Accuracy of FVIII:C assay by one-stage method can be improved using hemophilic plasma as diluent



    1. Agency for Hemophilia and Regional Reference Center for Inherited Bleeding Disorders, Azienda Ospedaliera Universitaria Careggi, Florence, Italy
    Search for more papers by this author

    1. Agency for Hemophilia and Regional Reference Center for Inherited Bleeding Disorders, Azienda Ospedaliera Universitaria Careggi, Florence, Italy
    Search for more papers by this author

    1. Agency for Hemophilia and Regional Reference Center for Inherited Bleeding Disorders, Azienda Ospedaliera Universitaria Careggi, Florence, Italy
    Search for more papers by this author

M. Morfini, Agency for Hemophilia, Azienda Ospedaliero-Universitaria Careggi, Viale G.B.Morgagni, n.85, I-50134, Florence, Italy.
Tel.: +39 055 7947587; fax: +39 055 7947794; e-mail:


Summary. Background: The basic prerequisite of factor (F) VIII clotting assay (FVIII:C) by one-stage method is that all clotting factors other than FVIII are present in constant concentration in each dilution of both standard reference and patient plasma curves. Indeed, the plasma content of each dilution should decrease as the dilution factor increases. Objectives and methods: To keep the plasma content exactly constant in each mixture, we performed all dilutions of both standard reference and patient plasma with FVIII-deficient plasma and with a fixed amount of buffer (method B). To show the discrepancies between this method and regular method A, using buffer to make dilutions, a comparative study was conducted on FVIII:C assay on samples at known FVIII concentration and in patient plasma. Imidazole or Owren's buffers and five different activated partial thromboplastin time reagents were employed, both in method A and B. Results: A discrepancy between FVIII:C assays obtained by methods A and B was observed, mainly when Pathrontin SL and Imidazole buffer were used. The assays derived from method B always fitted better with the expected, calculated, values of FVIII:C concentrations. Furthermore, FVIII:C was assayed in 60 patients: the outcome of method A was always higher than values of method B. The discrepancy between the two methods was higher at FVIII concentrations below 50 U dL−1 but nil at 100 U dL−1. The A slope was steeper than B slope and the difference was statistically significant starting from 1/10 dilution. Accordingly, FVIII:C of patients’ plasma obtained by method A was always higher than those obtained by method B, even two or three times for FVIII level ≤10 U dL−1 or 1.4–1.6 times for FVIII levels between 10 and 25 U dL−1. Conclusions: Only method B is able to give FVIII:C assays in agreement with the expected values. The dilution of reference standards and samples with FVIII-deficient plasma is crucial to accurately evaluate the postinfusion FVIII concentrations in pharmacokinetic studies or the trough level during prophylactic therapy and to investigate the discrepancy among different FVIII:C assays. In addition, the assessment of severity and classification of hemophilia should be reviewed.


One-stage method for factor (F)VIII assay has been well established since 1962 [1]. Many variables are involved in this method but the most important are the following: (i) factor-sensitivity and phospholipids pattern of partial thromboplastin (platelet substitute) [2]; (ii) addition of albumin 1% to buffer used in performing the dilutions [3]; and (iii) von Willebrand factor (VWF) content of substrate plasma [4].

Standard procedure for FVIII:C assay requires four or more doubling dilutions of reference plasma and three of the patient plasma, all in buffer. A fixed amount of each dilution is added to a tube containing substrate plasma, depleted of FVIII, activated partial thromboplastin. CaCl2 is added at the end as trigger. The clotting times (CTs) of three dilutions of the sample are compared with the CTs of three dilutions of the reference standard, by a simple mathematical procedure according to the parallel line method [5]. A semi-log plot (dilutions or FVIII concentrations on the log scale of x-axis and CTs on the linear scale of y-axis) may be useful to check graphically the parallelism, the most important prerequisite for the accuracy of the assay. According to Ingram [5], the slope index (a ratio between the slopes of standard and sample curves) should not be lower than 0.8 when FVIII/IX-deficient plasma is tested.

Notwithstanding this standardization, a large inter- and intra-laboratory variability of FVIII:C assay was observed in several multi-center exercises [6–10]. As far as the FVIII:C potency evaluation of concentrates is concerned, the variability of the assay has been greatly reduced by predilution with hemophilic plasma, up to 100 U dL−1, and by usage of product-specific standard according to the ‘like vs. like’ principle [11–17]. Unfortunately, this improvement has not been achieved in assay of FVIII:C in patient plasma, especially for low concentrations. Even recently, very disappointing results have been reported in a multi-center quality control study on FVIII:C assay and classification of hemophilia A, conducted among Haemophilia Centres in the UK [18,19].

There is no doubt that the assay of FVIII:C is still a very hot issue in coagulation laboratory practice [20]. In this regard, the preanalytical conditions play a very important role in the accuracy of clotting factor assay: among those, the handling and dilution of standard and samples are the most crucial. Many years ago, Frandsen et al. [21] observed that the predilution of both FVIII concentrate and reference plasma with reconstituted lyophilized or fresh FVIII-deficient plasma resulted in activated partial thromboplastin time (APTT) values shorter than those obtained by predilution in buffer alone. The author showed that this difference was because of the regular content of fibrinogen, FV, and vitamin-K-dependent factors in dilutions made by FVIII-deficient plasma. For this reason, the use of FVIII-deficient plasma to dilute not only the reconstituted concentrate up to 100 U dL−1, but also the FVIII standard to build the reference curve was strongly recommended. To date, these findings have been carefully considered for FVIII:C potency estimation of concentrates but very frequently disregarded as far as FVIII:C assay in patients is concerned, both in building of reference curve and in preparing the dilutions of patient's plasma.

Another source of variability has been introduced by the usage of FVIII-immunodepleted substrate plasma, sometimes made free by this procedure is VWF. In this case, the content of VWF of the final mixture is not constant because it depends only on the amount carried into the final mixture by the dilution of patient's plasma, reference standard or FVIII concentrate sample, when diluted up to 100 U dL−1 with hemophilic plasma.

Finally, the chemical characteristics of the buffers used to perform the dilution may be different and may interfere with the reaction of commercial platelet substitutes and clotting factors of the final mixture.

Aims of the study

Following some of our observations on the discrepancy of potency assay of FVIII concentrate prediluted with buffer or hemophilia A plasma (data not shown), we designed a study on the impact of some components of the clotting mixture that can affect the FVIII:C assay as plasma content, different commercial partial thromboplastins, and buffers. To keep all clotting factors exactly constant in each mixture, we implemented a new dilution procedure using commercial FVIII-immunodepleted plasma, with normal VWF content, and a fixed amount of buffer to dilute both FVIII:C reference standards and samples.

To validate this modification of one-stage method, we compared the assays of FVIII:C in hemophilia-simulated plasma and in patient plasma, performed according to both the traditional method (A) and the modified method (B), in different arrays of partial thromboplastin and buffer.

Materials and methods

Equipment and reagents

Coagulometer: Behring Coagulation Timer (BCT) by Dade-Behring .

FVIII-deficient plasma: immunodepleted commercial plasma from Dade-Behring has been employed both for prediluting samples and as substrate plasma. The concentration of clotting factors was checked and the results were the following: fibrinogen 232 mg dL−1, FII 96 U dL−1, FV 95 U dL−1, FVII 80 U dL−1; FX 93 U dL−1, VWF:Ag 78 U dL−1, VWF:RCo 44 U dL−1, FIX 62 U dL−1, FXI 62 U dL−1, and FXII 93 U dL−1. The FVIII content was far below 1 U dL−1 because the APTT, after addition of buffer (the so-called ‘buffer time’) ranged from 120 to 140 s. The APTT of 1/640 dilution of reference plasma, corresponding to FVIII 0.8 U dL−1, ranged from 95 to 110 s.

APTT reagents: we used the following partial thromboplastins (lipid source + activator).

  • 1Pathromtin SL from Dade-Behring (vegetable phospholipids + silica dioxide);
  • 2Actin from Dade-Behring (rabbit brain thromboplastin + ellagic acid);
  • 3Actin FS from Dade-Behring (soya phosphatides + ellagic acid);
  • 4Actin FSL from Dade-Behring (soya phosphatides + rabbit brain thromboplastin + ellagic acid);
  • 5SynthASil from Instrumentation Laboratory (synthetic phospholipds + colloidal silica).

FVIII standard: commercial plasma standard (Dade-Behring), calibrated against the WHO-standard.

FVIII concentrates: Beriate P (ZLB Behring ), Emoclot (Kedrion ).

Buffers: 1. Imidazole buffer (5.85 g NaCl, 3.4 g imidazole in 500 mL of distilled water, 18.6 mL of 1 N HCl, filled to 1000 mL with distilled water), pH 7.35 (DADE-Behring) with or without bovine serum albumin (BSA) 1%.

2. Owren's buffer (125 mm NaCl, 28 mm Barbital-Na, adjusted to pH 7.35 with 1 N HCl) (Dade-Behring).

Calcium chloride: 0.025 m solution (Dade-Behring).

Calibration curves

Method A (dilution of FVIII standard in buffer)  Eight calibrating samples were obtained by double dilution of plasma standard in imidazole buffer or in Owren's buffer, starting from 1/5 to 1/640.

Method B (dilution of FVIII standard in FVIII-deficient plasma)  Eight calibrating samples were obtained by doubling predilutions of plasma standard in FVIII-deficient plasma, starting from undiluted standard to 1/128; each sample was further diluted 1/5 with imidazole or Owren's buffer automatically by BCT.

One-stage clotting method  FVIII-deficient plasma, standard or sample dilution, and partial thromboplastin were automatically dispensed and mixed in equal volumes (50 μL) in the reaction tube of BCT. After 120 s, 50 μL of CaCl2 0.025 m were added and the CT was recorded until an increase of light absorbance of 100 mOD was reached. For assessment of FVIII:C activity, three dilutions, 1/5, 1/10 and 1/20 of each sample, obtained by method A or B, have been tested in duplicate, in a crossover design. The mean of CTs and each dilution factor have been entered in the ‘parallel line method’ calculation program together with three corresponding pair of data of the calibration curve of methods A or B, respectively [5].

FVIII samples for testing

Commercial FVIII plasma standard  Three samples of known FVIII content have been prepared according to the following procedure: Sample 1 (FVIII level: 43 U dL−1) was obtained by 1/2 dilution of the FVIII Commercial Standard (FVIII content: 86 U dL−1) with FVIII-immunodepleted plasma (Dade-Behring), Sample 2 (FVIII level: 8.6 U dL−1) by 1/10 dilution of commercial standard, as for Sample 1. Sample 3 (FVIII level: 0.86 U dL−1) by 1/10 dilution of Sample 2 with FVIII-immunodepleted plasma (Dade-Behring). The samples were prepared ex novo on each session and assayed in duplicate by methods A and B, according to a crossover design where five different partial thromboplastins (Pathromtin SL, Actin, Actin FS, Actin FSL and Synthasil) and two buffers (Owren's and imidazole buffers) have been used.

Simulated plasma  Both Beriate P and Emoclot were diluted with FVIII-deficient plasma (Dade-Behring) up to 100 and 10 U dL−1 and assayed according to methods A and B, using Pathromtin SL and imidazole buffer. FVIII:C activities were calculated vs. the 100 U dL−1 sample, regarded as concentrate standard.

Patient samples plasma  Fifty citrated plasma (venous blood was collected in 3.8% citrate Na, 4.5 and 0.5 mL, respectively) from hemophilia A patients, nine from non-hemophilic patients, and one from von Willebrand disease type 1 patient were tested according to methods A and B, using Pathromtin SL and imidazole buffer. Three dilutions (1/5, 1/10, and 1/20) of each plasma were performed with imidazole buffer in method A or undiluted (1/2 and 1/4) with FVIII-deficient plasma and further 1/5 with imidazole buffer in method B.


Final FVIII content was exactly identical in each corresponding dilution of both method but the total plasma content in the final mixture is decreasing from 60 μL, in the first dilution, to about 50 μL in the last dilution (17% less) in method A, even though the final volume of each mixture was the same. Conversely, the plasma content was constant in each final mixture in method B. The VWF:Ag content of Dade-Behring plasma substrate was quite good (73 U dL−1) and for this reason we always used this reagent in our assays.

Methods A and B calibration curves

The reference curves of both methods A and B showed a very good dose–response relationship, on a linear/log scale. Correlation coefficients of regression lines A and B were always very high (R2 >0.995), showing a good linearity between CTs and log of FVIII:C activities. When Pathromtin SL and Imidazole buffer were used ( Fig. 1), the slope of reference curve of method A was notably steeper than that of method B and the difference between the points of the two curves was statistically significant just from the second (1/10) dilution (paired and one-tailed Student's t-test, P < 0.001). Similar discrepancies between methods A and B were observed, even though at a lower degree, when other partial thromboplastins were used.

Figure 1.

 Semi-log plot of clotting times (CTs) vs. factor VIII (FVIII) concentrations, routinely in use for graphical calculation of FVIII assay, according to methods A (broken line) and B (continuous line), when Pathromtin SL and imidazole buffer were used. The calibration curves have been built with eight dilutions (1/5 to 1/640) of the same reference FVIII standard: squares for method B and diamonds for method A. Three dilutions (1/5, 1/10, and 1/20) of the same FVIII sample have been tested according to methods A and B. Even though the curves of each sample are parallel to the corresponding calibration curves, the FVIII assay from method A is higher than that from method B. In addition, the same CTs of 1/5 dilution can give rise to different results if read on reference curves with different slopes.

Samples from commercial FVIII standard

The results, mean of five sessions, are shown in Table 1. The FVIII concentration, assayed according to method A, was always higher than the expected values: observed/expected ratio (O/E) was always >1 (1.24–2.08) when Pathromtin SL and imidazole buffer, with or without BSA, have been used. The ratio O/E was nearer to 1 (at least for S1 and S2) when Owren's buffer replaced imidazole buffer. O/E ratios >1 (1.10–2.23) were obtained also with Actin for S1, S2, and S3, independently of the buffer used. On the contrary, Synthasil, Actin FS, but not Actin FSL, showed better results even in the presence of imidazole buffer. In each session, a very good agreement between observed and expected FVIII concentrations (ratio O/E ≈ 1) has always been achieved with method B, independently of the partial thromboplastin and buffer used.

Table 1.   Different arrangements of five activated partial thromboplastin time reagents and two buffers to assay three simulated factor (F)VIII-deficient plasma (S1 = 43.0 U dL−1, S2 = 8.6 U dL−1, and S3 = 0.86 U dL−1), derived from a plasma reference standard (86 U dL−1) diluted 1/2, 1/10, and 1/100 with FVIII-depleted plasma
FVIII-deficient plasmaPlatelet substituteBufferObserved/expected ratio
Method AMethod B
Dade-BehringPathromtin SLimidazole1.241.492.
Dade-BehringPathromtin SLOwren's1.001.071.591.001.001.25
Dade-BehringActin FSLimidazole1.071.272.600.990.951.20
Dade-BehringActin FSimidazole1.
Dade-BehringPathromtin SLimidazole + BSA1.161.361.771.030.981.02

Samples from FVIII concentrates

The results, mean of five sessions, are shown in Table 2. Both concentrate showed identical behavior: a very good agreement between methods A and B was observed for FVIII activities of 100 U dL−1 but in samples containing 10 U dL−1, the O/E ratio was 1.42 for method A and 1.0 for method B.

Table 2.   Two factor (F)VIII concentrates have been diluted to 100 and 10 U dL−1 with Dade-Behring FVIII-deficient plasma. FVIII activities have been calculated vs. calibration curve from 100 U dL−1 as standard. Data show that method B only is able to give expected FVIII activities, particularly in the sample 10 U dL−1. Reported values are means of two different sessions. Pathromtin SL and imidazole buffer were employed
Product (method)FVIII Concentrate 100 U dL−1FVIII Concentrate 10 U dL−1
FVIII assayParallel indexO/E ratioFVIII assayParallel indexO/E ratio
  1. O/E ratio, observed/expected ratio.

Beriate P (A)95.20.960.9514.20.831.42
Emoclot (A)99.70.990.9914.70.801.42
Beriate P (B)100.40.951.0010.00.951.00
Emoclot (B)98.20.950.989.90.800.99

Patient plasma

Results of FVIII assay by means of methods A and B are shown in Table 3 (mean of three sessions). The samples have been ranked in six increasing groups (<1.0; 1.0–<5.0; 5.0–<10.0; 10.0–<25.0; 25.0–50.0; >50.0 U dL−1) according to the FVIII:C level as resulted by method A. On each sample, the FVIII:C concentration from method B is always significantly lower than that obtained by means of method A, except two samples with FVIII concentration >50.0 U dL−1. The ratio between the FVIII:C assay by method A vs. method B was ranging from 1.2 to 1.5 for FVIII interval 25–50 U dL−1 to >9.5 for FVIII <1 U dL−1.

Table 3.   Factor VIII:C assay (mean ± 1 SD) in 60 patients, ordered in five different class (minimum–maximum FVIII:C level according to method A) as resulting from methods A and B. Corresponding ratios between the results of methods A and B are also displayed. Reagents: Pathromtin SL and imidazole buffer
FVIII:C range (U dL−1) (according to method A)nFVIII:C (U dL−1)Method A/B ratioStudent's paired t-test (P)
Method A, mean ± SD (U dL−1)Method B, mean ± SD (U dL−1)
<1.0220.55 ± 0.180.05 ± 0.049.5 – >10< 0.01
1.0–<5.0182.46 ± 1.121.11 ± 0.681.93–3.19< 0.01
5.0 –<10.0107.13 ± 1.594.15 ± 1.131.78–2.23< 0.01
10.0–<25.0414.83 ± 6.039.74 ± 4.411.46–1.59< 0.01
25.0–50.0439.68 ± 8.8429.68 ± 8.781.24–1.49< 0.01


Clotting times of calibration curves of FVIII standard and samples performed according to method A always resulted longer than the corresponding ones obtained by method B, especially when Pathromtin SL and imidazole buffer were used. A high significant difference between the CTs of the two methods was observed, just starting from the second dilution (1/10). The discrepancy was nil at 100 U dL−1 but about 20% at 40 U dL−1 or 40% at 10 U dL−1. Furthermore, the comparison of assays performed on samples at known FVIII content showed a very good agreement between observed and expected FVIII:C values exclusively for the assays obtained with method B.

To speculate about the cause of this discrepancy, we should take into account the relative composition of the reagent mixture in the reaction tubes. Even though the FVIII-deficient plasma is present as substrate in the same amount in both methods, the final content, that is, of the other than FVIII clotting factors, is variable along with the doubling dilutions in method A but is nearly constant in method B. The final FVIII concentration is the same in each corresponding dilution of the reference curves of both methods: they differ only for final amount of the total plasma volume, which was variable in method A but constant in all dilutions of method B. Similar observations, on the discrepancy between reference curves obtained by dilution with buffer or FVIII-deficient plasma, have been made many years ago by Frandsen et al. [21], as far as the FVIII potency of concentrates was concerned. For this reason, the author recommended the use of FVIII-deficient plasma to dilute not only the reconstituted concentrate, but also the FVIII standard to build the reference curve, even though the three dilutions of sample were made with buffer. Predilution with FVIII-deficient plasma was afterwards a very well stated procedure to obtain physiological FVIII level (100 U dL−1) from reconstituted concentrates [3,11,17].

The Frandsen's observation has been omitted when FVIII:C assay was carried out in patient plasma by traditional one-stage method, the buffer being used to dilute both the reference standard and the sample to be assayed. Really, in the range 100–50 U dL−1, the effect of dilution is poorly significant: the first dilution (1/5) of both methods A and B showed identical CTs in agreement with the same content of FVIII standard or plasma sample and FVIII-deficient plasma. The disagreement between the curves A and B becomes significant at the dilution 1/10 that is corresponding to 50 U dL−1. Afterward, the slopes of methods A, B curves and angular coefficient were completely different, the curve A being definitively more steep. When the reference curves of two partial thromboplastins are divergent, the lines obtained by fitting the points of three sample dilutions, even parallel to the corresponding reference curve, can give different results. Also, the same CT of 1/5 dilution can give different FVIII results when read on the method A or B curve, as shown in Fig. 1. Because the FVIII content is absolutely the same in each corresponding dilution of methods A and B, the different CTs should be due only to the difference in total plasma volume and its content of clotting factors other than FVIII.

Frandsen et al. showed that the discrepancy between dilutions with buffer or FVIII-deficient plasma is because of different content in fibrinogen, FV and vitamin K-dependent factors. These data raise questions about the principle on which the one-stage method is based, stating that FVIII:C is the only variable inside each reaction mixture because the plasma substrate content of clotting factors other than FVIII is optimal and constant in each dilution. According to our findings, plasma content of dilutions of routine, however, one-stage method is not constant and this inconsistency can affect the CTs of the calibration curve as well as of samples to be assayed. This phenomenon seems to be more evident when Pathrontin SL or Actin and imidazole buffer were used. Also, the addition of BSA to buffer did not decrease the discrepancy between methods A and B. Actin FS and Owren's buffer combination seems to be less affected by the differences in the final mixture. Also, the different activators present in the APTT reagents do not play a crucial role with respect to buffer composition, as showed by assays with Pathromtin SL. Probably, the phospholipid characteristics are important because it seems that synthetic phospholipids work better (Syntasil ) in method A, but we do not know how this is because of their concentration and/or qualitative composition.

Accurate standardization of reagents in FVIII assay should be recommended in order to decrease the huge inter-assay variability generally reported in multi-center exercises [13,19]. When, as reported by Barrowcliffe, 25 different APTT reagents are in use in the UK, there is no surprise if the CV of FVIII:C assay ranges from 20.7% [13] to 35% [16]. The lack of standardization may cause anomalies in FVIII:C assay. The most important is that traditional method A overestimated FVIII:C concentration in patient's plasma. The rate of overestimation rises in inverse proportion with FVIII:C level: overestimation is lacking for levels near to 100 U dL−1, but it is 20% near to 40 U dL−1 and 40%, or even higher, near to 10 U dL−1. In testing plasma of severe hemophilia A patients, we have found that method A/B ratio ranged from 1.78 to 3.19 for FVIII:C concentrations 1–10 U dL−1. The discrepancy between method A and B may play an important role in assaying the FVIII:C in postinfusion samples during pharmacokinetic studies or in checking the trough level in patients on prophylaxis. In this case, the discrepancy in plasma content between the three dilutions of patient plasma and those of FVIII standard of calibration curve may differ significantly, if traditional method A is in use. The risk of underestimating this error may be even higher if the FVIII:C assay is derived from only one dilution. Unfortunately, in some general laboratories, the assay of FVIII has performed very frequently by means of an automated procedure based on only one dilution of plasma sample. In our opinion, the findings of our study may contribute to have a FVIII assay ‘equally sensitive and accurate with therapeutic concentrates and patient samples after infusion’ [10].


Different partial thromboplastins and buffers, in heterogeneous array assembled to perform routine FVIII one-stage clotting assay, are able, individually or in combination, to give different FVIII concentration results on the same samples. In our opinion, these two reagents represent the main sources of the very well-known variability of FVIII assay in different laboratories. This phenomenon has been greatly reduced when FVIII-deficient plasma has been used not only for preparing the dilutions of reference and FVIII concentrates, but also those of plasma samples to be assayed. This strictly homogeneous handling of both samples and standards may ensure a more accurate measurement of FVIII:C activity. In addition, the effect of different chemical composition of buffers can be reduced in a matrix containing a constant clotting factors concentration. In fact, the results of method B best fitted the expected values. These findings demonstrate the significant influence of other factors, probably fibrinogen, FV, and vitamin K-dependent factors, contained in samples to be assayed as well as in reference standards, in different concentrations in the final mixtures of each dilution when traditional method A has used. Without the predilution with FVIII-deficient plasma, the FVIII level in hemophilia A patients was greatly overestimated, especially at low values. The great discrepancy between the traditional methods A and B may be very important especially when an accurate assay of low level FVIII has required. This generally occurs in defining the severity of the disease, in assessing the pharmacokinetics of FVIII concentrates and in tailoring the replacement therapy according to the trough level in recipients. In our opinion, the use of FVIII-deficient plasma in predilution and in working dilutions of both reference standard and patient plasma should be put into practice when the assay of FVIII activity is performed by one-stage method. Furthermore, the recent observation of the discrepancy between FVIII assay by one-stage method and by chromogenic substrate method [22] should probably be reviewed because of these findings.