control authorities or transfusion services) at a national level. The relative instability of Factor VIII in freeze-dried plasma would not matter unduly if there were a high rate of consumption and replace- ment of such working standards. The assay of bovine and porcine preparations used for treatment under special circumstances are separate, much smaller-though important-prob- lems that require separate treatment. SOURCE AND PREPARATION OF MATERIALS Preparation offreeze-dried plasma P1 (66/355) In November 1966, a pool of 4.2 litres of fresh plasma from 20 donors, kindly provided by Dr J. M. Grant, Regional Blood Transfusion Centre, Oxford, England, was received at the National Institute for Medical Research, London, and distrib- uted into ampoules and freeze-dried on the same day. For details of this material and its distribution into some 3 000 ampoules, referred to in the studyas P1, see Annex 2. Preparation offreeze-dried concentrate C (67/19) In March 1967, the National Institute for Medical Research, London, received 88 bottles of freeze-dried concentrate of Factor VIII of intermediate purity made by Dr A. J. Johnson.2 For details of this material, referred to in the study as C, and its distribution into some 3 600 ampoules, see Annex 2. A secondpreparation offreeze-driedplasma P2 A second small batch of pooled plasma was collected and distributed into ampoules and freeze- ' The preparation of concentrate coded 67/19 in this paper was established as the International Standard for Coagulation Factor VIII by the WHO Expert Committee on Biological Standardization (1971) at its 23rd meeting in 1970. The International Unit for Blood Coagulation Fac- tor VIII was defined as the activity contained in 14.365 mg of the International Standard. The Expert Committee Report recommends that since the stocks of the International Standard are limited it should be reserved primarily for the calibration of national and working standards (used for monitoring treatment) organized on a centralized national basis. 2 This material was processed by E. R. Squibb and Son, New York, and was generously donated by the American Red Cross, through the good offices of Dr A. J. Johnson. dried under the same conditions as 66/355, and referred to in this study as P2 THE COLLABORATIVE STUDY Altogether 20 laboratories in 9 countries took part in the study. The names of the participants are listed in Annex 1, but elsewhere in this report they are distinguished by a laboratory number, which is not necessarily related to the order of listing in the Annex. Each laboratory sent to the National Institute of Medical Research, London (NIMR) details of the assay method it intended to use and of the number and size of the assays it was able to perform. In the light of this information, suitable assay designs were then sent to each participant together with the materials described in Table 1, labelled with the code letters only. In addition, participants were asked to include in their assays any local laboratory standards they used (referred to in this report as L) and a number of samples of fresh normal plasma. The assay results were entered on special forms, which were returned to the NIMR for statistical analysis of the data. Aims of the study The overall plan of this study was designed to determine: (1) whether ampoules of freeze-dried whole plasma, P1, or of a freeze-dried concentrate, C, would be suitable to serve as a standard for the assay of Factor VIII; Table 1. Materials in ampoules distributed to participants Code Material Referred to inletter this report as A freeze-dried plasma 1 (66/355) Pi B freeze-dried plasma 2 P2 c freeze-dried concentrate 67/19 c D ampoules of C kept at 37°C for 6 months E duplicate coded ampoules of 66/355 P F duplicate coded ampoules of 67/19 C G ampoules of Pi kept at 37°C for 10 months STANDARD FOR MEASUREMENT OF BLOOD COAGULATION FACTOR VIII ACTIVITY (2) whether it would be possible to obtain compa- rative assays of whole plasma and concentrate that would satisfy certain statistical criteria of validity -e.g., parallelism of transformed dose-response lines; (3) estimates of the stability of the two freeze- dried preparations P1 and C, by accelerated degrada- tion studies; (4) any systematic differences between estimates made with the one-stage and the two-stage assay procedures; (5) estimates of the precision obtainable with each method; (6) estimates of the relative Factor VIII activity of 1 ml of " fresh normal plasma " obtained in each laboratory to provide a basis for the unitage to be assigned to a reference material; and (7) the relative potency of local Factor VIII standards in terms of ampoules of each of the two preparations. METHODS Biological assays Factor VIII activity was estimated by two types of technique: one-stage methods consisting essen- tially of modification of the partial thromboplastin time, and two-stage methods, based on thrombo- plastin generation methods. Details of the techniques used in the study were submitted by 19 of the 20 centres. Table 2 shows the basic ingredients of the one- stage methods used: 6 of the 10 centres used a kaolin-cephalin clotting-time technique. The time of incubation of kaolin with plasma varied from 2 min to 30 min and various sources of phospholipid were used. One centre used platelet-containing plasma and added neither kaolin nor phospholipid. Two centres diluted the samples in haemophilic plasma, whereas others used buffer. Table 3 gives a similar analysis of the two-stage results: the procedures and reagents varied in detail. Statistical analysis The results obtained by both one-stage and two- stage methods were analysed as parallel-line assays by classical statistical procedures and expressed as relative potencies (or potency ratios) of the total contents of one ampoule in terms of the total contents of the other ampoule. For the purpose of this study the word " assay refers to the results of tests performed under the same experimental conditions by the same operator at the same time, from which an independent estimate of potency of at least one material relative to another may be calculated without reference to other tests. The total number of assays performed was 248; of these, 230 were valid and 18 were invalid owing to non-parallelism or curvature of dose-response curves. RESULTS Suiitability of plasma and conicentrate to serve as standards Precision of assays: replicate assays of coded duplicates ofidentical materials. One aim of the study was to assess the relative precision of assay proce- dures. To avoid a tendency to get readings of known replicates closer together than would be found if the identity of the materials were known, additional ampoules of both P1 and C bearing different code letters were included. All laboratories assayed plasma and concentrate coded duplicates. Most laboratories obtained mean potency ratios (of P1 in terms P1 and C in terms of C) close to unity; the 95 % confidence intervals for relative potencies varied from laboratory to labora- tory, as shown in Table 4. Precision of assays: the effects of comparing similar and dissimnilar materials. The two plasma preparations P1 and P, were compared with each other and with the concentrate C. The log potencies of each preparation in terms of others were calculated and the distribution of log potencies is illustrated as bistograms in Fig. I and expressed numerically in Table 5. The estimates of the relative potencies of identical materials, the plasma P1 to its coded duplicates and the concentrate C to its duplicates, are shown in the first and last histograms of Fig. 1. The distribution is normal and the variance small and similar in each case. The histogram of the plasma P2 in terms of the plasma P1 showed similar although somewhat wider distribution; the variance was appreciably greater. The distribution of estimated potencies of con- centrate C was expressed relative to the plasma P1; it was irregular and wider although still essentially normal. Likewise, the potency ratio of the concen- trate C relative to plasma P2 showed a wider distribu- tion; the variance was still greater than that of P, against P2, but much less than the variance of P1 with C. 339 340 D. R. BANGHAM AND OTHERS Table 2. One-stage methods a Lab. Haemophilic substrate plasma Phospholipid Methods of activation Dilutions 2 platelet-containing plasma, none none buffer -30 to-40°C 4 platelet-poor plasma, -20°C Bell & Alton extract BDH kaolin for 10 min then diluted buffer with H-plasma 6 H-plasma stored frozen inosithin Baker kaolin, 5 ml/ml buffer 7 platelet-poor plasma, -20°C Thrombofax kaolin added to H-plasma and incu- citrate saline bated 15 min with dilution 9 b platelet-poor plasma, -25'C Milstone extract kaolin for 30 min then dilution with buffer H-plasma and addition of phospho- lipid 15 platelet-poor, -20'C inosithin BDH 10 min. with dilution and H- buffer saline plasma 16 platelet-poor, -20'C dilute thromboplastin none saline serum reagent 17 a platelet-poor, -20'C inosithin or Thrombofax activation product H-plasma 17b platelet-poor, -20'C inosithin or Platelin activation product buffer 17c platelet-poor, -20'C Platelin or Thrombofax activation product H-plasma, not adsorbed 18 platelet-poor, -20'C not stated kaolin 2 min with H-plasma and buffer dilution 19 platelet-poor, -25°C Tachostyptan none buffer 20a platelet-poor, -20'C cephalin kaolin 2 min incubation with dilu- buffer for plasma tion and H-plasma 20b platelet-poor, -20'C cephalin kaolin 2 min incubation with dilu- H-plasma for concentrates tion and H-plasma U H-plasma = haemophilic plasma. h) End-point assessed instrumentally. Addlitionial comparisons ofdifferent types ofpprepar-a- tio)1. To test the observation that the comparison of the two dissimilar types of preparation gave rise to less precise estimates, a small study was carried out in three laboratories. Concentrate C and a cryopre- cipitate preparation were assayed against each other and against the plasma P1.' l An additional concentrate preparation was included in the study, but there were difficulties in reconstituting it and the results obtained were unreliable; these results are therefore excluded. Histograms of the relative potencies are shown in Fig. 2 and Table 6 summarizes these results numeric- ally. Whereas a wide scatter of results was obtained in comparisons of both concentrated materials with P1, the comparison ofC with the cryoprecipitate gave figures that agreed much more closely. Sources of variationi Differentces between estimates made with one-stage methods anid two-stage methods. The histograms STANDARD FOR MEASUREMENT OF BLOOD COAGULATION FACTOR VIII ACTIVITY Table 3. Two-stage methods Lab. Phospholipid Serum 1 Factor V Buffer Dilution of concentrate 1 inosithin human, stored at haemophilic plasma (1 :20) imidazole buffer -20°C 3 brain cephalin human freeze-dried bovine plasma, fractionated citrate-imidazole 1 :5 in haemophilic plasma 5 commercial mixture of bovine and human serum citrate-barbital (various) 8 Bell & Alton freeze-dried, human bovine plasma, fractionated imidazole citrate saline 10 cephalin as 3 above freeze-dried, human bovine plasma, fractionated citrate saline 1 :10 in haemophilic plasma 12 human brain freeze-dried, human bovine plasma, fractionated barbital 1 :10 in haemophilic plasma cephalin 13a inosithin human, stored at bovine serum, stored -20°C imidazole buffer -20°C 14 cephalin human, stored at bovine serum, -20'C barbital citrate saline -20'C a End-point assessed instrumentally. comparing results with two-stage methods and one- stage methods are shown in Fig. 3, using the same data as in Fig. 1. Normal distribution with little scatter was obtained with both one-stage and two-stage methods in comparison of identical materials and materials of similar type (plasma 1 with its coded duplicates, concentrate with its coded duplicates, and plasma 1 with plasma 2). When different types of materials were compared (concentrate with plasma 1 and concentrate with plasma 2), the distribution appeared less regular and wider. This was most striking with one-stage methods in the comparison of concentrate with plasma 1. The numerical results are given in Table 7. The overall mean potency ratios for P2 in terms of P1 and for C in terms of P1 (see Table 6) were lower for one-stage methods than for two-stage methods, but these differences were not statistically significant (P > 0.05). The operator effect. In 5 laboratories assays were carried out by 2 or more operators using the same technique. Variations in relative potencies obtained by different operators are compared with variations observed with one operator. Of the 22 comparisons only 3 showed significant difference (see Table 8). In some comparisons the variation observed with one operator is greater than that with different operators, although presumably this is the result of chance variation. The method effect. In 4 laboratories assays were carried out using more than one technique. Table 9 summarizes the results in 3 laboratories and some of the details of the techniques used. Laboratory 17 used 3 different modifications of one-stage methods carried out by 3 operators. The effect of different sources of phospholipid was investigated at the same time and these results are also shown in Table 9. When Laboratory 8 assayed the concentrate preparation using a two-stage method, but without haemophilic plasma as diluent, the estimated potency was only 1/3 of that found when haemophilic plasma was added. All other laboratories using two-stage methods added haemophilic plasma to the concentrate. The within-day and between-day variability. The variation in estimates obtained on the same day and on different days varied widely from laboratory to laboratory. It was greater between days than within 341 D. R. BANGHAM AND OTHERS Table 4. Coded duplicates of plasma Pi and concentrate C: results and statistical analysis Plasma Concentrate Lab. No. "' No. of Mean 95% fi cal 95% confidence N.o Mean 95% fiducial 95% confidenceLab N ." ~N o. o 95% fiduc interval ofor intervalufor ,sasPote.nicy limits .fma f Iassays. potency' limits ofmen fa say ratio ofm individual assay ratio individual assay 1 1 2 1 3 2 4 1 5 7 .2 8 1 9 3 10 1 11 method A 1 method B 1 12 1 13 3 14 2 15 2 16 2 1.09 0.98 1.09 0.97 1.01 1 .32 0.99 0.79 0.63 0.98 1.00 1.84 1.18 1 .08 1.01 1.10 0.80 0.69 2 2 2 3 3 0.85 0.82 0.73 1.31 1.16 1.23 1.10 0.94-1.25 0.79-1.22 1.03-1.16 0.63-1 .44 0.85-1.19 1.1 5-1 .52 0.87-1.12 0.59-1.06 0.45-0.89 0.82-1 .19 0.86-1.16 1.64-2.10 0.95-1.42 1.00-1.18 0.82-1.26 0.76-1.59 0.39-1.63 0.45-1.06 0.64-1.12 0.62-1.10 0.54-0.97 0.91-2.14 1.08-1.25 0.89-1.71 0.89-1 .37 0.94-1.25 0.79-1.22 1.01-1 .18 0.63-1 .44 0.85-1 .19 1.08-1.62 0.87-1.12 0.48-1.30 0.45-0.89 0.82-1.19 0.86-1.16 1.64-2.10 0.85-1.52 0.96-1.22 0.75-1.37 0.65-1 .86 1 .2 1 i 1 13 2 13 0.39-1.63 1 0.45-1.06 1 0.57-1 .25 2 0.55-1.23 2 0.48-1.10 2 0.91-2.14 1 1.08-1.25 1 0.70-2.18 2 0.76-1.60 4 0.81-1.38 - 0.84-1.17 2 0.97 1.04 0.88 1.13 1.06 1.22 0.82 0.80 1.12 1.16 1.20 0.86 0.98 1.03 0.77 1.11 0.98 0.81 0.80 1.08 0.99 1.60 0.91 1.06 0.85-1.09 0.91-1.18 0.65-1 .17 0.82-1.59 0.85-1 .32 1.07-1.38 0.72-0.94 0.38-1.70 0.74-1 .73 0.90-1.51 1.13-1.29 0.84-0.87 0.90-1.06 0.98-1 .09 0.41 -1.42 0.81 -1.74 0.74-1.29 0.61 -1.08 0.61 -1.04 0.88-1 32 0.91 -1.09 1.31 -1.95 0.74-1.11 0.87-1.30 0.85-1 .09 0.91 -1 .18 0.65-1.17 0.82-1.59 0.85-1.32 1 .02-1 .46 0.72-0.94 0.28-2.33 0.74-1.73 0,90-1.51 1.1 3-1 .29 0.83-0.88 0.87-1.10 0.95-1 .13 0.41 -1.42 0.81 -1 .74 0.66-1 .44 0.55-1 .21 0.54-1.16 0.88-1.32 0.91 -1.09 1.21-2.11 0.60-1 .36 0.80-1 .41 ("For details of methods see Appendix Table 1. 342 1 7 operator 1 method A method B operator 2 method A method B operator 3 18 19 20 operator 1 operator 2 operator 3 operator 4 1 STANDARD FOR MEASUREMENT OF BLOOD COAGULATION FACTOR VIII ACTIVITY Fig. 1. Frequency distribution of estimates of log potencies: (A) plasma 1 relative to plasma 1 (coded duplicates); (B) plasma 1 relative to plasma 2; (C) concentrate relative to plasma 1; (D) concentrate relative to plasma 2; and (E) concentrate relative to concentrate (coded duplicates). Table 5. Mean potency ratios and measures of variation for each comparison of preparations tested in the collaborative study Comparison assNoof MeancCompaison assays potency ratioI Pi with Pi P2 with Pi C with Pi C with P2 C with C 42 93 130 81 38 0.84 4.43 5.95 Variance (logio) 0.01 0.10 0.34 0.17 0.007 S. D. as percentage of mean 27 107 283 158 22 6 A 4- 2 Log potency 0 Potency ratio 1.0 i C >1 m tn ~0 F= _ ~~~~~~~I 0.2 0.4 0.6 0.8 4 1.0 3.2 10.0 4 B 2 0 p. Log potency 0.2 0.4 0.6 0.8 1.0 6 4 2 0 c Log potency 0 0.2 Potency ratio 1.0 3.2 1.8 +1.0 10.0 _ 3.6 Fig. 2. Frequency distribution of estimates of log potencies: (A) cryoprecipitate relative to concentrale; (B) cryoprecipitate relative to plasma 1; and (C) concentrate relative to plasma 1. a day, and wider with one-stage methods than with two-stage methods. The numerical values for these observed effects are shown in Table 10. Results obtained with local laboratory standards Each laboratory included its local standard or standards in the trial. Table 11 lists various types of local standard used. It was assumed that local laboratory standards were equal in activity to " average " fresh normal plasma, unless otherwise stated, in which case the activity was adjusted accordingly. When log potencies of local standards were ex- pressed in terms of fresh normal plasma (see Fig. 4) the distribution of estimates was symmetrical and narrow. All types of local standard showed potencies approaching unity. The mean potency calculated A -~~~~~~C10 _ 5 -I Log potency 0 +0.25 20- IC 15 ; Log potency 0 +0.! Log potency 0 +0.50 +1.0 +1.50 E 5- n - 0.25 0 +0.25Log potency . __O 10_,67 ... ... ^ KU _ ---- 343 X t0.50i Flarlm .50 +1.0 +1.50 m im D. R. BANGHAM AND OTHERS Table 6. Additional comparisons of dissimilar types of preparations Comparison cryoprecipitate with concentrate cryoprecipitate with plasma 1 concentrate with plasma 1 No. Mean of potency assaysl ratio 11 11 15 1.026 3.95 4.38 Limits of distribution (95% confidence interval) 0.71-1.47 1.49-10.5 1.49-12.8 from the results from all laboratories for all local standards was 0.978 ml of " average fresh normal plasma ". This indicates that the participating laboratories were accurate in assessing the activity of their local standards. Comparison of local standards to plasma 1 gave a wide range of estimates. Pooled normal plasma standards and plasma from one donor showed a particularly wide spread of relative potencies (see Fig. 4). Local standards were also expressed in terms of concentrate C. The distribution was normal and the extent of the range was comparable to that observed with fresh normal plasma. Plasma P1 was calculated to be equivalent to 0.63 ml and concentrate C to 2.9 ml of " average" laboratory standard. Activity of P1 anid C relative to fresh normal plasma The preparations P1 and C were assayed against a total of 289 fresh normal plasmas. Fig. 3. Comparison of frequency distribution of esti- mates of log potencies obtained by one-stage and two- stage methods. The histograms of results with two- stage methods are superimposed over the histograms of results with one-stage methods: (A) plasma 1 relative to plasma 1 (coded duplicates); (B) plasma 1 relative to plasma 2; (C) concentrate relative to plasma 1; (D) concentrate relative to plasma 2; and (E) con- centrate relative to concentrate (coded duplicates). Unshaded areas = one-stage methods; diagonal shading = two-stage methods. Table 7. Sources of variation: comparison of one-stage and two-stage methods (combined results from all laboratories) One-stage methods Mean Variance S. D. as potency ratio (logIo) percentage 0.80 4.19 Two-stage methods No. of Mean Variance S. D. as assays potency ratio (log,o) percentageof mean 0.01 26 11 - 0.14 138 24 0.99 0.46 376 33 5.23 0.20 180 1 6 5.00 0.01 26 10 0.01 0.007 0.05 0.06 0.003 15.-15 10 ai 2'1 Log potency 0 +025 -V 0 +0.50 15 0 10 5- 0 'n Log potency 0 +0.50 + 1.0 +1.5 10 Log potency -0.25 0 +0.25 - - -------- . ......__hCz~~~~~0 10369 Comparison No. of assays P1 with P1 P2 with Pi C with Pi C with P2 C with C 26 22 68 76 13 344 STANDARD FOR MEASUREMENT OF BLOOD COAGULATION FACTOR Vill ACTIVITY Table 8. Comparison of variations in relative potencies obtained by one operator (within operator) and by different operators (between operators) as measured by standard deviation as percentage of mean Lab. No. Comparison a P2 with Pi C with Pi C with P2 L with Pi L with P2 W B W B W B W B W BI~ _ 7 13% 9% 54% 88% 1% b 9 12% 13% 7% 7% 6% b 14 13% 8% 10% 11% 1% 16 15% 5% 23% 12% 60% 20 30% 63% 41% 55% J 34% b 22% b 14% 6% 10% 8% 14% 84% 25% 1% 23% 24% 22% b 7% 15% 47% (c W = within operator; B = between operators b Significant differences (0.01. P<0.05) Table 9. Comparisons of different methods Methods or modifications compared 82% b 6% 41% 85% Results 3 two-stage assay, Biggs & Macfarlane two-stage assay, modification with combined reagent no significant difference 1 0 one -stage assay two-stage assay no significant difference 11 one-stage method with artificial haemophilic plasma one-stage method with haemophilic plasma no significant difference except for comparison of C with Pi 1 7 different sources of phospholipid (inosithin, Platelin, Thrombofax) no significant difference 17 effect of diluting with buffer instead of with haemophilic plasma no significant difference for compari- son of Pi with P2; significant differ- ences for comparisons of C with Pi and of C with P2 no significant difference for compari- son of Pi with P2; significant differ- ences for comparisons of C with Pi and of C with P2 1 7 effect of absorption of samples with aluminium hydroxide Fig. 5 shows the histogram of estimates of plasma I expressed in terms of 1 ml of 233 fresh normal plasmas. The distribution of log potencies is fairly wide. The overall mean potency ratio is 0.63 ml (95% confidence limits: 0.59-0.67 ml) of " average fresh normal plasma ". Likewise the histogram of the concentrate prepara- tion expressed in terms of I ml of 167 fresh normal plasmas shows a similarly wide distribution. The overall mean is 2.58 ml (2.34-2.84 ml) of " average fresh normal plasma ". Stability of freeze-dried plasma (66/355) and conI- centrate (67/19) on accelerated degradation In order to make the changes associated with a loss of biological activity proceed at a measurable rate, samples of both materials were subjected to higher temperatures for different times and their potency was estimated by direct comparison with ampoules stored at -20°C. Samples of both plasma 66/355 (P1) and concen- trate 67/19 (C) that had been stored at + 37°C for 10l/2 months and 6 months, respectively, were in Lab. No. 345 D. R. BANGHAM AND OTHERS Table 10. Within-day and between-day variation in estimates of Factor Vil activity within a laboratory, measured as standard deviation as percentage of mean One-stage methods Two-stage methods Compa- rison Between Within-day| Between _ _ Within-day days WihndyL days P2 with Pi 19% 36% 7% 19% C with Pi 24% 45% 9% 12% C with P2 26% 44% 11% 20% L with Pi 17% 46% 8% 16% L with C 20% 47% 12% 23% A +0.50 3.2 B 3050 1.0 3.2 15 r C hRM.'r o in 1 15 10 5 0 Log potency -0.50 Polency ratio 0.3 Log potency - 0.50 Potency ratio 0.3 0 +0.50 1.0 3.2 WII3 Q1031C Fig. 4. Frequency distribution of estimates of log potencies: (A) local laboratory standards relative to fresh normal plasmas; (B) local laboratory standards relative to plasma 1; and (C) local laboratory standards relative to concentrate. Unshaded = one-donor plasma; stippled = pooled normal plasma; diagonal shading = reference plasma; cross-hatched = concentrate. Table 11. Local standards used in the trial Type of local standard No. of laboratories concentrate preparations (animal and human) 5 a reference plasmas 7 a pooled normal plasma 5 plasma from one donor 3 no data 1 a One laboratory used concentrate and reference plasma simultaneously as local standard. fact included as coded samples in the international collaborative study. The results of the tests on these samples, although generally reassuring, did not provide enough evidence from which to calculate a sufficiently reliable estimate of stability for each preparation. A 1 I, lea- Log potency - 0.50 Potency ratio 0.31 U U u 0.6 1.0 B J Log potency -0.50 0 Potency ratio 0.31 1.0 3.2 +0.50 +10 3.2 10.0 _"O 103 Fig. 5. Frequency distribution of estimates of log potencies: (A) plasma 1 relative to 233 fresh normal plasmas; and (B) concentrate relative to 167 fresh normal plasmas. VA U~~~~~~~~~~~ _.- ... ..... 346 m STANDARD FOR MEASUREMENT OF BLOOD COAGULATION FACTOR VIII ACTIVITY Table 12. Accelerated degradation studies Tempe- Length Total No. of Preparation rature of no. of valid (~) storage as|( C) (months) assays assays plasma (66/355) +4 26 9 9 +20 26 9 9 +37 10% 14 2 +37 12% 9 9 concentrate (67/19) +20 15 6 6 +37 6 19 19 +37 1 5 6 6 The evidence on the relative stability of the two preparations was therefore supplemented with experiments additional to those planned in the collaborative study, and carried out by 3 laboratories in Great Britain. Table 12 shows the details of samples subjected to accelerated degradation as well as the number of assays performed on each. It is of interest to note the relatively large number of invalid assays with plasma kept at +37°C for 101/2 months. This may be taken to suggest that components in plasma 66/355 underwent alteration at 37'C, which made the preparation substantially dissimilar from the preparation stored at lower temperatures. There were no invalid assays with plasma at lower temperatures, or with the concentrate at any temperature of storage. Using the Arrhenius plot (Jerne & Perry, 1956) the degradation constants for each preparation have been calculated from these estimates, and the loss of potency calculated by extrapolation to the tempe- rature and conditions under which the bulk of the ampoules are stored (-20°C). The results obtained with these samples are summarized in Table 13. These results indicate that the plasma preparation will lose about 22% of its potency in 10 years at -20°C, or about 2%0 each year. The concentrate preparation will lose less than 0.1% of its potency under the same conditions of storage. DISCUSSION AND CONCLUSIONS Comparison of identical materials gave a mean potency of 1.0 in almost all laboratories and a narrow distribution of log potencies. This provided a useful starting point in assessing the significance of variations observed in comparisons of log potencies of non-identical materials. Comparison of materials of similar type and with similar Factor VIII content (plasma with plasma or concentrate with concentrate preparation) appeared to give satisfactory results irrespective of the type of method used, although the range of estimates was narrower with two-stage methods. When two dissimilar materials (concentrate and plasma) were compared with each other the results obtained were more variable. Although both two-stage and one- stage methods gave less precise answers, the varia- bility within and particularly between laboratories was more striking with one-stage tests (see Table 7 and Table 10). This may result from the use of different diluents for the concentrate preparations, to inherent differences in the haemophilic plasmas used, to the sensitivity of one-stage methods to " activated Factor VIII ", or-which is more likely-to a combination of many factors (Kekwick & Walton, 1964; Niemetz & Nossel, 1969; Rapaport et al., 1963). It is evident that more reproducible results were obtained in one laboratory on one day using the same set of reagents, than on different days under different experimental conditions (Table 10). The change of operator had little or no effect on the reproducibility of the results within a laboratory (Table 8). The comparison of local laboratory standards with fresh normal plasma samples showed that the Table 13. Degradation constants 347 D. R. BANGHAM AND OTHERS participating laboratories were accurate in assessing the activity of their local standards in terms of " average" normal plasma. It is of interest that greater variation was noted in comparisons of local standards with plasma 1 than in comparison with concentrate. This was most marked in laboratories using pooled normal plasma and plasma from one donor as local standard. The comparison of freeze-dried plasma P1 with fresh normal plasmas emphasized once more the variability of Factor VIII content in fresh normal plasmas. Similar variability, although on a smaller number of assays, was, not surprisingly, observed when concentrate C was compared with fresh normal plasma. The extent of the effect of the instability of plasma P1 in the comparisons with plasma P2 and with concentrate C, local laboratory standards, and normal plasmas should be considered. Most of the assays were performed during the first year of storage, when the expected loss was only 2% at -200C. Some laboratories kept the freeze-dried samples at +40C for a short time: even if kept for a year at +4°C the loss would have been only 6o%. It can be assumed, therefore, that the instability of plasma contributes little to the variations observed in comparisons of different materials. Analysis of stability tests on the concentrate and plasma preparations showed: (1) that the Factor VIII activity in the preparation of plasma 66/355 is not stable enough for it to serve as a long-term standard for Factor VIII activity, and (2) that the concentrate 67/19 is evidently stable enough for this purpose. All participants agreed that the figure assigned for the unitage of a standard should be based on the number of ml of " average fresh normal plasma " estimated to contain the Factor VIII activity of one ampoule of the preparation. In the collaborative study the overall mean estimate of the assays in which the concentrate 67/19 was compared with 167 samples of fresh normal plasma showed that each ampoule contained Factor VIII activity equivalent to that found in 2.6 ml of average fresh normal plasma and this is the recommended unitage for this preparation. On some grounds it might be considered more desirable to have a standard consisting of a plasma rather than a concentrate. Nevertheless stability is one of the most important attributes of a standard and the evidence is now unequivocal that the concentrate is stable and the plasma preparation is less stable. However, successive batches of freeze- dried plasma can be prepared, calibrated in terms of concentrate preparation and used as a working standard for the assay of Factor VIII in plasma. The lifetime of each such batch should be treated as not more than about one year unless its stability has been found to be adequate. ACKNOWLEDGEMENTS The authors thank Dr J. M. Grant, Regional Blood Transfusion Centre, Oxford, England, and Dr A. J. Johnson, American National Red Cross Research Laboratories, New York, USA, for arranging for the supply of the materials; Miss M. V. Mussett and Miss I. Robertson, National Institute for Medical Research, London, for additional statistical advice; and Dr P. J. Campbell, National Institute for Medical Research, London, for ampouling the materials. RESUME tTALON BIOLOGIQUE POUR LE TITRAGE DE L'ACTIVITI, DU FACTEUR VIII DE LA COAGULATION SANGUINE On a procede, dans 20 laboratoires de 9 pays, a une etude collective de deux preparations de facteur VIII, un plasma lyophilise (66/355) et un concentre lyophilise (67/19). Au total, 248 essais biologiques ont e effectues; les deux preparations ont e eprouvees l'une vis-a-vis de l'autre ou au regard d'etalons de travail locaux et d'ehantillons de plasma normal frais. Le titrage compa- ratif a porte en outre sur une autre preparation de plasma lyophilis6 et sur un cryopreeipite. Lorsqu'on a compare des materiels de meme type (plasma-plasma ou concentr&concentre), on a obtenu des estimations preeises avec toutes les techniques utilisees. Par contre, les essais sur des materiels differant quant au type et a la teneur en facteur VIII (concentre-plasma) ont donne de moins bons resultats, surtout avec la methode 348 STANDARD FOR MEASUREMENT OF BLOOD COAGIJLATION FACTOR VIII ACTIVITY 349 en un temps. D'autres elements susceptibles d'influencer les resultats (moment du titrage, changement d'operateur ou de technique) ont e aussi analyses. On a calcule que l'activite d'une ampoule de plasma 66/355 correspondait a 0,63 ml, et celle d'une ampoule de concentre 67/19 a 2,9 ml de l'etalon de travail ( moyen ). Un millilitre de l'etalon de travail << moyen )) utilise au cours des essais equivalait a 0,978 ml de plasma normal frais # moyen )>. Des tests de stabilite ont montre que l'activite du facteur VIII dans le plasma 66/355 n'etait pas suffisamment stable pour que cette preparation convienne comme etalon de longue duree. En revanche, la stabilite du concentre 67/19 est telle qu'on peut envisager de l'employer pour cet usage. Les essais comparatifs du concentre et de 167 echantillons de plasma normal frais ont conduit a attribuer a chaque ampoule de concentre une activite equivalant a l'activite de 2,6 ml de plasma. 11 peut y avoir interet a disposer de plasma comme etalon. On peut en preparer des lots successifs, etalonnes par rapport au concentre, mais en veillant a ne pas les utiliser comme reactifs de travail pendant plus d'un an, en raison de leur stabilite limitee. REFERENCES Biggs, R. & Macfarlane, R. G. (1966) Treatment of haemophilia and related disorders, Oxford, Blackwell Scientific Publications Biggs, R., Macfarlane, R. G., Denson, K. W. E. & Ash, B. J. (1965) Brit. J. Haemat. 11, 276 Blomback, B. (1958) Ark. Kemi, 12, 387 Britten, A. & Grove-Rasmussen, M. (1966) Transfusion, 6, 230 Brown, D. L., Hardisty, R. M., Kosoy, M. H. & Bracken, C. (1967) Brit. med. J., 2, 79 Finney, D. J. (1964) Statistical method in biological assay, London, Charles Griffin & Co. Ltd., London, 1964 Hardisty, R. M. & Ingram, G. I. C. (1965) Bleeding disorders, Oxford, Blackwell Scientific Publications Jerne, N. K. & Perry, W. L. M. (1956) Bull. Wld. Hlih. Org., 14, 167 Johnson, A. J., Newman, J., Howell, M. B. & Puszkin, S. (1967) In: K. M. Brinkhous et al., ed., Platelets: their role in haemostasis and thrombosis, Stuttgart, Schattauer, p. 377 Kekwick, R. A. & Walton, P. L. (1964) Brit. J. Haemat., 10, 299 Kekwick, R. A. & Wolf, P. (1957) Lancet, 1, 647 Michael, S. E. & Tunnah, G. W. (1966) Brit. J. Haemat., 12, 115 Niemetz, J. & Nossel, H. L. (1969) Brit. J. Haemat. 16, 337 Pavlovsky, A., de Tezanos Pinto, M. & Martinez Cana- veri, A. (1967) In: K. M. Brinkhous et al.,ed., Platelets: their role in thrombosis and haemostasis, Stuttgart, Schattauer, p. 383 Pool, J. G., Hershsold, E. J. & Pappenhasen, A. (1964) Nature (Lond.), 203, 312 Pool, J. G. & Shannon, A. E. (1965) New Engl. J. Med., 273, 1443 Preston, A. E. (1967) Brit. J. Haemat., 13, 42 Preston, A. E. & Barr, A. (1964) Brit. J. Haernat., 10, 238 Rapaport, S. I., Schiffman, S., Patch, M. J. & Ames, S. B. (1963) Blood, 21, 221 Spaet, T. H. & Carner, E. S. (1955) J. Lab. Clin. Med., 46, 111 Thelin, G. M. (1968) Thrombos. Diathes. haemorrh. (Stuttg.), 19, 423 Verstraete, M., Olislaegers, P., Itterbeek, van H., Waumans, P. & Lust, A. (1969) Vox Sang. (Basel), 16, 382 Wagner, R. H., McLester, W. D., Schmith, M. & Brinkhous, K. M. (1964) Thrombos. Diathes. haemorrh. (Stuttg.), 11, 64 Annex 1 PARTICIPANTS IN THE COLLABORATIVE ASSAY ARGENTINA Dr A. Pavlovsky & Dr A. Martinez Canaveri Instituto de Investigaciones Hematologicas Buenos Aires AUSTRIA Professor E. Deutsch First Department of Internal Medicine University of Vienna Vienna BELGIUM Professor Marc Verstraete Kliniek voor Inwendige Geneeskunde Leuven FRANCE Dr M. J. Larrieu Institut de Recherches sur les Maladies-du Sang H6pital St. Louis Paris D. R. BANGHAM AND OTHERS Professor J. P. Soulier & Dr D. Mdnache Centre National de Transfusion Sanguine Paris NETHERLANDS Dr K. Deggeller Central Laboratory Blood Transfusion Service of the Dutch Red Cross Amsterdam Dr E. A. Loeliger & Dr J. J. Veltkamp Interne Geneeskunde Academisch Ziekenhuis Leiden SWEDEN Dr Birger Blomback Department of Blood Coagulation Research Karolinska Institute Stockholm SWITZERLAND Dr R. Butler Central Laboratory Blood-Donor Service Swiss Red Cross Berne UNITED KINGDOM OF GREAT BRITAIN AND NORTHERN IRELAND Dr Rosemary Biggs Oxford Haemophilia Centre Churchill Hospital Oxford Dr Katharine M. Dormandy Haemophilia Centre Royal Free Hospital (N.W. Branch) London An, A preparation offreeze-dried plasma labelled 66/355 In November 1966, a pool of 4.2 litres of fresh plasma, kindly provided by Dr J. M. Grant of the Regional Transfusion Centre, Oxford, was received at the National Institute for Medical Research, London. The blood was centrifuged twice at +4°C to get rid of cells before pooling. In the afternoon of the same day the pooled unfiltered plasma was distributed at room tempera- ture, in equal amounts, into some 3 000 hard-glass ampoules. Vessels used for the plasma were pre- Professor A. S. Douglas & Dr C. D. Forbes Department of Medicine, Royal Infirmary Glasgow Dr R. M. Hardisty Department of Haematology Hospital for Sick Children London Dr C. I. C. Ingram & Dr J. M. Ziemski Louis Jenner Laboratory St. Thomas' Hospital London Dr W. d'A. Maycock Lister Institute of Preventive Medicine Elstree, Herts. UNITED STATES OF AMERICA Dr B. Alexander New York Blood Center New York Dr C. McMillan Department of Paediatrics University of North Carolina Chapel Hill, N.C. Dr R. Murray Division of Biological Standards Department of Health, Education, and Welfare National Institutes of Health Bethesda, M. Dr Judith Pool Division of Haematology S - No. 161 Department of Medicine Stanford Medical Center Stanford University California Dr H. R. Roberts Department of Pathology University of North Carolina Chapel Hill, N.C. iex 2 viously rinsed out with sodium citrate; the ampoules were steamed in acid, thoroughly rinsed, and sterilized by dry heat in a clean oven. Check weighings of 47 ampoules showed an average content of 1.016 g ± 0.65% plasma per ampoule. Approximately 10-12 hours after the blood had been obtained from the donors the plasma in the ampoules was frozen to - 35°C and placed in liquid nitrogen and then freeze- dried as one batch at -35°C. Thereafter the ampoules were fitted with plastic plugs designed to limit gaseous diffusion and subjected to secondary desiccation for 13 days. After the ampoules had been 350 STANDARD FOR MEASUREMENT OF BLOOD COAGULATION FACTOR Vill ACTIVITY filled with pure dry nitrogen they were sealed by fusion of the glass. Test weights on 6 ampoules showed a mean of 91.33 mg of solid contents per ampoule (see Table 14). The moisture content was estimated on 3 ampoules as loss of weight by the material after heating to 56°C over P2O5 in a vacuum for 5 hours, and was less than 1 %. The oxygen content of ampoules was less than 0.150% (mean of 3 ampoules, estimated by mass spectrometer). After they had been tested for cracks and pinholes, the ampoules were stored at -20°C in the dark. This material is referred to in the study as P1. A prepar-ation offreeze-dried concentrate labelled 67/19 In March 1967, the National Institute for Medical Research, London, received 88 bottles of a freeze- dried concentrate of Factor VIII of intermediate purity, made by Dr A. J. Johnson. The material, processed by E. R. Squibb and Son, New York, was generously donated by the American National Red Cross through the good offices of Dr A. J. Johnson. It consisted of 46 vacuum-sealed screw-capped bottles labelled Batch 565 and 42 bottles labelled Batch 571. After breaking the vacuum-seal, the material was dissolved in twice glass-distilled sterile water at room temperature. The contents of each bottle was passed through sterile acid-washed glass wool and then pooled. This solution was then distributed in equal amounts into some 3 660 ampoules and sub- jected to the same conditions of prefreezing and freeze-drying, secondary desiccation, and sealing as for preparation 66/355. The average wet weight contents of 55 ampoules was 1.023 g ± 0.94%. The mean dry weight of the contents of sealed ampoules was 37.35 mg and the moisture content was estimated to be less than 0.5 %. The mean oxygen content was 0.06%. After they had been tested for pinholes and cracks, the ampoules were stored at -20°C in the dark. This material is referred to in the study as C. 6 351