The role of ADAMTS‐13 and von Willebrand factor in cancer patients: Results from the Vienna Cancer and Thrombosis Study

Abstract Essentials Cancer is associated with increased risk of developing venous thrombosis. Cancer patients were studied for ADAMTS‐13 and VWF levels and occurrence of venous thrombosis. Increased VWF in cancer patients is associated with a higher risk of venous thrombosis. Low levels of ADAMTS‐13 and/or increased VWF in cancer patients are associated with worse survival. Background Cancer‐associated venous thromboembolism (VTE) is an important complication in the course of a malignant disease. Low ADAMTS‐13 (a disintegrin‐like and metalloproteinase with thrombospondin type 1 motif 13) and increased von Willebrand Factor (VWF) levels in cancer patients have been described numerously. Objectives Investigation of the influence of ADAMTS‐13 and VWF on the probability of VTE and survival in malignancy. Patients/Methods In the framework of the ongoing prospective Cancer and Thrombosis Study (CATS) ADAMTS‐13 activity and VWF antigen levels were investigated in cancer patients. Results In total, 795 patients with various tumor types (364 female/431 male, median age 62 years) were included; of those, 56 developed VTE and 359 patients died during a median follow‐up time of 730 days. The hazard ratio (HR) of VTE per doubling of VWF level was 1.56 (95% confidence interval [CI] 1.13‐2.16) in multivariable competing risk analysis. ADAMTS‐13 levels showed no correlation with the incidence of VTE in univariate competing risk analysis. The HR of mortality per doubling of VWF level was 1.46 (95% CI 1.28‐1.66) and per SD increment of ADAMTS‐13was 0.90 (95% CI 0.81‐1.00) in multivariable Cox regression analysis. Patients with VWF >75th percentile and concomitant low (<25th percentile) or medium (25‐75th percentile) ADAMTS‐13 values had the highest probability of mortality (HR 4.31 and 4.75, respectively). Conclusions High VWF levels were significantly associated with the risk of developing VTE in cancer patients, whereas ADAMTS‐13 was not. Low ADAMTS‐13 and increased VWF levels were independently associated with worse overall survival.


| INTRODUCTION
Malignancy is an acquired hypercoaguable state with increased risk for venous thromboembolism (VTE). 1 Cancer-related VTE is associated with higher mortality 2 and disease progression. 3 In a recent Cochrane review it was discussed that thromboprophylaxis may potentially alleviate this frequent clinical problem, however, it comes with an increased risk of bleeding complications in cancer patients. 4 Therefore, tailor-made treatment based on reliable, individual risk stratification is required. Several risk factors that may contribute to the development of VTE in cancer patients have been described and different risk prediction scores were established. 5 Von Willebrand factor (VWF) is a large polymeric glycoprotein involved in the adhesion and aggregation of platelets, particularly during vascular endothelial lesions. VWF is secreted into the plasma in the form of hemostatically highly active, "ultra-large"-VWF multimers (ULVWF). 6 After conformational unfolding these multimers are cleaved by ADAMTS-13 (a disintegrin-like and metalloproteinase with thrombospondin type 1, motif 13). 7 Cleavage results in smaller, less active VWF subunits. Complete deficiency in ADAMTS-13 is the cause of thrombotic thrombocytopenic purpura (TTP), a thrombotic microangiopathy (TMA) whose hallmark symptoms are platelet-and VWF-rich microvascular thrombi. 8 Cancer patients have been shown to have higher VWF levels and lower ADAMTS-13 levels than the general population, [9][10][11][12][13][14][15][16][17][18][19][20][21] often also in a stage-dependent intensity. [9][10][11][12]16,20,[22][23][24][25][26][27][28] In a recent study VWF and ADAMTS-13 were shown to be associated with occurrence of VTE in cancer patients, with ADAMTS-13 exhibiting predictive potential in risk scores. 21 Patients with VTE or history of VTE have been repeatedly shown to have higher mean levels of VWF, [22][23][24][25][26][27] however, patients with underlying malignancy have always been excluded. Data on ADAMTS-13 and its association with VTE, on the other hand, is ambiguous. There are studies showing decreased 28 or increased 25 ADAMTS-13 activity with venous thrombosis. With next-generation sequencing, one scientific group found an excess of rare coding single-nucleotide variants of the ADAMTS-13 gene in patients with deep vein thrombosis (DVT) 29 while another group failed to find a link between DVT and a polymorphism associated with reduced levels of ADAMTS-13. 26 The interrelationship and calculated ratio between ADAMTS-13 and VWF are also of particular interest. This topic has recently been studied in several settings and has been shown to be of potential use for predicting survival in patients with lung cancer 30 or thrombotic complications in patients after hepatectomy. 31 The aim of this investigation was to study the potential predictive value of ADAMTS-13, VWF and their interrelationship for development of VTE in cancer patients and the correlation of values with survival probability in a large patient cohort.

| Study design
The Vienna Cancer and Thrombosis Study (CATS) is an ongoing, prospective cohort study, approved by the local ethics committee and performed in accordance with the Declaration of Helsinki. It investigates potential predictive parameters for cancer-related VTE. Cancer patients from different departments of the General Hospital of Vienna were included according to the criteria given in the consort diagram ( Figure 1).
Study inclusion was performed via personal interview by a trained physician. Venous blood was drawn, mixed with one-tenth volume sodium citrate stock solution to prevent clotting, centrifuged twice to obtain platelet-free plasma (at 1500 g for 15 minutes and then 13 400 g for 2 minutes), frozen and stored at −80°C.
Patients were educated on the possible signs and forms of presentation of VTE and advised to contact the study administration upon occurrence of any symptoms. Questionnaires regarding current medical status and possible VTE were sent by postal mail to patients in 3-to 4-monthly intervals. If there was no response from the patient, information was sought by contacting family members, general practitioners or attending oncologists, and by annual check of the Austrian death registry regarding included study participants.
The study end point was an objectively confirmed VTE within a 2-year observation period. Objective imaging methods to confirm VTE upon symptoms were Duplex sonography or venography for DVT or computed tomography or ventilation/perfusion lung scan for pulmonary embolism (PE). In patients that had died during follow-up, death certificates and, if available, autopsy findings were reviewed to establish a diagnosis of fatal PE. All thrombotic events had to be confirmed by an adjudication committee, comprising independent specialists in angiology, radiology or nuclear medicine.
All authors had access to primary clinical data, which were analyzed by H.L.O, J.R., C.A., and I.P.

Essentials
• Cancer is associated with increased risk of developing venous thrombosis.
• Cancer patients were studied for ADAMTS-13 and VWF levels and occurrence of venous thrombosis.
• Increased VWF in cancer patients is associated with a higher risk of venous thrombosis.
• Low levels of ADAMTS-13 and/or increased VWF in cancer patients are associated with worse survival.

| Statistical analysis
Continuous variables were described with the median and the interquartile range (IQR). Categorical variables were described by the absolute number and percentages. Spearman correlation coefficient was used to describe the correlation between continuous variables.
Median follow-up time was calculated by the reverse Kaplan-Meier method. 35 The analysis of variance models with the Tukey-HSD test were applied to test for differences between tumor groups and stages.
Competing-risk analysis for estimating the relative risk of VTE in the observation period was calculated according to the Fine and Gray proportional hazard subdistribution model. 36 Within these regression models objectively confirmed VTEs were considered to be the event of interest, whereas deaths without developing VTE were treated as competing events and patients having reached the end of the observation period or being lost to follow-up alive and without developing VTE were included as censored observations. As the amount of events limited the number of prognostic factors to be considered simultaneously, two multivariable models were designed. For the evaluation of the risk of VTE the first model F I G U R E 1 Consort diagram. *Antiplatelet therapy as well as temporary prophylactic treatment with low molecular weight heparin was accepted. Patients under prior or concomitant antineoplastic therapy were excluded for reason of possible transient effect on the coagulation system Evaluation Study Population Assessed for eligibility (n = 891 ) Inclusion criteria (i) newly diagnosed cancer or progression of the disease after complete or partial remission of the brain, breast, lung, stomach, lower gastrointestinal tract, pancreas, kidney, prostate or other sites and hematologic malignancies (multiple myeloma, high and low grad lymphoma) (ii) a histological confirmation of diagnosis, (iii) age 18 or older, (iv) willingness to participate, (v) provision of written informed consent Exlusion criteria* (i) venous or arterial thromboembolism within the past 3 months, (ii) overt viral or bacterial infection or (iii) major surgery within the past 2 weeks (iv) chemo-or radiation therapy within the 3 months prior to inclusion or (v) continuous anticoagulant therapy To evaluate the influence on overall survival, three Cox regression models with the same parameters as above were applied; one for ADAMTS-13 and VWF each and one with both factors combined.
The event of interest was death of any cause and data were considered to be censored for patients having reached the end of the observation period or who were lost to follow-up.
Because the distribution of values of VWF and ADAMTS-13/ VWF over the study population were skewed to the right, they needed logarithmic scaling to ensure correct statistical analysis. Values were scaled to the base of "2" which caused VWF and ADAMTS-13/VWF to have the unit "per double increase" within the regression models by default. ADAMTS-13 values, however, were distributed symmetrically, therefore the selected unit for ADAMTS-13 values was a standard deviation (SD) increment.
To assess a possible non-additive effect of ADAMTS-13 and VWF on VTE and mortality, an interaction analysis was performed.
Therefore, the product of ADAMTS- 13  To minimize potential bias laboratory analysis was blinded, which means that the persons who performed the laboratory analysis were not aware of the patients' outcomes.
Two-sided P-values less than 0.05 were regarded as statistically significant.
Statistical analysis was performed with SAS 9.4 (SAS Institute Inc., Cary, North Carolina) and IBM SPSS Statistics 20.0 statistical software.

| Characteristics of study participants
The total study population included 795 cancer patients; the basic demographic data are summarized in Table 1. Patients were followed up prospectively over a median observation period of 730 days (IQR 273-731), 20 patients (2.5%) were lost for follow-up.

| ADAMTS-13 activity and VWF antigen levels
Median values for ADAMTS-13 were lowest in lung and colorectal cancers but within normal range of 40%-130% for most patients (Table 2). Only one patient with severe deficiency in ADAMTS-13 (3% ADAMTS-13 activity) was observed, another nine patients had moderately low levels (10%-25%) and 25 patients had mildly reduced levels (25%-40%), which together made up only 4.4% of the total study population. VWF values were highest in lung and pancreatic cancers. The ADAMTS-13/VWF ratio was lowest in those with lung, colorectal and pancreatic cancers ( Table 2).
ADAMTS-13 and ADAMTS-13/VWF ratio were significantly lower in metastatic compared to localized disease (P < 0.01 and P < 0.001, respectively). By contrast, VWF was significantly higher in patients with metastases compared to those without (P < 0.001) ( Table 2).
Very low ADAMTS-13 values were not necessarily associated with very high VWF. There was only a weak correlation between ADAMTS-13 and VWF (r = −0.126, P < 0.001), whereas VWF had a strong correlation with FVIII (r = 0.65, P < 0.001) and a positive correlation with D-dimer (0.356, P < 0.001).

| Relationship between ADAMTS-13 and VWF
It could be shown that ADAMTS-13 and VWF do not interact with each other in our analyses. Investigation for a possible non-addi- a 199 patients were not classifiable according to degree of metastasis (brain cancers, lymphomas and multiple myelomas). *P < 0.05, † P < 0.01, ‡ P < 0.001 for analysis of variance models for differences between tumor groups and stages.

F I G U R E 2 Cumulative incidence of VTE in competing risk analysis according to (A) ADAMTS-13 values, (B) VWF values, and (C)
ADAMTS-13/VWF ratio classified into three groups of values above the 75th percentile, from the 25-75th percentile, and below the 25th percentile. Patients at risk are given at 6-mo intervals, observation period was 2 y (731 d). Patients with VWF levels >75th percentile and patients with ADAMTS-13/VWF levels <25th percentile had higher probability of developing VTE. Difference in cumulative incidence was statistically significant for VWF and ADAMTS-13/VWF (P < 0.001 and P < 0.05, respectively). No statistically significant association for levels of ADAMTS-13 and VTE could be found. Kaplan-Meier plots for survival probability of cancer patients according to (D) ADAMTS-13 levels, (E) VWF levels, and (F) the ADAMTS-13/VWF ratio classified into three groups of values above the 75th percentile, from the 25-75th percentile and below the 25th percentile. Patients at risk are given at 6-mo intervals, observation period was 2 y (731 d). Patients with higher ADAMTS-13 and ADAMTS-13/VWF and lower VWF levels had the best overall survival probability. Levels of ADAMTS-13 and ADAMTS-13/VWF <25th percentile or VWF >75th percentile were associated with worse survival. Differences in survival probability among groups compared with the logrank test were all highly statistically significant with P < 0.  Figure 3). When interpreting this data it has to be kept in mind that, because of the division of the patient cohort according to the levels of both factors, group sizes are not equally distributed. The absolute patient numbers are given in Table 5.

| DISCUSSION
In the present cancer patient cohort a clear association between elevated levels of VWF and decreased levels of the ADAMTS-13/ VWF ratio and occurrence of VTE was found. In competing risk analysis patients with VWF values above the 75th percentile had a roughly three-fold risk of developing VTE compared to those below the 25th percentile after 2 years. For the ADAMTS-13/ VWF ratio this difference was even four-fold between levels <25th percentile and >75th percentile, respectively. VWF remained an independent predictor of VTE occurrence in cancer patients even after adjustment for patient-and cancer-related factors in the first multivariable model or adjustment for previously validated biomarkers of cancer-associated VTE 3 in the second model. As the impact of VWF levels on risk of VTE was time-dependent with the highest HR at baseline, its potential importance for prediction of cancer-associated VTE at cancer diagnosis is emphasized. Overall, this may identify VWF as a possible laboratory tool for risk stratifi- Therefore, it may be hypothesized that ADAMTS-13 does influence the predictive potential of the ratio for cancer-associated VTE, particularly concerning the stratification of low-risk patients.
As a mechanistic explanation of this association it was hypothesized that cancer-related VTE may be based on VWF-mediated platelet aggregation. 38 Bauer et al 39 recently demonstrated in vitro that melanoma cells can activate vascular endothelial cells and prompt them to release ULVWF which is followed by platelet aggregation.
They further showed that the combination of VWF release and decreased local ADAMTS-13 in the tumor tissue is likely to cause a procoagulatory milieu. After infusion of recombinant ADAMTS-13 (rADAMTS-13) the formation of ULVWF networks and platelet aggregation was reduced. The potential benefit of rADAMTS-13 for inhibiting thrombus growth and down-regulating platelet adhesion to the endothelium was also established in other experimental studies. [40][41][42] An approach via rADAMTS-13 would be advantageous, as it cleaves only the highly active forms of the protein, leaving basal levels of VWF in the circulation. 41 For a possible future outlook, in a recent review of therapies for TTP there was also promising data presented for both, rADAMTS-13, which showed good tolerance in a phase 1 clinical trial, and for caplacizumab which is a nanobody that inhibits ULVWF mediated platelet aggregation under high shear rates, 43 therefore theoretically also not interfering with basal VWF function.
One main point up for debate is the relationship and the (individual) cause and effect of either VWF and/or FVIII on thrombosis incidence, as one study found VWF to be an independent risk factor, 27 whereas another found the correlation of VWF with F I G U R E 3 Hazard ratio for mortality according to the relationship between ADAMTS-13 and VWF adjusted for categories age, sex, and cancer type/stage. Numerical data in Table 4, number of patients for respective categories in Table 5.
Patients with values of VWF >75th percentile and concomitantly low (<25th percentile) or average (25-75th percentile) ADAMTS-13 values had the highest probability of mortality (HR 4.31 and 4.75, respectively). The negative impact of high VWF on survival probability seems most prominent, but a high ADAMTS-13 also shows protective tendencies. Patients who had values of ADAMTS-13 > 75th percentile and VWF <25th percentile and were used as reference (set as HR = 1). *P < 0.05, † P < 0.01, ‡ P < 0.001. HR, hazard ratio; VWF, von Willebrand factor VTE to be largely explained by FVIII in multivariable analysis. 23 Both studies were performed in non-cancer patients. VWF is the carrier molecule of FVIII in the circulation and also the main determinant of FVIII plasma levels. 44 Because of the strong association between their values, we deliberately decided not to include FVIII in the multivariable models, as this causes both factors to mathematically nearly cancel out each other's predictive potential.
Their close molecular relationship is also reflected by the strong correlation we found, so that a clear distinction regarding the individual cause and effect of VWF and FVIII on thrombosis incidence appears to be difficult at present. So the "chicken-or-egg causality dilemma" remains unanswered.
In the present cohort of patients with cancer ADAMTS-13 was lowest in patients with lung, colorectal, and metastatic cancers but within the normal range in most individuals. Particularly high VWF levels were observed in pancreatic, lung, brain, stomach, and colorectal cancer patients and in those with distant metastases. This is in accordance with the literature, which frequently describes lower ADAMTS-13 and higher VWF levels in cancer patients, often also in a stage-dependent intensity. 9 One limitation of our study was that only VWF antigen was tested, but not VWF activity. The amount of frozen plasma and the dimension of the study allowed for only one factor to be evaluated. However, in most cases VWF antigen and activity levels are highly correlated (r = 0.82), as recently demonstrated in a study with 432 patients, indicating the impact on the results are thus likely to be limited. 59 Furthermore, patients were not tested for ADAMTS-13 antibodies. Therefore, it cannot be excluded that the few patients with severe to moderate ADAMTS-13 deficiency might have been positive for an inhibitor directed against the protease. Another limitation is the statistical modelling. Because of the comparably smaller amount of events in the competing risk analyses two separate models had to be devised. Also, in regard to the heterogeneity of our study population, cancer type and stage could not be evaluated in all subcategories. The design of dividing the models into four groups tries to address these limitations. The principal strength of our study, however, lies in the large number of diverse cancer patients studied for ADAMTS-13 and VWF and in its long follow-up period.
In conclusion, in this study we could demonstrate that levels of