Potential conflict of interest: Nothing to report.
von Willebrand factor antigen (vWF-Ag) is elevated in patients with liver cirrhosis, but the clinical significance is unclear. We hypothesized that vWF-Ag levels may correlate with portal pressure, measured by hepatic venous pressure gradient (HVPG), and predict clinically significant portal hypertension (CSPH; HVPG ≥10 mmHg), decompensation and mortality. Portal hemodynamics were assessed by HVPG measurement, whereas vWF-Ag levels were measured by enzyme-linked immunosorbent assay. During follow-up, complications of liver cirrhosis, death or transplantation were recorded. Two hundred and eighty-six patients (205 male and 81 female; mean age, 56 years) with liver cirrhosis were included. vWF-Ag correlated with HVPG (r = 0.69; P < 0.0001) and predicted CSPH independently of Child Pugh score. Higher vWF-Ag levels were associated with varices (odds ratio [OR] = 3.27; P < 0.001), ascites (OR = 3.93; P < 0.001) and mortality (hazard ratio: 4.41; P < 0.001). Using a vWF-Ag cut-off value of ≥241%, the AUC for detection of CSPH in compensated patients was 0.85, with a positive predictive value and negative predictive value of 87% and 80%, respectively. Compensated patients had 25% mortality after 53 months if the vWF-Ag was <315% compared to 15 months in patients with vWF-Ag >315% (P < 0.001). Decompensated patients had a mortality of 25% after 37 and 7 months if their vWF-Ag was <315% and >315%, respectively (P = 0.002). In compensated patients with a vWF-Ag >315% median time to decompensation or death was 32 months compared with 59 months in patients with vWF-Ag <315%. vWF-Ag equals Model for End-Stage Liver Disease (MELD) in mortality prediction (area under the curve [AUC] = 0.71 for vWF-Ag versus AUC = 0.65 for MELD; P = 0.2). Conclusion: vWF-Ag is a new, simple and noninvasive predictor of CSPH. A vWF-Ag cut–off value at 315% can clearly stratify patients with compensated and decompensated liver cirrhosis in two groups with completely different survival. vWF-Ag may become a valuable marker for the prediction of mortality in patients with liver cirrhosis in clinical practice. (HEPATOLOGY 2012)
Portal hypertension (PH) accounts for the major complications of liver cirrhosis, such as ascites, variceal hemorrhage and decompensation. Early diagnosis of PH is essential for the management of patients with cirrhosis. In previous studies, it has been shown that early diagnosis, leading to adequate treatment, can significantly reduce the mortality rate of PH-related complications.1, 2 Recent guidelines recommend the diagnosis of PH by the measurement of hepatic venous pressure gradient (HVPG).3 Clinically significant portal hypertension (CSPH; HVPG ≥10 mmHg) is associated with a higher risk of liver-related mortality, development of varices, and other PH-related complications. An HVPG ≥12 mmHg is associated with a higher risk of bleeding from varices.1 Measurement of HVPG is an invasive procedure and is only available in specialized centers. Noninvasive markers could be a clear advantage for the management of patients with cirrhosis, but none of the markers investigated, so far, have shown satisfactory specificity and sensitivity to enter clinical routine. More recently, transient elastography (TE) was described as a noninvasive tool for the diagnosis of PH in patients with liver cirrhosis,4 but the costs and availability of TE represent limiting factors in smaller hospitals. Thus, the recent Baveno V consensus conference on PH recommended to investigate and identify further noninvasive markers for PH.3 Hepatic decompensation is the most important predictor of prognosis and mortality in patients with liver cirrhosis, with several precipitating factors contributing to the first event of decompensation.5
Endothelial dysfunction is considered as an important determinant of the increased intrahepatic vascular resistance in cirrhotic livers.6, 7 von Willebrand factor antigen (vWF-Ag) is released by activated endothelial cells (ECs) and therefore represents an indicator of EC activation8 and plays a crucial role in high shear stress, depending on primary hemostasis. Furthermore, in patients with liver cirrhosis, elevated levels of vWF-Ag are frequently observed.9 vWF-Ag levels were also shown to be an independent risk factor of myocardial infarction and mortality in patients with angina pectoris.10, 11 Although it is established that VWF-Ag is increased in patients with cirrhosis, no data on the association of vWF-Ag and portal pressure exist. One recent study describes a correlation between vWF-Ag and HVPG in patients with CSPH, but patients without PH were not included, and thus the diagnostic power of vWF-Ag for CSPH could not be evaluated.12
Because vWF-Ag plays a crucial role in primary hemostasis and is an indicator of endothelial activation and development of thrombotic vascular obliteration, which are all discussed as possible mechanisms leading to PH,13 we hypothesized that patients with CSPH have increased vWF-Ag levels, compared to patients without CSPH. Thus, the aims of our study were (1) to evaluate the diagnostic performance of vWF-Ag to detect clinically significant PH defined by HVPG, compared to TE in patients with compensated liver cirrhosis (i.e., when CSPH is not clinically evident), and (2) to evaluate vWF-Ag levels in the prediction of mortality and decompensation in patients with liver cirrhosis.
Patients referred to the hepatic hemodynamic laboratory at the Department of Internal Medicine III, Division of Gastroenterology, Medical University of Vienna (Vienna, Austria) were included between September 2006 and December 2009. All patients with liver cirrhosis scheduled for baseline HVPG measurements were included. Liver cirrhosis was diagnosed histologically, clinically, or by typical radiological findings. Exclusion criteria were presence of pre- and posthepatic causes of PH, severe cardiopulmonary or renal impairment, active infections, diabetes, anticoagulant therapy, antiplatelet drugs, as well as current treatment with beta-blockers, statins, or interferon (IFN).14, 15 Patients with alcoholic liver disease had to be abstinent from alcohol for at least 3 months. Etiology of liver disease, age, HVPG, medical history, including the presence of esophageal varices, ascites, Child Pugh score (CPS), hematological status, including vWF-Ag, clinical chemistry, and liver stiffness (measured by FibroScan; Echosens, Paris, France) were recorded for each patient at the day of HVPG measurement. The study was approved by the local ethics committee and was conducted according to the principles of the Declaration of Helsinki.
Plasma levels of vWF-Ag were measured as previously described14 using a fully automated STA analyzer and vWF-Liatest (Diagnostica Stago, Paris, France).
Measurement of HVPG.
Portal pressure was evaluated by measurement of HVPG according to international standards, as described previously.16, 17 At least three repeated measurements of free and wedged hepatic vein pressure were performed to calculate the HVPG. Continuous tracing of pressure curves were electronically recorded using a pressure transducer and S5 collect software. Normal portal pressure was defined as an HVPG of 1-5 mmHg, whereas elevated portal pressure defined as an HVPG of 6-9 mmHg. CSPH was diagnosed at an HVPG ≥10 mmHg, and severe CSPH was diagnosed at an HVPG ≥12 mmHg. All measurements were performed by two hepatologists, each with a personal experience of more than 500 HVPG measurements.
Measurement of liver stiffness was performed by transient elastography (FibroScan; Echosens) after an overnight fast, as previously described in detail.17 Results of liver stiffness were considered as adequate if the interquartile range (IQR) was within the 30% interval of the median value and if the success rate was ≥70%. Results of the median value and IQR were recorded in kPa.
Follow-up and Mortality.
Patients were followed prospectively at least every 6 months at the outpatient clinic of the Medical University of Vienna until December 2011. During follow-up, all events, especially decompensation by ascites, jaundice, grade 3/4 hepatic encephalopathy, variceal bleeding, death, and liver transplantation (LT), were recorded. Because many of our patients were from foreign origin (mostly from Turkey and former Yugoslavia), we were not able to prevent all study participants being lost to follow-up because of to remigration. However, if a patient was not seen at our outpatient department within the preceding 6 months, telephone contact (to the subject or relatives or to their primary care physicians) was additionally established to check on the patients' status. Additionally, the national register of death was screened, including screening for the specific cause of death (according to the International Statistical Classification of Diseases and Related Health Problems 10th Revision codes). Transplanted patients (n = 24) were censored at the time of transplantation in Kaplan-Meier;s analysis and Cox's regression. Thirty patients (10.4%) were lost to follow-up.
Statistical analyses were performed using SPSS 19.0 (SPSS, Inc., Chicago, IL). Descriptive statistics are provided as median and IQR. Differences between groups with and without PH were assessed by Mann-Whitney's U test. Correlation of portal pressure (i.e., HVPG) and vWF-Ag were assessed by Spearman's correlation and expressed by Spearman's correlation coefficient. Univariate regression analysis was performed to identify a relation between vWF-Ag and PH and its clinical consequences. Receiver operating characteristic (ROC) curves were created for the assessment of the predictive value of vWF-Ag and TE for PH and mortality, including the area under the curve (AUC), sensitivity, specificity, positive predictive value (PPV), and negative predictable value (NPV) calculation. PPV was defined as the likelihood of CSPH; NPV was defined as the likelihood of having HVPG levels below 10 mmHg. The value with the best sensitivity and specificity in AUC analysis (Youden's Index) was chosen for further analyses. AUCs were compared using Hanely and McNeil's approach.18 Independence of predictive factors was assessed by multivariate binary logistic regression.
Time-dependent variables were analyzed using Kaplan-Meier's method and compared by the log-rank test; patients were censored at the time of liver transplantation. In the case of a comparison of more than one group, Shaffer's correction was applied to the P values. Cox's multivariable proportional hazards models were applied, and results of Cox's models are presented as the hazard ratio (HR) and 95% confidence intervals (CIs). We assessed the overall model fit using Cox-Snell's residuals. Furthermore, we tested the proportional hazard assumption for all covariates using Schoenfeld's residuals (overall test) and Schoenfeld's scaled residuals (variable-by-variable testing). According to the tests, the proportional hazards assumption was not violated.
Because transient elastography was unsuccessful in 25% of cases, we calculated ROC curves with the intention-to-diagnose approach (AUC-ITD),19 including all liver stiffness results, regardless of success in the AUC analysis. All P values reported are two-sided, and P values <0.05 are considered significant.
Two hundred and eighty-six patients with liver cirrhosis were included. Two hundred and one males and 65 females were included in the study with a median age of 55 years (IQR, 48-62) and median body mass index was 26.1 (range, 23.2-29.7).
One hundred and forty-eight patients (51.7%) were classified as Child Pugh A, 104 (36.4%) as Child Pugh B and 34 (11.9%) as Child Pugh C. Two hundred and eleven patients (73.8%) had HVPG >=10 mmHg, of which 190 (66.4%) had HVPG >=12 mmHg. Fifty-one (17.8%) patients had hepatocellular carcinoma (HCC). vWF-Ag levels were similar in patients with and without HCC (mean vWF-Ag 342% [IQR 293.4%-391.1%] versus 323.6% [IQR 305.2%-342.0%]; P > 0.05).
vWF-Ag in Liver Cirrhosis.
vWF-Ag levels were increasing with Child Pugh stage: In patients with Child A vWF-Ag was 240% (IQR 181%-325%), in Child B 350% (IQR 288%-435%), and in Child C 452% (IQR 353%-594%) (Table 1). Median vWF-Ag levels were significantly lower in the 189 compensated, compared to 97 decompensated patients (P < 0.001).
Table 1. Patients Characteristics and vWF-Ag Values and Clinical Course of Liver Cirrhosis and PH*
HVPG <10 mmHg
HVPG ≥10 mmHg
Median ± IQR.
vWF-Ag and PH.
vWF-Ag was significantly higher in patients with CSPH, compared to patients without CSPH (median 346% [IQR 275%-441%] versus 197% [IQR 158%-228%]; P < 0.001) (Fig. 1). vWF-Ag values were higher in patients with esophageal varices (P < 0.0008) and history of ascites (P < 0.0001), compared to patients without. Higher vWF-Ag levels were significantly associated with varices (OR = 3.27; P < 0.001) and ascites (OR = 3.93; P < 0.001).
There was a significant difference of vWF-Ag between patients with and without CSPH within the CPS stages. In CPS A, median vWF in CSPH was 302% (IQR 242%-364%), compared to a median vWF of 195% (IQR 158%-226%) (P < 0.001) in CPS A patients without CSPH. Similarly, in CPS B patients, median vWF-Ag was significantly higher in patients with CSPH than in patients without CSPH (367% [IQR 299%-454%] versus 205% [IQR 162%-283%]; P < 0.001). All CPS C patients had CSPH.
vWF and HVPG values correlated significantly (r = 0.643, P < 0.001). Linear regression showed an increase of HVPG values of 2.9 mmHg per increase of vWF-Ag level of 100 points (P < 0.0001). AUC for the diagnosis of CSPH was 0.884 (CI: 0.841-0.928) and 0.88 (CI: 0.84-0.92) for the diagnosis of severe PH (HVPG ≥12 mmHg) (Table 2). A cut-off value of 241% provided optimal sensitivity and specificity to discriminate between patients with and without CSPH.
Table 2. Diagnostic Performance of vWF-Ag and MELD for Mortality and of vWF-Ag and TE for HVPG >=10 mmHg in All Patients (n = 286) and for HVPG >=10 mmHg, HVPG >=12 mmHg, and Decompensation in Compensated Cirrhotic Patients (N = 145)
Relationship Between vWF-Ag and HVPG in Compensated Cirrhosis.
Among compensated patients with CSPH, vWF-Ag levels were significantly higher compared to patients without CSPH (median 323% [IQR 251%-389%] versus median 197% [IQR 158%-228%]; P < 0.001) (Figs. 2 and 3). Furthermore, in compensated patients, vWF and HVPG values correlated significantly (Spearman's r = 0.660; P < 0.001).
AUC for the diagnosis of CSPH in compensated patients was 0.850 (CI: 0.793-0.907) for vWF-Ag, and AUC for the diagnosis of severe PH in compensated patients was 0.847 (CI: 0.789-0.905) (Table 2). A cut-off value of 241% yielded the most accurate sensitivity and specificity to discriminate patients with and without CSPH. We further found a significant relationship of vWF-Ag and HVPG. Linear regression showed an increase of HVPG values of 3.3 mmHg per increase of vWF-Ag level of 100 points (P < 0.0001).
In univariate analysis CPS, vWF-Ag, platelets and liver stiffness were significantly associated with CSPH. In multivariate linear regression, considering significant predictors identified at univariate analysis, vWF-Ag (P < 0.001), platelets (P < 0.001) and stiffness (P < 0.001) were independently associated with CSPH.
When categorizing patients according to vWF-Ag levels below or above 241%, OR for the presence of CSPH was 4.17 (P = 0.017; corrected for liver stiffness and platelet count).
Liver Stiffness and PH.
Using Liver stiffness measured by TE for the diagnosis of CSPH resulted in an AUC of 0.907 (CI: 0.855-0.960) and in an AUC of 0.886 (CI: 0.829-0.943) for the diagnosis of severe PH in compensated cirrhosis (Table 2). Comparing the AUC in patients with valid TE measurement and vWF-Ag, there was no significant difference in the AUCs (P = 0.08). Because TE was unsuccessful in 41 of 128 of compensated patients (mainly because of obesity), we calculated ROC curves with the ITD approach. AUC-ITD for TE—including all patients regardless of success of TE—was 0.79 (CI: 0.71-0.86).
To compare the predictive power of TE and vWF-Ag for noninvasive diagnosis of CSPH and severe PH, a comparison of AUCs (AUC-ITD for TE and AUC for vWF-Ag) was computed, resulting in a similar performance of vWF-Ag if the unsuccessful cases of TE were included (P = 0.135).
vWF-Ag for Prediction of Mortality in Patients With Liver Cirrhosis.
The median follow-up time was 33 months (95% CI: 30-36 months). Overall, 91 patients died during follow-up. Overall median transplant-free survival time was 59 months. Median vWF-Ag was significantly higher in patients who died during follow-up period, compared to patients who were still alive at the end of follow-up (387% [IQR 288%-457%] versus 271% [IQR 200%-358%]; P < 0.0001).
The adjusted HR for mortality was 4.1% per increase of vWF-Ag of 10% points. At a cutoff of vWF-Ag of 315% for mortality, the HR was 3.37 (95% CI: 2.21-5.15). This cutoff also predicted mortality (HR: 2.92; 95% CI: 1.72-4.97) (P < 0.001) independently of CPS, MELD, HCC, and HVPG in multivariate analysis.
In patients without HCC (n = 235), the HR ratio (univariate) for mortality in case of vWF-Ag >315% was 4.42 (95% CI: 2.5-7.7; P < 0.001), whereas in multivariate analysis, considering CPS, MELD, HVPG, and vWF-Ag >315% as a categorical variable, the HR was 3.49 (95% CI: 1.8-6.6; P < 0.001).
vWF-Ag equals MELD in mortality prediction (AUC = 0.71 [95% CI: 0.65-0.77] versus AUC = 0.65 [95% CI: 0.58-0.72]; P = 0.197) (Fig. 4). Considering only patients with CSPH, AUC of vWF-Ag was not significantly different, compared to MELD (0.690 [95% CI: 0.618-0.763] versus 0.620 [95% CI: 0.540-0.699]; P = 0.222).
vWF-Ag and Mortality in Compensated and Decompensated Cirrhosis.
Twenty-five percent mortality level was reached after 53 months if the baseline vWF-Ag was <315%, compared to 9 months in patients with vWF-Ag >315% (P < 0.001) (Fig. 5A–C).
Compensated patients had 25% mortality after 53 months if the baseline vWF-Ag was <315%, compared to 15 months in patients with vWF-Ag >315% (P < 0.001). Decompensated patients had a mortality of 25% after 37 and 7 months if their vWF Ag was <315% and if vWF-Ag was >315%, respectively (P = 0.002).
In compensated patients with a vWF-Ag value <315%, median time to decompensation or death was 59 months, compared to 32 months in patients with vWF-Ag levels >315%.
This prospective study demonstrates the value of vWF-Ag as a novel, noninvasive marker in patients with liver cirrhosis. We could show a clear correlation to PH assessed by the gold standard, HVPG, in cirrhosis and its clinical consequences. In addition, our data suggest that vWF-Ag may be a valuable noninvasive marker for the prediction of mortality in compensated and decompensated patients, independent of established models, such as the CPS or MELD.
The diagnostic performance of a vWF-Ag cut-off value at 241% for noninvasive diagnosis of CSPH in compensated and decompensated patients was excellent, as shown by accurate PPV (92%) and NPV (76%). The data clearly demonstrate that a linear increase of vWF-Ag elevates the risk for CSPH and severe PH, as well as associated complications, demonstrated by the linear regression showing an increase of HVPG of 3.3 mmHg per increase of vWF-Ag level of 100. This finding may have a profound effect on the clinical management of patients with cirrhosis by allowing further risk stratification.
Most interestingly, vWF-Ag level (notably as a single parameter) was significantly associated with mortality, and the predictive performance was similar to MELD. vWF-Ag levels >315% seem to identify a risk group of higher mortality and add prognostic information on top of MELD. Furthermore, MELD has to be calculated with a cumbersome formula, which is not feasible at the bedside. In this context, further studies are warranted to assess whether the addition of vWF-Ag to MELD may improve the prediction of short- or long-term mortality.
The most important finding of our study is that a vWF-Ag cutoff at 315% can clearly stratify patients with compensated and decompensated liver cirrhosis in two groups with completely different survival. Mortality rates in compensated patients were significantly lower if vWF-Ag was <315%, with similar results for decompensated patients (Fig. 5B). It may be attractive to speculate on whether vWF-Ag may help to select the optimal point in time for listing for LT or initiation of alternative treatment options in both patients with compensated and decompensated cirrhosis.
A universal explanation on the pathophysiologic mechanisms of elevated vWF-Ag in patients with cirrhosis cannot be provided by our data. Thrombotic risk factors in patients with chronic viral hepatitis are associated with more advanced fibrosis, and platelets themselves seem to play a role in promoting liver injury in the last years.20, 21 However, the elevated levels of vWF-Ag in cirrhosis may be a consequence of endothelial perturbation, caused by increased shear stress, bacterial infection,22 or induction of the synthesis of vWF-Ag in the cirrhotic liver itself.23 Reduced clearance of vWF-Ag resulting from decreased expression or activity of ADAMTS13 (vWF-Ag cleaving protease)9 may further increase vWF-Ag levels in patients with cirrhosis with PH. Effects of antiportal hypertensive therapy, such as nonselective beta-blockers or transjugular intrahepatic portosystemic shunt (TIPS) implantation, on vWF-Ag levels will provide further mechanistic insights in the mechanism regulating vWF-Ag levels in patients with cirrhosis and PH.
vWF-Ag levels can be easily determined in standard laboratories at a cost of less than 6 US$€ per patient sample. To increase the probability of a proper testing, it is advisable that the measurement should be done at a facility with immediate on-site processing in their own specialized coagulation laboratory,24 which is not always found in small hospitals. However, certain known limitations to the application of vWF-Ag levels in patients with cirrhosis have to be considerd (e.g., infections, malignancies, physical training or IFN therapy),14, 15 which have been shown to elevate vWF-Ag levels. Hereditary vWF-Ag deficiency or acute bleeding may diminish vWF-Ag levels and the degree of PH might be underestimated. On the other hand, measurement of HVPG is invasive, expensive, not widely available and technically not successful in up to 4% of patients.25 TE, investigated as another promising noninvasive tool for the assessment of patients with liver disease, may not be successful in up to 25% of cases, not to mention the cost of the system and maintenance, and is therefore inferior when standard probes are used.
Clinical consequences of cirrhosis are foremost related to CSPH more than to any other cause,26 which prompted the proposal of a new staging system for patients with cirrhosis.5 The invasiveness and lack of general availability of HVPG measurement prevents the broad use of pressure-guided diagnostic and therapeutic algorithms in patients with cirrhosis. In our large cohort, we could show an impressive correlation between portal pressure and vWF-Ag levels, which is independent of CPS.8 Thus, vWF-Ag can be used for the selection of high-risk patients within respective Child Pugh stages. This is of particular importance in patients with CPS A and B, who might not be considered for liver transplantation. We additionally could show that the reported increase of vWF-Ag with higher CPS stages9 is probably more related to PH, because patients with cirrhosis without PH had only slightly elevated vWF-Ag levels. Furthermore, we demonstrated a correlation of vWF-Ag with clinical outcome parameters and a high predictive value of disease-related mortality. In line with our results, La Mura et al.12 investigated the effect of vWF-Ag levels on clinical outcome in 42 patients with cirrhosis and PH. The investigators reported a vWF-Ag cut-off value of 216 U/dL to disclose between patients with cirrhosis with a highly different probability of survival free of PH-related events and transplantation. This cut-off level is similar to our 241%, which represents the optimal cutoff to discriminate between the presence or absence of CSPH in patients with cirrhosis. The study by La Mura et al.12 included only patients with CSPH, which precludes the exploration of the role of vWF-Ag for noninvasive prediction of CSPH in patients with cirrhosis. In addition, no data on overall mortality were reported.
Our findings could open up new strategies for the clinical management of patients with cirrhosis: vWF-Ag levels could represent a noninvasive marker of CSPH and could be used to predict survival of patients with liver cirrhosis independent of CPS stages. It could be used to select suitable patients for procedures, such as early TIPS implantation or prophylaxis of spontaneous bacterial peritonitis. Because all our patients with cirrhosis were treated with indicated prophylactic as well as therapeutic regimens (e.g., beta-blockers and/or band ligation), hard endpoints, such as variceal bleeding, were very rare and were not separately assessed. Thus, the prognostic value of vWF-Ag for the occurrence of cirrhosis-related complications, such as variceal bleeding, hepatorenal syndrome, or spontaneous bacterial peritonitis, needs to be confirmed in multicenter trials.
In conclusion, the measurement of vWF-Ag represents a valuable, accessible, and affordable noninvasive predictor of CSPH and mortality in compensated and decompensated liver cirrhosis. It has the potential to enter clinically relevant diagnostic and therapeutic algorithms for patients with cirrhosis. Further prospective studies on the prognostic value of vWF-Ag levels are warranted to assess their role in the potential risk stratification of patients with cirrhosis with PH.