The impact of ABO blood type on the prevalence of portal vein thrombosis in patients with advanced chronic liver disease

Abstract Background and aims Non‐O blood type (BT) is a risk factor for thromboses, which has been attributed to its effects on von Willebrand factor (VWF)/factor VIII (FVIII) levels. Although high VWF/FVIII may be risk factors for portal vein thrombosis (PVT) in patients with advanced chronic liver disease (ACLD), the impact of BT on PVT is unknown. We aimed to assess (I) whether non‐O‐BT is a risk factor for PVT and (II) whether non‐O‐BT impacts VWF/factor VIII in patients with ACLD. Methods Retrospective analysis comprising two cohorts: (I) “US” including all adult liver transplantations in the US in the MELD era and (II) “Vienna” comprising patients with a hepatic venous pressure gradient (HVPG) ≥6 mmHg. Results (I) The “US cohort” included 84 947 patients (non‐O: 55.43%). The prevalence of PVT at the time of listing (4.37% vs 4.56%; P = .1762) and at liver transplantation (9.56% vs 9.33%; P = .2546) was similar in patients with O‐ and non‐O‐BT. (II) 411 patients were included in the “Vienna cohort” (non‐O: 64%). Mean HVPG was 18(9) mmHg and 90% had an HVPG ≥10 mmHg. Patients with non‐O‐BT had slightly increased VWF levels (318(164)% vs 309(176)%; P = .048; increase of 23.8%‐23.9% in adjusted analyses), but this difference was driven by patients with less advanced disease. However, non‐O‐BT explained only 1% of the variation in VWF and had no effect on FVIII. Conclusions Although non‐O‐BT impacts VWF in patients with early stage ACLD, its contribution to VWF variation is considerably smaller than in the general population. Moreover, non‐O‐BT had no impact on FVIII. These findings may explain the absence of an association between non‐O‐BT and PVT in patients with advanced cirrhosis.


| INTRODUC TI ON
As a consequence of abnormal routine coagulation tests and thrombocytopenia, cirrhosis has long been considered an acquired bleeding disorder. 1 In fact, the liver plays a central role in coagulation and plasma levels of most procoagulant factors are significantly reduced in patients with cirrhosis. However, these changes are balanced by decreased levels of anticoagulant proteins 2 and highly elevated levels of the platelet adhesive protein von Willebrand factor (VWF). 3 Therefore, patients with cirrhosis are nowadays considered to have a rebalanced haemostatic equilibrium. 4 However, when compared to liver-healthy subjects, this equilibrium seems to be instable and easily tips in one direction, which may lead to bleeding or thrombosis. 5,6 However, while clinically relevant non-portal hypertensive bleeding is rare, the prevalence/incidence of portal vein thrombosis (PVT) in patients with advanced cirrhosis is considerably high. 7 PVT is a clinically relevant complication in patients with cirrhosis, 7 especially in the liver transplant waiting list setting. While there is an ongoing debate regarding a potential causal relationship between PVT and deterioration of liver function, 8 extensive PVT has been shown to worsen transplant outcomes in an analysis based on the OPTN/ UNOS data 9 or may even preclude liver transplantation. Moreover, the risk of venous thromboembolism (VTE; ie, deep vein thrombosis and pulmonary embolism) is significantly increased in patients with cirrhosis. 10 As a result, cirrhosis may even be considered as a prothrombotic condition, with thrombophilic changes such as elevated VWF/factor VIII levels potentially contributing to the high risk of thrombotic events. 4 The multimeric glycoprotein VWF is released from endothelial cells upon activation and is cleaved by the ADAMTS13 protease into smaller, haemostatically less potent multimers. 11 Apart from portal hypertension, 12 several other factors influencing VWF levels have been identified. For instance, in the general population/ studies not focusing on chronic liver disease (CLD), VWF increases with age 13 or in the presence of the metabolic syndrome (MetS). 14 Moreover, VWF levels are determined by ABO blood type. VWF and factor VIII levels are about 25% higher in non-O individuals, as compared to O individuals, 15 which translates into a clinically significantly increased risk of VTE (eg, incidence rate ratio of 1.8 and population attributable risk of 32%), 15,16 and to a smaller extent, arterial thrombosis (in particular, myocardial infarction: incidence rate ratio of 1.1 and population attributable risk of 6%). 16,17 Moreover, some evidence links VWF/factor VIII levels with incident PVT in patients with cirrhosis. 18 However, in patients with clinically significant portal hypertension (CSPH), VWF levels are highly elevated, most likely as a result of portal hypertension 19 and bacterial translocation-induced inflammation. 12 Accordingly, it is unclear if, and to what extent, ABO blood type impacts on VWF/factor VIII levels in patients with cirrhosis.
Furthermore, the effect of ABO blood type on the development of PVT, the most common thrombotic event in patients with cirrhosis, has yet to be studied. Prevention of PVT development with prophylactic enoxaparin seems to be effective and safe in patients awaiting liver transplantation. 20 Identification of patients at particularly high risk for PVT may facilitate patient selection for future studies on prophylactic anticoagulation. Therefore, we aimed to analyse the impact of ABO blood type on (I) the development of PVT and (II) VWF/factor VIII levels in patients with advanced chronic liver disease (ACLD).

| Study design and population
This retrospective analysis included two cohorts: were excluded from the data set for the main analyses. Finally, the analyses were repeated after re-including these patients.
Blood type A subgroups (A1, A1B, A2 and A2B) were referred to as blood type A in all analyses.
(II) The "Vienna cohort" comprised prospectively characterised patients with ACLD who underwent hepatic venous pressure gradient (HVPG) measurement at the Medical University of Vienna

Key points
• Non-O blood type is a risk factor for thromboses in the general population.
• This association has been attributed to increased von Willebrand Factor (VWF)/factor VIII levels in subjects with non-O blood type.
• The impact of blood type on portal vein thrombosis (PVT) in patients with advanced chronic liver disease (ACLD) is unknown.
• In this study, the prevalence of PVT was comparable between patients with O and non-O blood type undergoing liver transplantation.
• The contribution of non-O blood type to VWF variation was considerably smaller than in the general population and was limited to patients with early stage ACLD.
• Non-O blood type had no effect on factor VIII levels in ACLD patients.
• These findings may explain the absence of an association between blood type and PVT in advanced cirrhosis. between 01/2006 and 02/2018 and had a HVPG ≥6 mmHg and available information on ABO blood type and plasma VWF levels.
Patients with active bacterial infection or hepatocellular carcinoma (HCC) were excluded. The final cohort comprised 411 patients.

| Evaluation of portal vein thrombosis
In the "US cohort," portal vein thrombosis was evaluated at the time of listing as well as at the time of liver transplantation. The required information was independently reported to the OPTN by each transplant centre as part of the routine transplantation waiting list registry and the transplant event registry.

| HVPG measurement
Hepatic venous pressure gradient was measured in clinical routine for diagnostic or prognostic purposes or HVPG-guided non-selective betablocker therapy, 21 as supported by the Austrian consensus recommendations for the treatment of portal hypertension. [22][23][24] HVPG measurements were performed in the absence of portal pressure-lowering medications (ie, non-selective beta blockers and nitrates) or transjugular intrahepatic portosystemic shunt (TIPS) and according to a standardised protocol, as previously described. 25,26 HVPG values ≥10 mmHg denoted CSPH.

| Assessment of VWF and factor VIII levels
Labouratory tests in the "Vienna cohort" were performed using blood samples obtained at the time of HVPG measurement. VWF antigen levels were measured by a latex agglutination assay (STA LIATEST vWF, Diagnostica Stago), while factor VIII activity was assessed by a one-stage clotting assay using a fully automated CS-5100 coagulation analyser (Sysmex). Plasma protein C and antithrombin activity were measured using chromogenic assays. were performed to evaluate factors independently associated with VWF/factor VIII levels. Variables showing a trend in univariate analysis (P < .1) as well as the factor of interest (non-O blood type) were included into the multivariate models. Multicollinearity was detected by variable inflation factor (VIF). Accordingly, either decompensated cirrhosis and MELD, or Child-Turcotte-Pugh (CTP) score were included as covariates. A P ≤ .05 was considered statistically significant.

| Ethics
The sub-study based on the "Vienna cohort" was conducted in accord-

| Patient characteristics of the "US cohort" according to ABO blood type
Overall, the majority of patients was male (63.2%) with a mean age of 52.55 ± 11.68 years. While 26 481 patients had blood type O, 21 902 had A, 7915 B and 2994 patients harboured blood type AB.
Detailed patient characteristics of the "US cohort" stratified by blood type are displayed in Table S1.
As this study aimed to evaluate differences between O vs non-O individuals, a comparison of baseline characteristics between these two groups is shown in Table 1. While 55.34% of patients (n = 32 811) had non-O blood type, 44.66% (n = 26 481) harboured O blood type. As a result of the large sample size, aetiology of liver disease was statistically significantly different between the two groups (P < .0001), however, the differences were small and judged not to be clinically significant (maximum between group difference: 1.06%). MELD score at the time of liver transplantation was statistically significantly higher in patients with blood type O Importantly, despite being statistically significant, the differences in MELD were only modest. Moreover, patients with blood type O showed a lower prevalence of diabetes mellitus (22.40% vs 23.26%; P = .0132), but again, this difference was not considered clinically significant. Importantly, the proportions of overweight and obese patients as well as other baseline characteristics were well balanced between the two groups.

| Prevalence of PVT at the time of listing for liver transplantation and transplantation in the "US cohort"
At the time of listing for liver transplantation, 4 4.4% vs 4.69% (P = .384) and >20 points: 4.17% vs 4.49% (P = .208).
Finally, we re-included the patients with HCC and/or TIPS, which did not affect our results regarding the impact of ABO blood type on PVT (Tables S2 and S3).

| Patient characteristics of the "Vienna cohort" according to ABO blood type
The majority of patients was male (70%) with a mean age of 54.1 ± 11.4 years ( Table 2). Viral hepatitis (38%) and alcoholic liver disease (ALD; 35%) were the most common aetiologies. While 16%  (Tables S4 and S5). When stratifying by CTP score stage, there was a numerical difference in CTP A patients, however, this trend did not attain statistical significance.
A similar pattern was observed when analyzing factor VIII levels, with trends towards higher factor VIII levels in blood type non-O patients with CTP A, MELD <10 points, or subclinical portal hypertension (HVPG 6-9 mmHg). However, these trends did not attain statistical significance. Of note, factor VIII values were very similar between O and non-O patients with more liver disease/ portal hypertension.

| Adjusted and unadjusted analyses on factors associated with VWF levels in the "Vienna cohort"
In unadjusted analysis, VWF levels showed a positive association with age, ALD as underlying aetiology, HVPG and factors closely related to portal hypertension (ie, varices and hepatic decompensation) and indicators of hepatic dysfunction (MELD and CTP score) ( Table 3). In addition, VWF was linked with liver enzyme levels (as-  VWF levels across HVPG strata as well as MELD and CRP terciles are depicted in Figure 3.

| Adjusted and unadjusted analyses on factors associated with factor VIII levels in the "Vienna cohort"
Factor VIII levels showed

| Correlation between VWF and factor VIII levels in the "Vienna cohort"
Von Willebrand factor and factor VIII showed a correlation of moderate strength (ρ = 0.466; P < .001) (

| Comparison of patient characteristics between the "US cohort" and the "Vienna cohort"
After excluding patients with HCC and/or TIPS, the male predominance was less pronounced in the "US cohort," as compared to the "Vienna cohort," which may be related to the lower prevalence of ALD in the "US cohort" (Table S7). In contrast, patients in the "US cohort" were more commonly overweight or obese. The most striking difference was the higher severity of underlying hepatic dysfunction in the "US cohort," as evidenced by a MELD score at the time of listing (20.7 ± 9.6 points), which was considerably higher than the MELD at the time of HVPG measurement in the "Vienna cohort" (11.7 ± 4.0 points; P < .0001). TA B L E 3 Simple (A) and multiple linear regression analysis (B model including decompensated cirrhosis and MELD; C model including CTP score) of factors associated with VWF levels in the "Vienna cohort"

| D ISCUSS I ON
donors, for instance, the risk of VTE was increased by 80% in patients with non-O blood type. 16 In contrast, the association between ABO blood type and PVT has yet to be investigated. In our large cohort of patients undergoing liver transplantation ("US cohort"), the prevalence of PVT at the time of listing as well as at the time of transplantation was similar between O and non-O individuals. Importantly, owing to the enormous sample size of the "US cohort," we can basically rule out type II error, as we were able to assess even very small differences. For instance, a 10% decrease in the prevalence of portal vein thrombosis in patients with O blood type at the time of listing (ie, from 4.48% to 4.032%) would have been detected with a statistical power of more than 99%. Finally, the OPTN/UNOS data have been used in a series of studies reporting information on PVT, [28][29][30][31][32][33][34] and thus, seems to be suitable for the purpose of our study.
The well-documented prothrombotic state observed in liver-healthy non-O subjects is attributed to the impact of ABO blood type on VWF, and consequently, factor VIII levels. 35 Even though the underlying mechanisms have yet to be fully elucidated, 36 the effects of ABO blood type on VWF and factor VIII levels seem to arise from differences in VWF clearance. 37 In-vitro data suggests that non-O blood type impacts the VWF glycosylation pattern and thereby alters the susceptibility of VWF to ADAMTS13-mediated proteolysis. 38 The impact of ABO blood type seemed to increase with age, as patients >55 years showed the highest ABO blood type-related difference in VWF. 13 Accordingly, with a mean age of 54.1 ± 11.4, we would have expected a profound impact of ABO blood type in our second cohort ("Vienna cohort"). In adjusted analysis, the non-O blood type-related increase in VWF antigen levels was about 24%, which, in absolute terms, was comparable to the differences reported in the general population. 15  increased VWF levels in these patients. 39 LSEC dysfunction impacts intrahepatic resistance, and thus, aggravates portal hypertension, which may further increase bacterial translocation. 40 In turn, bacterial translocation worsens endothelial dysfunction via toll-like receptor 4 activation by endotoxins/lipopolysaccharides, 41 leading to a perpetuation of these mechanisms, 42 and, because of the intense crosstalk between parenchymal and non-parenchymal cells in the liver, also hepatic inflammation. 43 Finally, bacterial translocation also induces systemic inflammation and hemodynamic derangements, which may trigger acute decompensation and acute-on-chronic liver failure, 44   We have to acknowledge that the two main analyses included in this study were performed in different cohorts, since the "US cohort" did not include information on VWF/factor VIII, while in the "Vienna cohort," screening for PVT was not standardised. However, the demographic characteristics of the two cohorts were comparable and both cohorts included patients with ACLD, although patients in the "Vienna cohort" had considerably less advanced disease. Importantly, this approach allowed to combine the strengths of the "US cohort" (very large sample size) and the "Vienna cohort" (in depth characterisation). However, the relationship between ABO blood type and PVT was evaluated in a sicker cohort than the association of ABO blood type and VWF/factor VIII levels, which is an important limitation of our study. Stratification by severity of hepatic dysfunction/portal hypertension revealed, that the impact of ABO blood type on VWF (and possibly, factor VIII) levels seemed to be limited to ACLD patients with less advanced disease (ie, patients with CTP A, subclinical portal hypertension, or MELD <10 points). As patients with these characteristics usually aren't listed for liver transplantation, they were not included in the "US cohort," and thus, our findings should not be extrapolated to these patient populations. Of note, these patients with less advanced disease are per se at low risk for PVT. 7 In conclusion, while ABO blood type contributes to the variation in VWF levels in patients with early stage ACLD, its overall impact is considerably smaller than in the general population, and may vanish in patients with advanced cirrhosis. Moreover, ABO blood type had no impact on factor VIII levels. These findings may explain the absence of an association between ABO blood type and PVT, even in a large data set of patients with advanced cirrhosis.

ACK N OWLED G EM ENTS
This work was supported in part by Health Resources and Services Administration contract 234-2005-37011C. The content is the responsibility of the authors alone and does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organisations imply endorsement by the US Government.

CO N FLI C T S O F I NTE R E S T
The authors have nothing to disclose regarding the work under consideration for publication. However, the authors disclose the fol-