Search tips
Search criteria 


Logo of cjvetresCVMACanadian Journal of Veterinary ResearchSee also Canadian Journal of Comparative MedicineJournal Web siteHow to Submit
Can J Vet Res. 2008 October; 72(5): 420–427.
PMCID: PMC2568046

Language: English | French

Validation of a von Willebrand factor antigen enzyme-linked immunosorbent assay and newly developed collagen-binding assay


No single test is comprehensive enough to detect all of the variants of von Willebrand Disease (VWD), making determination of both concentration and function of von Willebrand Factor (VWF) important for an accurate diagnosis. The objective of the study was to validate a newly developed VWF collagen binding assay (VWF:CB) and VWF antigen enzyme-linked immunosorbent assay (ELISA) developed at the Ontario Veterinary College (OVC VWF:Ag). Linearity, sensitivity, and coefficients of variation were determined. The Asserachrom VWF:Ag ELISA was used as the reference assay for this study. Concordance correlation and Bland-Altman plots were used to evaluate agreement between both VWF:Ag assays. The VWF:CB accuracy was assessed by degree of association with the VWF:Ag assays, and the VWF:Ag to VWF:CB ratio. All assays were assessed for their ability to distinguish between VWD negative and VWD positive patients. Linearity, intra-assay coefficients of variation, and inter-assay coefficients of variation were acceptable for both the newly developed VWF:CB (R2 = 0.97, average CV = 4.4, and 15, respectively) and OVC VWF:Ag assays (R2 = 0.96, average CV = 7.9, and 5.9, respectively). Agreement between the OVC VWF:Ag assay and reference assay was excellent (ρc = 0.89), and although differences between assay results precluded interchangeable use of the assays, both successfully distinguished VWD positive and VWD negative dogs (P < 0.0001). The VWF:CB showed a strong association with both VWF:Ag assays (R2 = 0.86, 0.82) and VWF:Ag to VWF:CB ratios (≤ 1) were as expected. The excellent performance of both assays in this validation study confirm their reliability and potential for clinical application.


Aucun test unique n’est assez complet pour permettre de détecter tous les variants de la maladie de von Willebrand (VWD), rendant ainsi la détermination de la concentration et de la fonction du facteur de von Willebrand (VWF) importante pour un diagnostic précis. L’objectif de l’étude était de valider une épreuve récemment développée d’attachement du collagène du VWF (VWF-CB) et une épreuve ELISA développée au Ontario Veterinary College pour l’antigène du VWF (OVC VWF-Ag). La linéarité, la sensibilité et les coefficients de variation ont été déterminés. Pour la présente étude l’épreuve de référence était l’ELISA VWF-Ag d’Asserachrom. La concordance de corrélation et les diagrammes de Bland-Altman ont été utilisés afin d’évaluer l’accord entre les deux épreuves VWF:Ag. La précision du VWF-CB était évaluée par le degré d’association avec les épreuves VWF-Ag, et le rapport VWF-Ag/VWF-CB. Toutes les épreuves ont été évaluées pour leur capacité à faire la distinction entre les patients VWD négatifs et VWD positifs. La linéarité, les coefficients de variation intra-épreuves et inter-épreuves étaient acceptables pour l’épreuve VWF-CB nouvellement développé (R2 = 0,97, CV moyen de 4,4 et 15, respectivement) et pour l’épreuve OVC VWF-Ag (R2 = 0,96, CV moyen de 7,9 et 5,9, respectivement). L’accord entre l’épreuve OVC VWF-Ag et l’épreuve de référence était excellent (ρc = 0,89) et bien que des différences entre les résultats des épreuves empêchent une utilisation interchangeable des épreuves, les deux permettent de distinguer avec succès les chiens VWD positifs et VWD négatifs (P < 0,0001). L’épreuve VWF-CB montrait une forte association avec les deux épreuves VWF-AF (R2 = 0,86, 0,82) et des rapports VWF-Ag/VWF-CB (≤ 1) étaient tels qu’attendus. L’excellent rendement des deux épreuves dans cette étude de validation confirme leur fiabilité et potentiel pour une application clinique.

(Traduit par Docteur Serge Messier)


Von Willebrand disease (VWD) is the most common inherited bleeding disorder in dogs (1). Three types of canine VWD have been identified: 2 are quantitative deficiencies (types 1 and 3), with a decrease in circulating factor, and 1 is qualitative (type 2) where there is a decrease in high molecular weight multimers of von Willebrand factor (VWF) (2). It can be challenging to both recognize and diagnose VWD due to the varied presentations of the 3 forms, and the incomplete penetrance of the genetic defect for the most common form (type 1). Routine coagulation tests (prothrombin time and partial thromboplastin time) are generally within reference intervals and clinical evidence of bleeding is quite variable. The current gold standard for confirming the disease is the von Willebrand factor antigen concentration (VWF:Ag) test. The coefficient of variation for this assay has been reported to be as low as 3.8% (3). The assay determines concentration (% normal concentration) of the factor as opposed to function. Reported ranges include VWD negative (> 70%), VWD positive (< 50%) and indeterminate values that comprise the difference (2). This test can identify some of the type 1 and 3 VWD positive animals, but can leave some diagnoses indeterminate. Type 1 disease can also be problematic to diagnose because of extragenic influences, including azotemia, liver disease, strenuous exercise, endotoxemia, parturition, and increased plasma vasopressin, that can temporarily increase VWF into the normal range (2,4). In addition, this test is not very effective for identifying type 2 disease as this is a qualitative rather than a quantitative deficiency.

Functional assays, such as ristocetin (VWF: RCof) or botrocetin-induced platelet aggregation, platelet function analyzer (PFA-100), and collagen binding assays (VWF:CB) are available. Ristocetin aggregation is inhibited by canine plasma regardless of the amount of functional VWF available in the plasma (2). Using high levels of ristocetin overcomes this issue, but the resulting aggregate is too fragile to withstand stirring in a platelet aggregometer (2). The botrocetin-induced aggregation assay has better correlation with VWF:Ag but is imprecise and difficult to standardize (2,4). The PFA-100 (Dade Behring, Marburg, Germany) is an automated analyzer, developed in 1995, that simulates vascular injury in order to assess platelet function (5). For the dog, when using adenosine diphosphate (ADP) cartridges to initiate platelet aggregation, the PFA-100 is specific in its detection of primary hemostatic disorders and can be used as a screening tool for cases of VWD with severely decreased VWF concentrations (6). In cases with a moderate decrease in VWF the closure time often remains within reference limits (6). In 1986, the use of a VWF:CB, which measures the quantity of VWF bound to immobilized collagen in a procedure similar to an enzyme-linked immunosorbent assay (ELISA), was proposed as an alternate to the VWF: RCof assay in human diagnostic laboratories (7). The coefficient of variation for the VWF:CB has been reported to be as low as 4.4% (3). This assay is superior to VWF:RCof in its detection of type 2 VWD, and has a decreased assay variability (interassay and interlaboratory) (8). Collagen has been shown to bind VWF with a preference for the high molecular weight forms; therefore, the VWF:CB can be used to assess the relative proportion of large VWF multimers (9,10). In patients with VWD, the collagen binding activity is significantly decreased compared with control patients (7). In types 1 and 3 VWD, the decrease in function of VWF parallels the level of VWF:Ag present, and in type 2 VWD there is a disproportionate decrease in activity (1). Collagen-binding activity, in conjunction with the antigen assay, can therefore be used to assess the quantity of VWF as well as its function (7). This makes these assays valuable not only as screening tests for VWD, but also to distinguish between types 1 and 2 VWD (7). The ratio of VWF:Ag to VWF:CB in VWD negative patients and type 1 VWD is generally ≤ 1 (11). In type 2 VWD this ratio has been reported to range from 2 to 8 (10,11).

As there is no single, comprehensive test to detect all of the variants of VWD, it is important to determine both the concentration and function of VWF in the diagnosis of VWD (7,12). Although the VWF:CB is extensively used in diagnostic evaluation of human VWD, it is not widely available for the diagnosis of canine VWD. As diagnostic testing in humans has demonstrated, a panel of tests is typically required for the accurate and complete diagnosis of VWD, and the VWF:CB has proven to be a valuable asset (13). The objective of this study was to validate a canine VWF:CB and the VWF antigen ELISA developed at the Ontario Veterinary College (OVC VWF:Ag), and create reference intervals for both.

Materials and methods

All samples were collected in accordance with the guidelines for the care and use of experimental animals established by the Canadian Council on Animal Care under protocols approved by the University of Guelph Animal Care Committee (14).

Normal plasma pool

Blood from 7 Ontario Veterinary College — Veterinary Teaching Hospital (OVC-VTH) blood donor dogs was collected into citrate tubes (9 parts blood:1 part citrate) through nontraumatic jugular venipuncture. The 3 female and 4 male dogs that comprised this pool ranged in age from 2 to 5 y, and included greyhounds and mixed-breed dogs. Within 1 h of collection, the citrated blood was centrifuged at 1500 × g for 20 min at room temperature; plasma was harvested, pooled, and then stored in 100 and 200 μL aliquots at −80°C until needed.

Patient samples

A total of 183 plasma samples ranging in VWF concentration from 1% to 125% were collected from dogs at veterinary clinics located near Guelph and the OVC-VTH. Eighty-two plasma samples had VWF concentrations < 50%, 55 had concentrations between 50% to 70%, and 46 samples had concentrations > 70%. The VWF concentrations were determined by the Hemostasis Reference Laboratory (Henderson Research Centre, Hamilton, Ontario) using the Asserachrom VWF:Ag ELISA assay (Diagnostica Stago, Asnières sur Seine, France). Of the 183 patient citrated plasma samples, 176 were collected through nontraumatic jugular venipuncture. The remaining 7 samples were collected through nontraumatic cephalic venipuncture due to either health issues precluding jugular sampling, such as cervical spinal disease — IVDD, or cervical vertebral instability, or restraint limitations. The sample population included 100 females and 83 males with an average age of 4.8 y (range: 0.25 to 13 y) from the following breeds: Doberman pinscher (n = 125), Manchester terrier (n = 13), mixed breed (n = 10), greyhound (n= 8), Irish setter (n = 6), Belgian tervuren (n = 6), golden retriever (n= 3), Bernese mountain dog (n = 2), english setter (n = 2), Labrador retriever (n = 1), standard schnauzer (n = 1), Shetland sheepdog (n = 1), flat-coated retriever (n = 1), Staffordshire terrier (n = 1), plott hound (n = 1), and great dane (n = 1). Since VWF:Ag concentrations generally approach mean adult values by 11 wk of age, patients <12 wk of age were excluded from the study to avoid any influence of age on sample concentrations. Samples were not collected from overtly ill patients; however, several chronic subclinical disorders were present in the population, for example dilated cardiomyopathy, hypothyroidism, cervical vertebral instability, urinary incontinence, intervertebral disc disease, food or skin allergies, and arthritis. Thus, medications present in the sampled population included diethylstilbestrol (n = 3), phenylpropanolamine (n = 4), levothyroxine (n = 13), cardiac medications (n = 2), nonsteroidal anti-inflammatories (n = 7), low dose corticosteroids (n = 1), antibiotics (n = 4), and a variety of nonmedicinal supplements (n = 17). Rare samples were collected during estrus as it has been shown that there is no significant change in VWF concentration during the estrus cycle (16). The citrated blood was centrifuged (1500 × g for 20 min at room temperature) within 3 h of collection, plasma was harvested, pooled, and then stored in aliquots ranging from 10 μL to 800 μL at −80°C until samples could be batch-processed and remaining assays performed (minimum 1 d, maximum 3.5 mo).

Assay methodology

All assays were performed by a single operator (HB). Except for plate preparation, both the VWF:CB and VWF:Ag assays were performed using similar methodologies, and they are described together. Costar 96-well flat bottom plates (Corning, Corning, New York, USA) were coated with a mixture of room temperature 0.05 M carbonate buffer (pH 9.6) and either Vitrogen Bovine Collagen (COHESION, Palo Alto, California, USA), containing 95% Type I and 5% Type III collagen in a 1:30 dilution, or Sheep Anti-Canine von Willebrand Factor IgG (Cedarlane Laboratories, Hornby, Ontario) in a 1:1000 dilution, for the VWF:CB and VWF:Ag assays, respectively. The plates were then incubated for 2 d at room temperature for the VWF:CB assay and for 16 to 18 h at 4°C for the VWF:Ag assay. After incubation, the plates were washed 3 times with a PBS/Tween/1% skim milk washing buffer. A standard curve, composed of 7 serial dilutions of a normal canine plasma pool representing relative VWF concentrations of 0%, 6.25%, 12.5%, 25%, 50%, 75%, and 100%, was applied to the plates along with patient samples at a 50% dilution. Normal canine plasma pool and patient samples were diluted using the PBS/Tween/1% skim milk washing buffer. After a 90-min incubation period at room temperature, the washing step was repeated and a PBS/Tween/5% skim milk blocking solution was applied for a 60-min incubation period. Again, the washing step was repeated and a solution of Sheep Anti-Canine von Willebrand Factor IgG — Peroxidase Conjugate (Cedarlane Laboratories) with PBS/Tween/1% skim milk washing buffer at a 1:1000 dilution was added and incubated for 90 min at room temperature. The plates were then washed 4 times with the PBS/Tween/1% skim milk washing buffer. After a 5 to 8 min incubation period with the peroxidase substrate, 3, 5, 3′, 5′ tetramethylbenzidine dihydrochloride (SIGMA, Sigma-Aldrich, St. Louis, Missouri, USA), a 2M solution of sulfuric acid was added to stop the reaction. The resultant color change was assessed with an EL-800 Biotek plate reader (Biotek Instruments, Winooski, Vermont, USA) at a 450 nm wavelength. Optical density was plotted against % dilution to form a standard curve, and the slope and intercept were determined by using computer software (Microsoft Excel 2002; Microsoft, Mississauga, Ontario). Patient samples with very low (< 12.5%) VWF:Ag concentrations and collagen-binding activity were repeated as 200% solutions (as opposed to the typical 50%), thereby increasing the sample’s absorbance into the optimal range for the assay.


Serial dilutions of the normal canine plasma pool were used to assess the linearity and sensitivity of the assay. The plasma was initially diluted to 1:166 with PBS/Tween/1% skim milk washing buffer, then serially diluted to achieve the following concentrations: 75%, 50%, 25%, 12.5%, 6.25%, and 0%. The initial dilution of 1:166 was chosen to maximize the number of dilution points within the optimal absorbance range for the assay. The results were plotted and the R2 value calculated.

Precision studies

Samples for intra-assay and inter-assay precision were collected as described for the patient samples and normal canine plasma pool, respectively. For VWF:CB and VWF:Ag intra-assay precision, 3 patient samples with low, medium, and high VWF:Ag concentrations (20%, 47%, and 85%, respectively) were repeated 14 to 18 times on the same 96-well plate, and coefficients of variation (CVs) were calculated using the optical density results. For VWF:CB and VWF:Ag inter-assay precision, the normal canine plasma pool was initially diluted to 1:166 with PBS/Tween/1% skim milk washing buffer (representing a relative concentration of 100%), then serially diluted to achieve concentrations of 50% and 25%. All 3 solutions were assayed, in duplicate, on 11 different 96-well plates over a 10-month period. The coefficients of variation, were calculated using the assay results; these are reported as percent of normal canine plasma pool.

Method comparison

The Asserachrom VWF:Ag ELISA assay used by the Hemostasis Reference Laboratory, was used as the reference assay for this study. Each of the 183 patient samples were analyzed using this assay, as well as the OVC VWF:Ag assay. The data were treated with a square root transformation in order to achieve normality. Agreement between the 2 assays was assessed by calculating the concordance correlation (SAS; SAS Institute, SAS Campus Drive, Cary, North Carolina, USA). In addition to the concordance correlation, Bland-Altman plots (difference plotted against the mean for each sample) were constructed using 95% confidence intervals (CI) to assess within-animal variation, or measurement error (Sigma Plot 8.0 for Windows, Systat Software, Richmond, California, USA) (17).

At this time, there is no appropriate reference test for VWF:CB method comparison; therefore, accuracy was assessed through the level of association with the Asserachrom VWF:Ag ELISA as well as the OVC VWF:Ag ELISA results (SAS Institute). A square root transformation was used to achieve normality of these data. The ratio of VWF:Ag to VWF:CB was also examined (Minitab for Windows, Student Release 12, Minitab, State College, Pennsylvania, USA). Data were assessed for outliers using the Grubb’s test (18). As an additional measure of accuracy and screening efficacy, the ability of each assay to distinguish between groups of non-Doberman pinscher dogs without a bleeding history (n = 35), Doberman pinscher dogs without a bleeding history and a history of a previous negative VWF:Ag ELISA (determined by another laboratory) or DNA test (n= 21), and Doberman pinschers with a history of a previous positive VWF:Ag ELISA (determined by another laboratory) or DNA test (n = 45) was assessed. The Tukey-Kramer test was used to detect significant differences between groups, due to its ability to adjust for multiple comparisons (SAS Institute).

Reference interval determination

During patient sampling, 56 dogs with a signalment of being healthy, non-Doberman pinschers without a bleeding history, or a healthy Doberman pinscher without a bleeding history and a history of a previous negative antigen ELISA (determined by another laboratory) or DNA VWD test, were selected from the previously described patient population to formulate the reference intervals. These samples were processed and stored as previously described for patient samples. Forty-five Doberman pinschers with a history of a previous positive antigen ELISA or DNA VWD test were selected to create reference intervals for VWD positive dogs, enabling the degree of overlap between positive and negative status to be evaluated. Again, these samples were processed and stored as previously described for patient samples. Grubb’s test was performed on all reference interval data to detect outliers (18) and normality of the data was assessed using the Anderson-Darling Normality test (Minitab for Windows). Nonparametric data were ranked, and the 2.5 and 97.5 percentiles were determined, defining the central 95% of the data (Minitab for Windows) (19). The mean and standard deviation (s) of the parametric data was determined, and the central 95% of the data was defined by calculating the mean (+/− 2 s) (Minitab for Windows).


Linearity results are reported graphically for the OVC VWF:Ag assay (Figure 1), and results for the collagen binding assay, although not reported graphically, were very similar in appearance. The R2values were 0.96 and 0.97, respectively. The relative concentration of 100% was removed from the VWF:Ag assay due to consistent loss of linearity at the higher optical density values. When sample results were calculated using the slope equation (Table I), it became clear that the assays were not accurate at the 6.25% dilution; therefore, the sensitivity of the assay was considered to be the 12.5% dilution, or optical densities of approximately 0.23 and 0.13, for the VWF:Ag assay and VWF:CB, respectively.

Figure 1
Linear regression scatter plot of optical density (OD) versus canine normal plasma pool sample dilution for the OVC VWF:Ag assay. R2 = 0.96.
Table I
Sample results for the canine normal pool, calculated using the slope equation for the OVC VWF:Ag and VWF:CB assays, with respect to their initial dilutions

Intra-assay CVs for the low, medium, and high VWF concentration samples were 3.7%, 7%, and 2.5%, respectively for the VWF:CB, and 8.6%, 11%, and 4.2%, respectively for the OVC VWF:Ag assay (Table II). Inter-assay CVs for the low, medium and high VWF concentration samples were 18%, 12%, and 14%, respectively for the VWF:CB, and 4.7%, 3.8%, and 9.2%, respectively for the VWF:Ag assay (Table II).

Table II
Intra-assay and inter-assay precision for the VWF:CB and OVC VWF:Ag assays

The concordance correlation, examining the agreement between the reference assay (Asserachrom VWF:Ag ELISA assay) and the OVC VWF:Ag assay, can be interpreted by referring to statistical strength of agreement as reported by Shoukri and Pause (20). Although these ranges were originally designed for the interpretation of kappa, the analogy between these statistical methods makes the strength of agreement indices appropriate for the concordance correlation. In this case there was almost perfect agreement, defined as 0.81–1.00, with a concordance correlation of 0.89 (20). The corresponding linear regression plot is shown in Figure 2. A Bland-Altman plot revealed the relationship between the differences and means of the VWF:Ag data, and showed that the scatter of the differences widened as the mean VWF:Ag concentration increased (Figure 3). The mean (+/− 2 s) was depicted for evaluation of clinical relevancy of the difference in results.

Figure 2
Linear regression scatter plot of VWF:Ag concentration, as determined by the OVC assay versus Asserachrom VWF:Ag concentration for 183 samples after square root transformation. Concordance correlation = 0.89.
Figure 3
Bland-Altman Plot of difference in VWF:Ag concentration for 183 samples, as determined by the OVC VWF:Ag assay and the Asserachrom VWF:Ag assay versus mean of the VWF:Ag concentration, as determined by the 2 assays. Data are not transformed. (s = standard ...

The coefficient of determination (R2) showed excellent correlation between the VWF:CB and Asserachrom VWF:Ag ELISA assays as well as between the VWF:CB and OVC VWF:Ag assays with values of 0.86 and 0.82, respectively, and represented the proportion of variation in collagen-binding activity that could be attributed to VWF concentration (21). Corresponding linear regression plots are shown in Figures 4 and and5,5, respectively. As there were no type 2 VWD patients within the test population, the ratio of VWF:Ag to VWF:CB was as expected, with a median value of 0.74 (range: 0.1–1.84). A single outlier was removed from these data (ratio of 4.23) based on a Grubb’s test Z-value of 9.12, while the critical value of Z for these data was 3.58. This particular data point was verified in the raw data, and was considered to be a true outlier, rather than an indication of type 2 VWD, as type 2 VWD has not been identified in that particular breed (Doberman pinscher).

Figure 4
Linear regression scatter plot of VWF:CB results versus Asserachrom VWF:Ag concentration for 183 samples, after square root transformation. R2 = 0.86.
Figure 5
Linear regression scatter plot of VWF:CB results versus VWF:Ag concentration for 183 samples as determined by the OVC assay, after square root transformation. R2 = 0.82.

Results from the Asserachrom VWF:Ag ELISA assay, the OVC VWF:Ag assay, and the collagen binding assay were compared between groups of dogs composed of Doberman pinschers with VWD negative history (group 1), non-Doberman dogs expected to be VWD negative (group 2), and Doberman pinschers with a VWD-positive history (group 3), as previously described. Using the Tukey-Kramer test, there was a significant difference detected between groups 1 and 3, as well as between groups 2 and 3, while there was no significant difference between groups 1 and 2 (P values: < 0.0001, < 0.0001, 0.7199, respectively) using results from the Asserachrom VWF:Ag ELISA assay. The same conclusions were found for the OVC VWF:Ag test and the collagen binding assay (VWF:Ag, P values: < 0.0001, < 0.0001, 0.1443, respectively; VWF:CB, P values: < 0.0001, < 0.0001, 1.0097, respectively).

Reference intervals for the Asserachrom VWF:Ag ELISA assay were previously determined at the Hemostasis Reference Laboratory and were reported as: 56% to 143%. Reference intervals, rounded to the nearest integer, for the OVC VWF:Ag assay and collagen binding assay are reported in Table III. Using the Grubb’s test, no outliers were detected in the data used to determine the reference intervals.

Table III
Established reference intervalsa for the OVC VWF:Ag and VWF:CB assays


The overall performance of the newly developed VWF:CB and OVC VWF:Ag assays was excellent, and the linearity study revealed an optimal range of optical densities (0.23–0.98). Use of the plasma pool relative concentration of 100% in the VWF:Ag assay for the calculation of R2 and slope was often excluded; the dilutions were not adjusted to include this concentration in the optimal absorbance range, in order to take advantage of this range for lower values where accuracy was more clinically imperative. The plasma pool relative concentration of 100% was typically included in the VWF:CB assay due to generally lower absorbances [optical density (OD)] in this assay. The linearity study also revealed poor accuracy of the VWF:Ag and VWF:CB assays in the lower optical density range (< 0.23 and 0.13, respectively). The sensitivity of the assays was thus determined to be limited to the 12.5% dilution corresponding to optical densities of 0.23 and 0.13 as mentioned previously.

The intra-assay CVs were excellent over a wide range of VWF:Ag concentrations with an average of 7.87% for the OVC VWF:Ag assay and 4.4% for the VWF:CB assay. Although previous reports have cited CVs as low as 3.8% for the VWF:Ag assay, the VWF:Ag concentration used for that determination was not stated (3). Using the assessment of repeatability from the “high” VWF:Ag concentration (85%) our assay performed with a CV of 4.2%, which is comparable to that found in the previous study (3). In addition, the average CV for the OVC VWF:Ag assay is comparable to a CV of 6.1% reported for the Asserachrom VWF:Ag ELISA assay by the Hemostasis Reference Laboratory. In previous studies, the CV for the VWF:CB has been reported as low as 4.4%, which was consistent with our findings (using the average CV) (3). Both assays performed well during the evaluation of inter-assay repeatability. Previously reported inter-assay CVs for the VWF:CB have ranged from 1.2% to 11.0% depending on the concentration of VWF present (22). In our study, the CVs for all VWF:CB concentrations were slightly higher that those of Sabino et al (22), possibly because of the different time periods used in each study. Sabino et al (22) used data from assays performed on 14 separate days, whereas the data used to calculate the inter-assay CVs herein were collected over a 10-month period. This approach was chosen because it coincided with sample evaluation, and more accurately reflected the degree of variation throughout our study. However, there are increased numbers of possible confounding factors encountered over such a prolonged period, such as numerous batches of reagents, changes in lots of antibodies, and variable environmental conditions, that may have further contributed to the degree of variation. The inter-assay CV for the OVC VWF:Ag assay (range: 3.8–9.2%) was similar to the 3.0–7.2% range reported for the Asserachrom VWF:Ag ELISA assay.

Agreement of the OVC VWF:Ag assay with the Asserachrom VWF:Ag ELISA assay was almost perfect when assessed using concordance correlation. This statistical test is superior to other types of analyses, such as the paired t-test and Pearson’s correlation, which only show association between data, not necessarily agreement (20). The concordance correlation is essentially analogous to Cohen’s weighted kappa and is used as a measure of agreement between 2 clinicians when the response is ordinal (20). The Bland-Altman plot is used for assessment of within-animal variation, as well as how interchangeable the assays are in a clinical context. Provided differences, within the mean difference (+/− 2 s) of the difference, are not clinically important, the 2 measurement methods can be used interchangeably (17). Our plot (Figure 3) showed systematic variation in the differences, that is, increased difference with increased mean. While this pattern often suggests that a logarithmic transformation is necessary, these data responded best visually to a square root transformation. Unfortunately, a log transformation is the only transformation giving back transformed differences that are easy to interpret, and use of any other transformation is not recommended in this context (17). Thus, nontransformed data showing systematic variation results in limits of agreement that are wider than necessary for small mean VWF:Ag concentrations, and narrower than they should be for larger mean VWF:Ag concentrations (17). Additional markers, therefore, were inserted to represent the mean (+/− 1 s) to assist in the visual assessment of clinical relevance of the difference between assays. For the smaller mean VWF:Ag concentrations, most differences were represented by a tight cluster of data within 1 standard deviation of the mean. Although the data were more loosely clustered as the mean VWF:Ag concentration increased, up to a mean concentration of approximately 80%, from 80% to 85% of the data points were within the mean (+/− 1 s). A difference of 40% between assays is clinically unacceptable; however, this degree of disagreement was only seen consistently in the higher VWF:Ag concentration ranges, where patients are considered VWD negative; the influence on clinical interpretation in this range is minimal. In contrast, the least amount of disagreement, as mentioned, was in the lower mean VWF:Ag concentrations where a high level of agreement is very important. Overall, this graphical assessment of the 2 assays suggested that while they showed excellent agreement through the concordance correlation, strictly speaking, these 2 assays are not interchangeable. This is somewhat expected as test results from different laboratories are generally not directly comparable due to differences in plasma standards, assay techniques, and reference intervals (23). However, the level of agreement in the range of assay results having the most effect on patient management was likely sufficient to justify using the OVC VWF:Ag assay in place of the Asserachrom VWF:Ag ELISA assay in a clinical context.

Diagnostic testing in humans has found that the decrease in function of VWF parallels the level of VWF:Ag concentration in type 1 and 3 disease, suggesting there should be a high degree of association between these tests (1). Due to the current lack of a reference assay for comparison with the newly developed VWF:CB assay, the relationship between the concentration and function of VWF was used to assess performance of the VWF:CB assay. The VWF:CB showed a strong association with both the Asserachrom VWF:Ag ELISA assay and the OVC VWF:Ag assay (R2 = 0.86 and 0.82, respectively). Similarly, assessment of the VWF:Ag to VWF:CB ratio in this study yielded expected results for a type 1 VWD population, as the ratio for these patients should generally be ≤ 1 (11). Based on the known relationship between concentration and function of VWF in type 1 VWD, the VWF:CB performance is excellent; however, further studies are required to confirm that the VWF:CB is capable of detecting the disproportionate decrease in VWF function associated with type 2 disease.

In addition to showing excellent agreement, or strong associations with each other, each assay used in this study was able to successfully differentiate groups of VWD negative and VWD positive patients. This finding reinforces the use of these assays in a clinical context for the identification of VWD.

Reference intervals were determined for each of the newly developed assays. The indeterminate values reflect the overlap of results from VWD positive and negative patients and reinforces the variable presentation of this disease. Although the number of Doberman pinschers used in the reference interval determination (VWD negative = 21, VWD positive = 45) was high and could be considered a bias, they were not excluded from the population. The well-known affiliation between this breed and particular disease suggested that their inclusion more accurately reflected the target population. For the VWF:Ag assay reference intervals, the lower cutoff for the VWD negative population was comparable to literature values. However, the reference interval determined for our indeterminate group was wider, making a lower result necessary to confirm disease. The value of a single sample in an animal with an indeterminate VWF:Ag ELISA result is limited by daily and weekly variation in VWF:Ag concentration, and patients falling within this area of the reference interval should be retested to reveal any trends that may be present (24). For the VWF:CB reference intervals, the lower cutoff for the VWD negative population was higher than that seen for the VWF:Ag assay, and was reflective of the tendency for VWF:Ag to VWF:CB ratios of < 1 in our study. While good screening efficacy of the VWF:CB has been determined in human laboratories, use of the reported reference intervals for screening purposes should be approached with caution (7). As mentioned, the tendency of our study population was to have a VWF:Ag to VWF:CB ratio < 1; therefore, the VWF:CB result was typically higher than the corresponding VWF:Ag concentration. However, a VWF:Ag to VWF:CB ratio up to one is quite acceptable in the literature, and with the reported reference ranges a patient determined to be VWD negative based on a VWF:Ag concentration of 80%, and a VWF:Ag to VWF:CB ratio of 1.0 could be inappropriately placed in the indeterminate group if the VWF:CB assay was used alone. As such, based on the study results herein, the VWF:CB appeared to be most valuable when used in conjunction with the VWF:Ag assay.

In conclusion, the overall performance of the newly developed VWF:CB and VWF:Ag assays was excellent. Although, strictly speaking, the OVC VWF:Ag assay cannot be freely interchanged with the Asserachrom VWF:Ag ELISA assay, it is capable of distinguishing VWD positive dogs from VWD negative dogs, and with knowledge of its limitations on either side of the optimal absorbance range, its otherwise excellent performance shows promise for clinical application.


Financial support for this study was provided by the Ontario Veterinary College Pet Trust Fund and the American Kennel Club Canine Health Foundation.


This full article has been published as a thesis chapter in partial fulfillment of requirements for the degree of Doctor of Veterinary Science, at the Ontario Veterinary College, University of Guelph.


1. Denis CV, Wagner DD. Insights from von Willebrand disease animal models. Cell Mol Life Sci. 1999;56:977–990. [PubMed]
2. Thomas JS. Von Willebrand’s disease in the dog and cat. Vet Clin North Am Small Anim Pract. 1996;26:1089–1111. [PubMed]
3. Callan MB, Giger U, Catalfamo JL. Effect of desmopressin on von Willebrand factor multimers in Doberman pinschers with type 1 von Willebrand disease. Am J Vet Res. 2005;66:861–867. [PubMed]
4. Ewenstein B. Von Willebrand’s disease. Annu Rev Med. 1997;48:525–542. [PubMed]
5. Franchini M. The platelet-function analyzer (PFA-100) for evaluating primary hemostasis. Hematology. 2005;10:177–181. [PubMed]
6. Mischke R, Keidel A. Influence of platelet count, acetylsalicylic acid, von Willebrand’s disease, coagulopathies and haematocrit on results obtained using a platelet function analyser in dogs. Vet J. 2003;165:43–52. [PubMed]
7. Paczuski R. Determination of von Willebrand factor activity with collagen-binding assay and diagnosis of von Willebrand disease: Effect of collagen source and coating conditions. J Lab Clin Med. 2002;140:250–254. [PubMed]
8. Favaloro EJ. Detection of von Willebrand disorder and identification of qualitative von Willebrand factor defects: Direct comparison of commercial ELISA-based von Willebrand factor activity options. Am J Clin Pathol. 2000;114:608–618. [PubMed]
9. Duggan MJ, DiMichele DM, Christian MJ, Fink LM, Hathaway WE. Collagen-binding of von Willebrand’s factor antigen in the classification of von Willebrand’s Disease. Am J Clin Pathol. 1987;88:97–102. [PubMed]
10. Brown JE, Bosak JO. An ELISA test for the binding of von Willebrand antigen to collagen. Thromb Res. 1986;43:303–311. [PubMed]
11. Favaloro EJ, Grispo L, Koutts J. Development of a simple collagen based ELISA assay aids in the diagnosis of, and permits sensitive discrimination between type I and type II, von Willebrand’s disease. Blood Coagul Fibrinolysis. 1991;2:285–291. [PubMed]
12. Favaloro EJ, Facey D, Henniker A. Use of a novel platelet function analyzer (PFA-100) with a high sensitivity to disturbances in von Willebrand factor to screen for von Willebrand’s disease and other disorders. Am J Hematol. 1999;62:165–174. [PubMed]
13. Favaloro EJ, Bonar R, Kershaw G, et al. Laboratory diagnosis of von Willebrand’s disorder: Quality and diagnostic improvements driven by peer review in a multilaboratory test process. Haemophilia. 2004;10:232–242. [PubMed]
14. Olfert ED, Cross BM, McWilliam AA, editors. Guide to the Care and Use of Experimental Animals. 2. Ottawa: Bradda Printing Services Inc; 1993. Canadian Council on Animal Care; pp. 95–97.
15. Mansell PD, Parry BW. Changes in factor VIII activity and von Willebrand factor antigen concentration with age in dogs. Br Vet J. 1992;148:329–337. [PubMed]
16. Moser J, Meyers KM, Russon RH, Reeves JJ. Plasma von Willebrand factor changes during various reproductive cycle stages in mixed-breed dogs with normal von Willebrand factor and in Doberman Pinschers with type-I von Willebrand’s disease. Am J Vet Res. 1998;59:111–118. [PubMed]
17. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1:307–310. [PubMed]
18. Graphpad Software, QuickCalcs. [Last accessed June 30, 2008]. [homepage on the Internet]. Available from:
19. Solberg HE. The theory of reference values. Part 5. Statistical treatment of collected reference values. Determination of reference limits. J Clin Chem Clin Biochem. 1983;21:749–760. [PubMed]
20. Shoukri MM, Pause CA. Statistical Methods for Health Sciences. 2. Boca Raton, Florida: CRC Pr; 1999. pp. 35–41.
21. Hopkins WG. A New View of Statistics. [Last accessed February 2007]. Web site
22. Sabino EP, Erb HN, Catalfamo JL. Development of a collagen-binding activity assay as a screening test for type II von Willebrand disease in dogs. Am J Vet Res. 2006;6:242–429. [PubMed]
23. Brooks M. Management of canine von Willebrand’s disease. Probl Vet Med. 1992;4:636–646. [PubMed]
24. Kramer JW, Venta PJ, Klein SR, Cao Y, Schall WD, Yuzbasiyan-Gurkan V. A von Willebrand’s factor genomic nucleotide variant and polymerase chain reaction diagnostic test associated with inheritable type-2 von Willebrand’s disease in a line of German shorthaired pointer dogs. Vet Pathol. 2004;41:221–228. [PubMed]

Articles from Canadian Journal of Veterinary Research are provided here courtesy of Canadian Veterinary Medical Association