A quantitative analysis of bleeding symptoms in type 1 von Willebrand disease: results from a multicenter European study (MCMDM-1 VWD)

Authors


Francesco Rodeghiero, Department of Hematology, S. Bortolo Hospital, 36100 Vicenza, Italy.
Tel.: +39 0444 753626; fax: +39 0444 753922; e-mail: rodeghiero@hemato.ven.it

Abstract

Summary. Background: A quantitative description of bleeding symptoms in type 1 von Willebrand disease (VWD) has never been reported. Objectives: The aim was to quantitatively evaluate the severity of bleeding symptoms in type 1 VWD and its correlation with clinical and laboratory features. Patients and methods: Bleeding symptoms were retrospectively recorded in a European cohort of VWD type 1 families, and for each subject a quantitative bleeding score (BS) was obtained together with phenotypic tests. Results: A total of 712 subjects belonging to 144 families and 195 controls were available for analysis. The BS was higher in index cases than in affected family members (BS 9 vs. 5, P < 0.0001) and in unaffected family members than in controls (BS 0 vs. −1, P < 0.0001). There was no effect of ABO blood group. BS showed a strong significant inverse relation with either von Willebrand ristocetin cofactor (VWF:RCo), von Willebrand antigen (VWF:Ag) or factor VIII procoagulant activity (FVIII:C) measured at time of enrollment, even after adjustment for age, sex and blood group (P < 0.001 for all the four upper quintiles of BS vs. the first quintile, for either VWF:RCo, VWF:Ag or FVIII:C). Higher BS was related with increasing likelihood of VWD, and a mucocutaneous BS (computed from spontaneous, mucocutaneous symptoms) was strongly associated with bleeding after surgery or tooth extraction. Conclusions: Quantitative analysis of bleeding symptoms is potentially useful for a more accurate diagnosis of type 1 VWD and to develop guidelines for its optimal treatment.

Introduction

Von Willebrand disease (VWD) is the most common inherited bleeding disorder [1–3], usually diagnosed on the basis of the presence of mucocutaneous bleeding symptoms and reduced circulating von Willebrand factor (VWF) levels [4,5]. The genetic transmission is generally co-dominant or dominant [6]. Although the first report of the disease dates back to 80 years ago [7], with mucocutaneous bleeding as the prominent clinical manifestation [8], a systematic, quantitative description of bleeding symptoms associated with VWD is still awaited. Therefore, it is not known how bleeding symptoms are influenced by age and gender in VWD patients, or whether non-affected family members may refer bleeding symptoms more than subjects not belonging to VWD families. It is also substantially unknown whether different bleeding symptoms may preferentially cluster together with some particular pattern within VWD patients, and whether the presence of some bleeding symptoms may actually predict occurrence of bleeding in circumstantial situations (e.g. surgery). Furthermore, we do not know the degree of association between bleeding symptoms and laboratory VWF-related measurements.

Several attempts have been made at assessing the usefulness of the bleeding history in different inherited bleeding disorders [9–12], but no standard criteria for the evaluation of bleeding in VWD is currently available. Recently, however, a quantitative bleeding score (BS), related to the number and the severity of bleeding symptoms, has been employed to validate clinical criteria for the diagnosis of type 1 VWD in an international multicenter study [13]. Using a simplified questionnaire and a very similar BS system, we have now evaluated bleeding severity in a large panel of type 1 VWD families enrolled in a European VWD study, the Molecular and Clinical Markers for the Diagnosis and Management of Type 1 von Willebrand disease (MCMDM-1 VWD). In particular, we aimed at evaluating the pattern of bleeding symptoms in type 1 VWD families and the association between bleeding score (BS) and both clinical and laboratory characteristics of enrolled patients.

Patients and methods

Subjects

The ‘MCMDM-1 VWD’ Study is a multicenter, EU funded, survey on type 1 VWD (see also http://www.shef.ac.uk/euvwd/ for further details; accessed 26 January 2006). In the study 12 centers, all experienced in the diagnosis and management of individuals and families with inherited bleeding disorders, recruited a total of 154 families from nine European countries. The aim of the MCMDM-1 VWD study is to determine the value of clinical, phenotypic and molecular markers for the diagnosis of type 1 VWD, through an extensive evaluation of phenotype, linkage and mutation analysis in unselected families diagnosed by the evaluating center as affected by type 1 VWD. The study intentionally sought to include also milder cases of type 1 VWD, and no diagnostic criteria were specified a priori for the diagnosis of type 1 VWD in patients. However, since linkage studies were planned, the only other recruitment requirement in addition to an institutional diagnosis was the presence within a family of at least two subjects classified as affected. Efforts were made however, to include as many family members as possible, possibly with at least two generations available. In addition, approximately 100 normal controls were recruited at each center, for a total of 1166 normal individuals. Normal controls were defined as subjects who never sought medical attention for a bleeding symptom before enrollment. As a matter of fact, no subject was excluded from enrollment as a normal control according to this criterion. Informed consent was obtained from all subjects. Within the families, subjects were classified by the enrolling center as index case (IC, subject who led to investigation) and affected or unaffected family member (AFM and UFM respectively), on the basis of VWF levels and bleeding symptoms at time of investigation of the IC and according to the enrolling institutional criteria. For the present study, we accepted the classification of enrolled subjects as affected or unaffected given by the center without further investigation of the true genetic status of the subjects.

Bleeding questionnaire and score

A bleeding questionnaire was administered by a physician to each enrolled family member and to 195 of the 1166 normal controls. The physician was aware of the diagnosis of the patient but unaware of his/her previous hemorrhagic history. The questionnaire was essentially similar to that used for a previous investigation [13], and all clinical information was entered in a centralized database maintained in Sheffield. From the database, we directly computed a symptom-specific score for each bleeding symptom, according to the grading criteria reported in Table 1 that were prespecified by the study steering committee and not available to the field physician. The grading criteria were essentially the same as previously reported [13], with slight modifications introduced at time of study design to improve the sensitivity of the questionnaire. First, an additional grade of bleeding severity was included for each bleeding symptom. Second, a −1 grade was introduced to account for circumstantial situations associated with an high bleeding risk (e.g. tooth extraction), in which nonetheless bleeding did not occur despite the fact that no anti-hemorrhagic prophylaxis was given. For each individual, a summative BS was computed as the sum of the each symptom-specific grading, and could theoretically range from −3 (no spontaneous bleeding symptom, no bleeding after surgeries, teeth extractions and deliveries) to +45 (major bleeding for all symptoms). Extraction of information from the database, generation of the symptom-specific grading and calculation of the final BS for each individual was automatically performed using an ad hoc software written in Stata [14]. As diagnosis may influence the severity of bleeding symptoms by introducing anti-hemorrhagic prophylaxis, the BS was computed using symptoms that occurred before diagnosis of VWD or in subjects who did not receive prophylaxis. The full bleeding questionnaire and criteria used to compute the bleeding score are available at http://www.shef.ac.uk/euvwd/bleed_score.htm.

Table 1.  Assigned score for each bleeding symptom
SymptomScore
−101234
EpistaxisNo or trivial (less than 5)>5 or more than 10′Consultation onlyPacking or cauterization or antifibrinolyticBlood transfusion or replacement therapy or desmopressin
CutaneousNo or trivial (<1 cm)>1 cm and no traumaConsultation only  
Bleeding from minor woundsNo or trivial (less than 5)>5 or more than 5′Consultation onlySurgical hemostasisBlood transfusion or replacement therapy or desmopressin
Oral cavityNoReferred at least oneConsultation onlySurgical hemostasis or antifibrinolyticBlood transfusion or replacement therapy or desmopressin
Gastrointestinal bleedingNoAssociated with ulcer, portal hypertension, hemorrhoids, angiodysplasiaSpontaneousSurgical hemostasis, blood transfusion, replacement therapy, desmopressin, antifibrinolytic 
Tooth extractionNo bleeding in at least two extractionNone done or no bleeding in one extractionReferred in <25% of all proceduresReferred in >25% of all procedures, no interventionResuturing or packingBlood transfusion or replacement therapy or desmopressin
SurgeryNo bleeding in at least two surgeriesNone done or no bleeding in one surgeryReferred in <25% of all surgeriesReferred in >25% of all procedures, no interventionSurgical hemostasis or antifibrinolyticBlood transfusion or replacement therapy or desmopressin
MenorrhagiaNoConsultation onlyAntifibrinolytics, pill useDilatation and currettage, iron therapyBlood transfusion or replacement therapy or desmopressin or hysterectomy
Postpartum hemorrhageNo bleeding in at least two deliveriesNo deliveries or no bleeding in one deliveryConsultation onlyDilatation and currettage, iron therapy, antifibrinolyticsBlood transfusion or replacement therapy or desmopressinHysterectomy
Muscle hematomasNeverPost trauma no therapySpontaneous, no therapySpontaneous or traumatic, requiring desmopressin or replacement therapySpontaneous or traumatic, requiring surgical intervention or blood transfusion
HemarthrosisNeverPost trauma no therapySpontaneous, no therapySpontaneous or traumatic, requiring desmopressin or replacement therapySpontaneous or traumatic, requiring Surgical intervention or blood transfusion
Central nervous system bleedingNeverSubdural, any interventionIntracerebral, any intervention

Laboratory tests

Twenty milliliters of citrated blood were obtained from all investigated family members and normal controls, and plasma was immediately aliquoted and stored at −80° for subsequent measurements. Each enrolling center was asked to measure with their own method and against the 4th WHO Standard both von Willebrand ristocetin cofactor (VWF:RCo), von Willebrand antigen (VWF:Ag) or factor VIII procoagulant activity (FVIII:C) in a plasma aliquot and to provide the median of the past measurements of the VWF:RCo, VWF:Ag and FVIII:C, when available as IU dL−1. VWF:RCo and VWF:Ag measurements were confirmed in a central laboratory (Vicenza) on separate aliquots. In the central laboratory, VWF:RCo activity was measured using ristocetin as the aggregating agent (final concentration, 1 mg mL−1) of a formalin-fixed platelet suspension; VWF:Ag was measured using a enzyme-linked commercial assay (VWF Assay, Stago, Asnières, France). Both measurements were carried out against the 4th WHO Standard for VWF/FVIII and expressed as IU dL−1. For the purpose of the present analysis, VWF:RCo and VWF:Ag values measured in Vicenza were considered together with FVIII:C values measured by each enrolling center.

Statistics

As the BS is a summative index of ordinal values not normally distributed, we used median values and a non-parametric test (Mann–Whitney ranksum test) to assess differences in BS between subgroups of enrolled subjects. Analysis of the association between bleeding symptoms and VWD carriership was performed using a logistic model, with the grading of each bleeding symptom being categorized in two values (below or above 1) to indicate absence or presence of a clinically relevant symptom. Age and gender were added in the logistic model as covariates. We subsequently hypothesized that bleeding symptoms do not occur independently within a subject, but do tend to cluster (e.g. a subject with a history of epistaxis is more likely to have another bleeding symptom). This hypothesis was tested using a log-linear model evaluating all the possible interactions between bleeding symptoms (coded for absence or presence of a specific symptom, as above). To test the relationship between BS and laboratory quantitative values, the BS was divided into quintiles and differences in mean VWF:RCo, VWF:Ag and FVIII:C were tested among the different quintiles of BS, using a multiple regression model adjusted for age, gender and O blood group. Likelihood ratios of VWD for each BS value and their 95% confidence intervals were computed as the ratio between the percentages of affected subjects (either IC or AFM) to controls with that particular BS value [15]. Finally, we tested whether the BS was associated with bleeding in circumstantial situations, that is, if a quantitative history of bleeding could be useful to anticipate bleeding after surgery or tooth extraction. To this purpose, the score of mucocutaneous bleeding symptoms only (epistaxis, cutaneous bleeding and bleeding from minor wounds) was considered in two distinct receiver-operator curve (ROC) analyses predicting bleeding after surgery or tooth extraction (categorized as no bleeding or bleeding of any severity). ROC analysis was used because of its ability to directly compare the amount of information contained in different diagnostic tests, the latter being expressed both in a quantitative or ordinal [15], therefore allowing to directly compare the predictive value of mucocutaneous bleeding symptoms with VWF/FVIII:C measurements. All computations were performed using the Stata software package [14].

Results

Enrolled subjects and laboratory data

A total of 744 subjects from 154 families with type 1 VWD were initially recruited. After analysis of the family data, 10 families were excluded from further analysis in the present study. Seven families were discarded because of deviation from the enrollment criteria (only one affected family member); three families were discarded because reduced VWF levels could not be demonstrated in any family member. An additional subject was discarded from a family because he was a hemophilia A patient. Therefore, data from 712 subjects belonging to 144 families were available for evaluation. At least one hemorrhagic symptom was present in 142/144 (98.6%) and in 249/273 (91.2%) of IC and AFM, respectively (P = 0.59), while three or more hemorrhagic symptoms were present in 114 of 144 (79.2%) and 156 of 273 (57.1%) of IC and AFM, respectively (P < 0.0001).

There was a good correlation between VWF:RCo and VWF:Ag as measured locally or centrally in Vicenza (r2 0.74 and 0.86, respectively). The median of the past measurements was lower than that measured at study enrollment, although this difference was significant only for FVIII:C values (VWF:RCo 56.0 vs. 60.3 IU dL−1, P = 0.185; VWF:Ag 59.0 vs. 62.5 IU dL−1, P = 0.243; FVIII:C 67.0 vs. 74.4 IU dL−1, P = 0.01). Table 2 reports the clinical characteristics of the investigated subjects in whom the BS was available. In the 144 investigated families, the average number of members was 4.9. In addition to the IC, the average number of AFM and UFM within the families was 1.9 and 2.0 respectively. There was an increasing fraction of females and blood group O subjects in AFM vs. UFM and in IC vs. AFM (see Table 2; P < 0.0001, chi-squared test), whereas there was no difference of either VWF:RCo, VWF:Ag or FVIII:C between IC and AFM within the families. Within the families, there was a significant relationship between the VWF:RCo, VWF:Ag and FVIII:C of the IC and that observed in other AFM (r2 = 0.23, 0.26 and 0.19 respectively, P < 0.0001 for each one). This relationship was not observed for BS (r2 = 0.005, P = 0.22).

Table 2.  Characteristics of the investigated subjects
 Index case (n = 144)Affected family members (n = 273)Unaffected family members (n = 295) PControls (n = 195)
  1. *Tests the null hypothesis of no differences between index cases, affected and unaffected family members; Age was not available for six controls; Tests the null hypothesis of equal means between index cases and affected family members (non-parametric Mann–Whitney test).

Females (%)91 (63.2)151 (55.3)145 (49.1)0.02*100 (51.3)
Median age, years (range)40 (1–80)32 (2–91)41 (3–90)0.002*40 (8–78)
Age groups
 Below 15 years155245 
 15–30 years347249 41
 30–50 years668392 55
 50–70 years244888 89
 >70 years51821 4
Blood group O (%)96 (66.6)161 (58.9)141 (47.8)0.001*75 (38.4)
Median VWF:RCo, IU dL−1 (25–75 percentile)34 (11–49)35 (12–54)87 (65–112)0.12106 (83–138)
Median VWF:Ag, IU dL−1 (25–75 percentile)33 (19–49)36 (21–54)92 (70–118)0.1199 (76–121)
Median FVIII:C, IU dL−1 (25–75 percentile)56 (28–77)63 (34–88)100 (78–133)0.08113 (95–113)
Median BS (range)9 (−1 to 23)4 (−2 to 27)0 (−2 to 14)−1 (−3 to 4)

Bleeding score and bleeding symptoms

Figure 1 summarizes the finding of the BS in the studied subjects. In 195 controls with an available bleeding history, the median BS was −1 in both males and females. Only one female control had a BS of 4 (0.5%), while all the remaining control subjects had a BS lower than 4. UFM had a BS higher than control subjects (median score 0 vs. −1, P < 0.0001), and IC had a BS significantly higher than AFM (median 9 vs. 4 respectively, P < 0.0001). There was a clear trend for increasing BS with age in AFM and IC but not in controls or UFM (Fig. 1) at time of diagnosis. Furthermore, in IC and AFM females had a higher BS than males (median score 11 vs. 6, P < 0.00001 in IC, 6 vs. 3, P < 0.0004 in AFM), a finding not confirmed in UFM and controls. Group O subjects had a higher median BS than non-O subjects within families (3 vs. 2, P = 0.001). However, this was related to the excess of blood group O in IC, as after adjustment for carriership status this difference disappeared.

Figure 1.

Bleeding score and age in von Willebrand disease families and normal controls. For each category, the boxes represent the 50% of subjects centered around the median. Age is presented on the x-axis.

Figure 2 reports the association between bleeding symptoms in AFM vs. UFM in VWD families, using a logistic model adjusted for age and gender. ICs were excluded from this analysis to avoid a selection bias. All bleeding symptoms, with the exception of postpartum hemorrhage and oral or gastrointestinal bleeding, were significantly associated with AFM, cutaneous bleeding being the symptoms most associated with VWD when compared with normal controls. Using a log-linear model, a statistically significant association of bleeding symptoms was evident for each possible combination, the lowest P-values being observed for epistaxis together with postsurgical or oral bleeding and for cutaneous bleeding and menorrhagia (P < 0.001). In women aged >14 years, menorrhagia accounted for 11.1%, 23.4% and 18.9% of the total BS in UFM, AFM and IC, respectively.

Figure 2.

Association between bleeding symptoms and type 1 von Willebrand disease in the enrolled families in an age-adjusted logistic model. Index cases were excluded from the analysis; a bleeding symptom was considered in the model for a symptom-specific score greater than one. For each bleeding symptom, the graph reports the logistic estimate and its 95% confidence interval.

The correlation between the BS and plasma levels of either VWF:RCo, VWF:Ag or FVIII:C was tested with three multiple regression models, modeling the levels of the above mentioned laboratory values as a function of quintile of the BS adjusted for age, sex and blood group. A strong, inverse correlation was found between all the quintiles of BS and VWF:RCo, VWF:Ag and FVIII:C (P < 0.001 for all the four upper quintiles of BS vs. the first quintile, for all considered laboratory measurements, see Fig. 3). The mean interquintile differences in VWF:RCo, VWF:Ag and FVIII:C were 14.1, 13.1 and 11.2 respectively.

Figure 3.

Association between bleeding score and von Willebrand factor (VWF)/FVIII:C levels. For each quintile of bleeding score, the boxes span from the 25th to the 75th percentile. The center line represents the median value; dark gray boxes: VWF:RCo; light gray boxes, FVIII:C; white boxes, VWF:Ag.

Table 3 reports the likelihood ratios of VWD computed for BS between −3 and 4, as no control subjects had a score higher than 4. Likelihood ratios were computed pooling data from both IC and AFM, as a separate analysis did not show significant differences between IC and AFM.

Table 3.  Likelihood ratios for von Willebrand factor at different levels of bleeding score
Bleeding scoreLikelihood ratio (95% confidence interval)Number of subjects (%)
ICAFMUFMControls
  1. AFM, affected family member; UFM, unaffected family member; IC, index case.

−30.00000 (0)0 (0)0 (0)9 (4.6)
−20.035 (0.011–0.113)0 (0)3 (1.1)25 (8.3)39 (20.0)
−10.103 (0.060–0.176)1 (0.7)14 (5.0)74 (24.7)67 (34.3)
00.129 (0.076–0.219)2 (1.4)14 (5.0)102 (34.0)57 (29.2)
11.590 (0.771–3.270)3 (2.1)28 (10.1)31 (10.3)9 (4.6)
22.240 (1.010–4.960)9 (6.1)25 (9.0)26 (8.7)7 (3.6)
33.000 (1.290–6.960)10 (6.8)29 (10.5)9 (3.0)6 (3.1)
415.70 (2.160–114.0)9 (6.1)25 (9.0)10 (3.3)1 (0.5)
≥5110 (76.8)135 (48.7)18 (9.5)– (0)

Mucocutaneous bleeding score and bleeding after surgery or tooth extraction

The mucocutaneous BS behaved similarly to VWF:RCo and VWF:Ag but was superior to FVIII:C for the prediction of bleeding after tooth extraction (ROC area 0.71 vs. 0.64, P =0.01; Fig. 4). The mucocutaneous BS was superior to either VWF:RCo, VWF:Ag and FVIII:C levels for the prediction of surgical bleeding (ROC area 0.78 vs. 0.67, 0.67 and 0.65, respectively; P = 0.02; Fig. 5). All these analyses were performed after exclusion of ICs to avoid a selection bias.

Figure 4.

Receiver-operator curve analysis for bleeding after tooth extraction as a function of the mucocutaneous bleeding score (bsl00043), von Willebrand factor (VWF):RCo (bsl00067), VWF:Ag (bsl00084), FVIII:C (□).

Figure 5.

Receiver-operator curve analysis for bleeding after surgery as a function of the mucocutaneous bleeding score (bsl00043), von Willebrand factor (VWF):RCo (bsl00067), VWF:Ag (bsl00084), FVIII:C (□).

Discussion

The present study is the first attempt at offering a quantitative evaluation of bleeding symptoms in type 1 VWD families. Most current knowledge on the clinical picture of VWD is based on descriptive reports of the presence of bleeding symptoms, lacking however any analysis of their severity. Recently, we designed a questionnaire together with a quantitative BS to validate the clinical diagnostic criteria of VWD in a sample of type 1 VWD obligatory carriers [13]. In that study, we demonstrated that the BS could be a reliable tool for the identification of VWD patients. Using basically the same methodology, we analyzed a large cohort of type 1 VWD families enrolled within a multicenter, EU funded study (MCMDM-1 VWD).

The symptoms that were mostly associated with VWD within these families were bleeding after minor wounds and cutaneous bleeding, whereas postpartum bleeding, bleeding from the gastrointestinal tract and oral bleeding had the same frequency observed in UFM. This finding is basically in keeping with those of the above mentioned multicenter study comparing obligatory VWD carriers and healthy subjects [13] and with those of Silwer [8], who also found that bleeding after minor wounds was the most distinctive feature of VWD. Furthermore, there was a clear trend for increasing likelihood ratios of VWD with an increasing BS, confirming our previous report in a more selected set of patients [13]. There was also a clear tendency for clustering of bleeding symptoms in VWD patients, in particular for epistaxis with postsurgical or oral bleeding and for cutaneous bleeding with menorrhagia. This indicates that in VWD patients bleeding symptoms tend to follow a particular pattern of mucocutaneous bleeding rather than occurring only by chance. From a quantitative point of view, we observed a wide variability of bleeding severity within the families. ICs were more symptomatic than the other AFM, while having similar levels of either VWF:RCo, VWF:Ag or FVIII:C. This finding is certainly suggestive for the presence of a selection bias for IC, but it is likely to reflect also a bias in the diagnostic process. IC are likely identified on the basis of a presenting bleeding symptom, while AFM are more often diagnosed on the basis of abnormal laboratory values alone rather than on the combination of both abnormal laboratory values and manifest bleeding symptoms. This hypothesis is also in keeping with the finding that a clear correlation of VWF levels was present between IC and AFM within families, whereas this relation was not confirmed for the BS (i.e. the bleeding severity in IC does not predict symptoms in the remaining AFM). On a pathogenetic basis one could also speculate that IC may be more symptomatic because of inheritance of additional pro-hemorrhagic defects, with a multigenic pattern of inheritance similar to that found for some thrombophilic disorders [16]. This latter possibility could also be purported by the finding that UFM have a slightly higher BS than normal controls, suggesting that other subtle abnormalities could be present in these families, as recently reported [17].

The risk of bleeding in VWD patients, either IC or AFM, appears to be constantly increased through the whole life, as suggested by the steady increasing BS with age and it seems to be similar in both male and females. By contrast, normal controls and UFM do not show an appreciable bleeding risk, as demonstrated by the near zero BS throughout all ages (Fig. 1). However, further prospective studies are needed to better estimate the absolute annual risk of bleeding in these patients.

The clinical severity of the disease was clearly related to the levels of circulating plasma VWF and FVIII, the highest quintiles of the BS distribution being associated with the lowest levels of either VWF:RCo, VWF:Ag or FVIII:C (Fig. 3). Interestingly, this association was linear and present also for VWF:RCo values still within the normal range.

A key question in VWD diagnosis and treatment is whether laboratory or clinical data could be useful for the prediction of bleeding under circumstantial situations (e.g. surgery or tooth extraction). To address this issue, a mucocutaneous BS, formed by summing only spontaneous bleeding symptoms, was compared with the circulating levels of VWF and FVIII:C in the prediction of bleeding after surgery or tooth extraction in a ROC analysis. Clinical assessment proved to be at least as effective as laboratory testing for the prediction of bleeding after tooth extraction, but it was superior in the prediction of bleeding after surgery. Although this finding comes from a retrospective assessment, and it could be therefore biased by a differential recall in those subjects having major hemorrhage after surgery, it suggests that adults with a VWD diagnosis but without a clear-cut history of bleeding could be possibly at lower risk of bleeding. This hypothesis should be prospectively validated.

Apart from the preferential inclusion of very symptomatic IC in the study, another limitation of the present study pertains to the classification of subjects as AFM or UFM. When the present study was undertaken, we were not aware of the true genetic status of most families, as linkage and mutation analyses were still in progress. Most of the families were diagnosed before consensus criteria for the diagnosis of VWD were available [4,5] so that the classification assigned by the enrolling center was used for the analysis. In most circumstances, the diagnoses were validated with repeated measurements over time at each enrolling center, but in some cases VWF measurements were taken only once and were not confirmed during the present reinvestigation. This finding, that is in keeping with the well-known fluctuation of VWF levels [18], could explain why in the present study some individuals were classified as IC or AFM even in the presence of completely normal laboratory values at the time of study enrollment.

In conclusion, a quantitative analysis of bleeding symptoms disclosed subtle differences in bleeding severity among IC and AFM and between UFM and controls. Likelihood ratios for type 1 VWD however rose significantly with an increasing BS. There was a close correlation of severity of bleeding symptoms with VWF levels, and AFM with more severe mucocutaneous symptoms had more bleeding complications after invasive procedures (tooth extraction or surgery). Thus, use of a standardized BS is potentially useful to further dissect the association between VWF function and bleeding, to establish an optimal diagnosis of type 1 VWD and to evaluate the bleeding risk in VWD patients.

Addendum

F.R., A.G. and I.P. were responsible for study initiation and coordination. F.R., G.C., A.G., A.B.F., J.B., D.M., E.F., C.M., J.G., J.E., R.S., U.B., J.I., Z.U., D.H., L.H., S.L., J.P., F.H. and I.P. were involved in study design, data collection and performing laboratory analyses. A.T., F.R., G.C., A.G. and I.P. were responsible for analysis and interpretation of results. A.T. performed statistical analysis. A.T. was the lead author of the initial manuscript. F.R., G.C., A.G., A.B.F., J.E., R.S. and I.P. were responsible for revisions of draft manuscripts. F.R., G.C., A.G., A.B.F., J.B., D.M., E.F., C.M., J.G., J.E., R.S., U.B., J.I., Z.V., D.H., L.H., S.L., J.P., F.H. and I.P. were responsible for review and approval of the final manuscript.

Acknowledgements

We wish to thank: Sara Roberti for her invaluable help as a data manager in Vicenza, particularly for writing most Stata procedures and for the log-linear modeling of bleeding symptoms; Anna Cappelletti for the general management of clinical data; Martina Bernardi and Kelly Bertoncello for their technical experience in VWF measurements.

Ancillary