For decades, hematologists have debated how to diagnose von Willebrand disease (VWD) type 1, which is characterized by partial quantitative deficiency of von Willebrand factor (VWF). Little progress has been made, mainly because the population distribution of VWF levels is very broad, low values seldom have a simple genetic basis, and many bleeding symptoms are very common. As a result, a relationship between low VWF and bleeding can be difficult to establish, and the boundary between ‘healthy’ and ‘diseased’ seems irreducibly arbitrary. A letter in this issue of the Journal clearly illustrates the problem − depending on the criteria, from 1 to 27 clinic patients in the French Basque country could be diagnosed with VWD type 1 . The question remains, can any criteria identify patients for whom this diagnosis is truly useful, or is a different approach required?
The mean plasma level of VWF is about 100 IU dL-1, by definition, although the precise value depends strongly on the ABO blood types of a population. In general, the mean for blood type O is 25 U dL-1 to 35 IU dL-1 lower than for other ABO types . More importantly, the range of values is very large. For populations of heterogeneous blood type, 95% of VWF values typically lie between 50 IU dL-1 and 200 IU dL-1. For blood type O the mean is 75 IU dL-1 with a range (± 2 SD) of 36–157 IU dL-1 and for type AB the mean is 123 IU dL-1 with a range (± 2 SD) of 64–238 IU dL-1 − the variation illustrates the effect of ABO type. Changes in the prevalence of blood type O will change the population distribution of VWF levels and alter the number at risk for a diagnosis of VWD. For example, data for healthy blood donors indicate that VWF levels < 50 IU dL-1 are expected in 14% of type O subjects, 2.9% of type A subjects, 0.9% of type B subjects, and only 0.3% of type AB subjects . There is no epidemic of bleeding among type O blood donors, and a VWF level at the low end of the type O distribution, between 36 IU dL-1 and 50 IU dL-1, is not very dangerous by itself.
The broad distribution of VWF levels is closely related to their low heritability. Family studies suggest that just 30% of the variance in VWF is heritable [3,4]. Twin studies report higher values of up to 75%[5,6], but these estimates may be increased by hidden environmental influences . If at least one-third of the variance in VWF level is attributable to genetic factors, then what genes are responsible? When VWF levels are very low, <20 IU dL-1, mutations in the VWF gene usually can be found [7,8]. But genome wide linkage analysis in 342 less selected individuals did not pick up an effect of the VWF gene; instead, the ABO blood group was the only identifiable influence , and it accounts for a minority of the variation in VWF level. Another modifier of oligosaccharide antigens, the Secretor locus, also has a small effect . More focused studies have shown a relationship of VWF level to the VWF gene, suggesting it accounts for 20% of the observed variance . In short, ABO type is a significant determinant of VWF level, the Secretor and VWF genes also have a minor influence, but up to two-thirds of the variance in VWF level cannot be attributed to genes. In this respect, VWF level is similar to many cardiovascular risk factors such as blood pressure or cholesterol level: rare genetic disorders are associated with hypertension or hypercholesterolemia, but most of the variation in the population cannot be explained simply by inheritance.
Bleeding also can be very hard to evaluate. Responses to standard questionnaires suggest that an astonishingly high fraction of the population has bleeding symptoms that could plausibly be caused by a defect in VWF-dependent platelet adhesion. In various reports, at least 12% of subjects had easy bruising, 7% had gum bleeding, 5% had frequent nosebleeds, 2.4% had bleeding after tooth extraction, 1.4% had excessive postoperative bleeding, and 0.2% had excessive bleeding from trivial wounds. Among women of childbearing age, 23% had menorrhagia and 6% had postpartum bleeding [reviewed in 12]. A family history of bleeding also is very common, and was recorded for 44% of healthy children undergoing tonsillectomy . The high prevalence of bleeding symptoms in ostensibly healthy persons complicates the attempt to attribute bleeding to a low VWF level − in most cases, an association of mild bleeding and low VWF will be coincidental .
Finally, low VWF confers only a modestly increased risk of bleeding and most persons with low VWF never bleed, as demonstrated by the usual lack of symptoms among the relatives of patients with VWD type 3. Bleeding histories and VWF levels have been collected from the literature for 191 obligate heterozygous carriers of such mutations . The mean VWF level was 47 IU dL-1, about half that of the general population, with a range (± 2 SD) of 16 IU dL-1 to 140 IU dL-1. The VWF level was <50 IU dL-1 in 117 subjects, and 31 had any bleeding symptoms; VWF was >50 IU dL-1 in 74, and 10 had bleeding. These data indicate a relative risk of bleeding for VWF <50 IU dL-1 of 1.9 (P = 0.051 by Chi square test, P = 0.046 by Fisher's exact test). The only subject with postoperative bleeding had VWF >50 IU dL-1. All other bleeding was mild and included bruising, epistaxis, menorrhagia, and bleeding after tooth extraction. Heterozygous persons with VWD levels <50% had no significant bleeding, on average, so requiring severe or specific symptoms could not identify a useful subgroup.
A similar analysis suggests that the relative risk of menorrhagia, for subjects with low VWF, is about 3.9 . In studies not limited to persons diagnosed with VWD type 1, about 9% of women with menorrhagia had VWF <50 IU dL-1. A corollary is that 90% of women with menorrhagia had normal VWF levels. There is no doubt that low VWF can exacerbate menorrhagia, but persons with apparently normal hemostasis still experience bleeding. The careful investigation of control groups is essential to understand the risks associated with low VWF and the potential benefits of treatment.
The report in this issue by Bauduer and Ducout  illustrates the problems associated with using thresholds for low VWF to define VWD type 1. Their center in the French Basque region registered 27 patients who meet a commonly used definition of VWD type 1, with a VWF level at least 2 SD below the population mean (<50 IU dL-1) and a personal or family history of bleeding (criteria A). If the required VWF level was changed to at least 2 SD below the blood type O mean (<37 IU dL-1), only eight patients continued to qualify (criteria B), and just one patient had a VWF level <15 IU dL-1 and frequent bleeding (criteria C). In addition, they identified four patients with VWD type 2 and none with VWD type 3. Using these numbers, the estimated referral-based prevalence of VWD varied from 100 per million for criteria A to 17 per million for criteria C. These values for the French Basque region span the range reported worldwide for VWD prevalence in a referral setting . In this short report, data were not included on coinheritance of bleeding and low VWF, or on linkage of low VWF to the VWF locus. Similar information is not available for healthy controls, so the relative risk of bleeding for the patients is uncertain. Either ‘a significant bleeding tendency or positive inheritance’ was attributed to 10 of the 19 patients excluded by criteria B. Therefore, at least nine of them had no significant bleeding, which raised questions about the specificity of criteria A . An earlier study also found a poor correlation between bleeding and low VWF − by screening 1218 Italian schoolchildren, nine new subjects were diagnosed with VWD type 1 using a definition somewhat more stringent than criteria A , and none of them had significant bleeding over the subsequent 13 years .
It is interesting to compare the numbers of French Basque patients who satisfy the different criteria with the total numbers in the population expected at each VWF level. The region includes approximately 300 000 persons, 168 000 of whom (56%) should be blood type O . Based on the distribution of VWF levels for type O , 23 000 of them would have VWF <50 IU dL-1 (criteria A), 4200 would have VWF <37 IU dL-1 (criteria B), and only one would have VWF <15 IU dL-1 (criteria C). Almost all of the other 132 000 persons would be blood type A ; because VWF levels are higher for blood type A, they would add only 10–15% to the totals and were ignored for these rough calculations. In any case, many persons appear to be at risk for VWD type 1 based on having low VWF, but the patients in the clinic numbered just 27 (criteria A), 8 (criteria B), and 1 (criteria C) . For those with VWF levels between 15 IU dL-1 and 50 IU dL-1, the chance of being registered with a hemorrhagic disorder was a very low 0.1%, but the one hypothetical person with a VWF level <15 IU dL−1 may have found his or her way to a hematologist.
What distinguishes the subjects with low VWF who are in the hemostasis clinic from the 99.9% of them who are not? A considerable body of data discussed above suggests that low VWF, from about 15 IU dL-1 to 50 IU dL-1, rarely is associated with medically important bleeding. However, the ascertainment of patients because they actually have bleeding, rather than by the discovery of a low VWF level, could select a subgroup with additional independent risk factors, some of which might interact with a low VWF level to increase the risk of bleeding. For example, patients with both low VWF and defects in platelet aggregation were reported to have more severe bleeding . A recent study of VWD families found that increased bleeding was associated with coinheritance of specific DNA markers for platelet membrane proteins α2β1, αIIbβ3, and GPVI . In the present article by Bauduer and Ducout , two symptomatic patients with low VWF also had mild factor (F)XI deficiency. Just as cardiovascular risk factors interact to increase the likelihood of myocardial infarction (MI) or stroke, many hemostatic risk factors must surely cooperate to influence the occurence of bleeding or thrombosis. Because bleeding is uncommon for persons with moderately low VWF, it is reasonable to consider other inherited or acquired risk factors when they do bleed.
A diagnosis of VWD type 1 remains useful for persons with significant bleeding and very low VWF, often <15 IU dL-1. The coinheritance of bleeding and VWF level usually can be demonstrated in their families, and mutations in the VWF gene usually can be found [7,8]. But for the more numerous persons with slightly low VWF, between about 15 IU dL-1 and 50 IU dL-1, the modest risk of bleeding varies continuously across the entire range, and no bright line divides them into binary categories of ‘diseased’ and ‘healthy’. Persons with low VWF rarely bleed and their low VWF usually is not heritable. Consequently, the diagnosis of an inherited bleeding disease, VWD type 1, seems arbitrary and inappropriate.
Alternatively, a low VWF level could be managed as a biomarker for an increased risk of bleeding, in the same way that blood pressure and cholesterol levels are managed as cardiovascular disease risk factors. Persons with low VWF could be counseled that they have ‘low VWF’, which is associated with a moderately increased risk of bleeding. For that matter, some persons with high VWF have a moderately increased risk of thromboembolic stroke , MI [20–22], or venous thromboembolism , mediated at least partly through changes in FVIII levels. Depending on the circumstance, low or high VWF levels could be incorporated into clinical decisions, without considering the diagnosis of a genetic disease in either case.
Such an epidemiological approach requires good data and is subject to continuous refinement. For example, the threshold for treating high blood pressure has changed over time as we have learned more about cardiovascular disease and as new antihypertensive medications have been developed. Similarly, the threshold for treating low (or high) VWF levels should change as we learn more about how VWF levels influence hemostasis, and about how VWF levels interact with other biomarkers of bleeding and thrombotic risk.
Two ongoing studies of VWD type 1 should move the field forward substantially. A Canadian study has recruited 110 families with at least one member previously diagnosed with VWD type 1 . A multicenter European study has recruited 154 more highly selected families that contain at least two persons diagnosed with VWD type 1 [8,24]. The total number of subjects with low VWF will be large − over 800 persons − but as important is the inclusion of an even larger number of controls, so that information will be obtained about the magnitude of bleeding risk and how much of it can be attributed specifically to variation in VWF level and the VWF gene. These data should help considerably in deciding when to diagnose VWD type 1, and when to treat a low VWF level as a biomarker for hemostatic risk.