Thrombosis and anticoagulation in liver disease


  • Dominique Charles Valla

    Corresponding author
    1. Service d'Hépatologie, Pôle des Maladies de l'Appareil Digestif, Hôpital Beaujon, Assistance Publique–Hôpitaux de Paris, Université Denis Diderot-Paris 7, and INSERM U773-CRB3, Clichy, France
    • Hépatologie, Hôpital Beaujon, 92118 Clichy, France
    Search for more papers by this author
    • fax: (33) 1-40-87-44-26.

  • Potential conflict of interest: Nothing to report.

Thrombosis and inflammation are closely related mechanisms. They are required to cope with various forms of aggression, but can produce damage when they occur or persist inappropriately. Multiple pathways normally ensure that the hemostatic process is properly triggered and controlled. An imbalance between the pathways that promote, and those that limit, thrombosis may produce a prothrombotic state in which the risk of thrombosis is increased. Thrombosis occurring in the vessels of a previously healthy liver produces so-called primary vascular diseases of the liver. A related issue is that of thrombosis occurring in the hepatic vessels of patients with a primary parenchymal liver disease. The purpose of this article is to discuss the role of prothrombotic disorders and anticoagulation therapy in these 2 different types of liver diseases: primary thrombosis of the hepatic vessels and primary parenchymal liver diseases.

The Impact of Liver Disease on the Diagnosis of Prothrombotic Disorders

Thrombosis is a multifactorial process in which local and general prothrombotic factors need to be combined.1 Most commonly identified local factors are malignant or inflammatory foci, venous injury, and mechanical obstruction of a venous segment. General prothrombotic factors include a number of inherited or acquired conditions, as well as external factors.1 When a patient presents with liver disease, whatever its cause, liver damage has induced changes that considerably hamper recognition of some underlying prothrombotic factors. Antithrombin, protein C, and protein S are synthesized by hepatic cells. Therefore, liver dysfunction decreases the levels of these inhibitors. This happens even in patients with portal vein thrombosis (PVT), in which liver damage is inconspicuous.2 In clinical practice, a low level of a coagulation inhibitor can be interpreted as a primary defect only when the levels of coagulation factors are entirely normal. Otherwise, either a familial deficiency can be documented through screening or the issue will remain unsettled. Indeed, molecular testing is made difficult by the multiple possible mutations that can cause a decrease in coagulation inhibitors.

Myeloproliferative diseases (MPDs) can be easily overlooked in patients with portal hypertension. Splenomegaly can be attributed to portal hypertension, and hemodilution and hypersplenism decrease peripheral blood cell counts and thus mask the peripheral blood features of myeloproliferation. Isotopic measurements show an increased plasma volume, explaining why hematocrit remains in the normal range despite an increased total red cell mass.3 Thus, conventional diagnostic criteria based on splenomegaly and peripheral blood cell counts cannot be applied to patients with portal hypertension. The assessment of the V617F mutation of Janus tyrosine kinase 2 in myeloid cells provides an easy means to reach diagnosis.4 In V617F Janus tyrosine kinase 2–negative patients, bone marrow biopsy looking for clusters of dystrophic megacaryocytes and myelofibrosis is needed.3 These 2 tests have replaced the assessment of endogenous erythroid colony formation in vitro, a demanding, nonstandardized technique of limited availability.3

Low levels of anticardiolipin antibodies are found in approximately 20% of patients with chronic liver disease, whatever its cause.5 Therefore, in a patient with liver disease, a diagnosis of antiphospholipid syndrome is best established on the detection of lupus anticoagulant or antibeta-2 glycoprotein I antibodies. Likewise, hyperhomocysteinemia is difficult to interpret as a cause or a consequence of liver disease. This difficulty is not completely overcome with the use of genotyping for homozygous C677T mutation in the methylene tetrahydrofolate reductase gene, because this homozygosity is only 1 of several determinants of hyperhomocysteinemia.1


BCS, Budd-Chiari syndrome; CI, confidence interval; IVC, inferior vena cava; MPD, myeloproliferative diseases; PAR, protease-activated receptor; PVT, portal vein thrombosis; TIPS, transjugular intrahepatic portosystemic shunt.

Primary Vascular Liver Diseases

Thrombosis affecting the extrahepatic portal venous system is the most common of the primary vascular diseases of the liver. Primary thrombosis of the intrahepatic portal venous system, and thrombosis involving the hepatic venous system or the suprahepatic inferior vena cava (IVC), are less common. Thrombosis of the hepatic arterial tree is even rarer, usually related to iatrogenic injury. For further discussion of this last entity, the reader is referred to a recent review.6

Extrahepatic Portal Vein Thrombosis

Extrahepatic PVT has been found at routine necropsy in approximately 1% of Swedish subjects in the 1970s.7 In developing countries, PVT has accounted for up to 30% of cases of portal hypertension in adults.8 From a clinical point of view, PVT consists of 2 different entities, acute PVT and chronic PVT, which represent successive stages of the same disease.8 Their common causes are discussed first and their management separately.

Prothrombotic Disorders.

A local factor can be identified in approximately 30% of patients (Table 1), and a general prothrombotic disorder in 70% (Table 2).8–12 Malignant tumors within the portal venous territory and cirrhosis are the leading local factors for PVT (Table 1).7 In the rest of the patients, the most common local factors for PVT are inflammatory foci in the abdomen and splenectomy.8 The local factor is more commonly recognized at the acute stage of PVT than later on, at the chronic stage. In patients without apparent local factor, minute inflammatory foci escaping imaging recognition might be involved. A general prothrombotic condition is found almost as frequently in patients with and without local factors.8, 10, 12

Table 1. Local Risk Factors for Portal Vein Thrombosis
  1. Abbreviation: TIPS, transjugular intrahepatic portosystemic shunt.

Cancer (without malignant invasion or constriction of the portal venous system)
 Any abdominal organ
Focal inflammatory lesions
 Neonatal omphalitis, umbilical vein catheterization
 Diverticulitis, appendicitis
 Duodenal ulcer
 Tuberculous lymphadenitis
 Crohn's disease, ulcerative colitis
 Cytomegalovirus hepatitis
Injury to the portal venous system
 Colectomy, gastrectomy
 Liver transplantation
 Abdominal trauma
 Surgical portosystemic shunting, TIPS, liver transplantation
 Preserved liver function with precipitating factors (splenectomy, surgical portosystemic shunting, TIPS dysfunction, thrombophilia)
 Advanced disease in the absence of obvious precipitating factors
Table 2. Average Prevalence and Odds Ratio of Underlying Prothrombotic Conditions in Patients with Nonmalignant Noncirrhotic PVT or BCS
  1. Data are averages from the reports reviewed elsewhere8, 43 and from more recent reports.10, 13–15, 36, 73 Abbreviations: OR, odds ratio; NA, not available.

Inherited conditions    
 Antithrombin deficiency2154
 Protein C deficiency205207
 Protein S deficiency20374
 Heterozygous factor V Leiden51.52513
 Heterozygous G20210A prothrombin17772
Acquired conditions    
 V617 JAK2 positive MPD35NA50NA
 Antiphospholipid syndrome10NA10NA
 Behcet's disease05NA
 Other general condition10NA5NA
External factors    
 Oral contraceptives in women30502.4 
Multiple factors including local factors35NA35NA
No factor25NA5NA

A high prevalence of underlying prothrombotic disorders has been consistently shown in patients without cirrhosis or malignant invasion of the portal vein (Table 2).8–15 PVT is related to MPD in approximately one-third of patients when diagnosis is based on sensitive criteria.13–15 This prevalence is much higher than in patients with thrombosis at other sites10, 15 or in healthy controls.16 By contrast, factor V Leiden is not overrepresented as compared with healthy controls. Conversely, G20210A prothrombin gene mutation is frequently found.10 Data on protein C, protein S, or antithrombin deficiency are difficult to interpret. A possible link with oral contraceptives has not been as well documented as for other venous territories.9–11

Acute Portal Vein Thrombosis: Outcome and Role of Anticoagulation.

Acute PVT is characterized by the recent formation of a thrombus, partially or totally occluding the portal vein. For a recent review on clinical and diagnostic aspects see Condat and Valla.8 Besides inducing severe and persistent abdominal pain, acute PVT is a serious condition for 2 reasons. First, when extensive thrombosis involves distal mesenteric veins, intestinal ischemia or infarction can occur. Infarction is probably related to complete obstruction of venous outflow and to a reflex decrease in mesenteric arterial blood flow.17 In the absence of treatment, intestinal perforation, peritonitis, shock, and death of multiorgan failure occur. Second, permanent obstruction caused by chronic PVT causes irreversible portal hypertension. Therefore, goals for treatment of acute PVT are to avoid extension of thrombosis and to recanalize the obstructed veins, which will prevent intestinal infarction and portal hypertension.

Spontaneous recanalization is likely infrequent in untreated patients. However, data on this crucial point are scarce.8 It is advisable to treat the causal factors as soon as possible. Many patients require specific hematological or surgical management, which is beyond the scope of this article. Because anticoagulation therapy is of proven benefit in patients with acute, deep vein thrombosis,18 extrapolation to patients with acute PVT is logical. Several retrospective surveys8, 12 included consecutive patients. Their combined data show that when initiated immediately, 6 months of anticoagulation therapy was associated with complete recanalization in approximately 45% of patients, partial recanalization in 35% of patients, and no recanalization in 25% of patients. Severe complications of anticoagulation therapy were reported in less than 5% of treated patients. These outcomes have recently been confirmed in a prospective study.19 The optimal duration of anticoagulation therapy for acute PVT has not been determined. Complete recanalization can be delayed until the 6th month of anticoagulation therapy. Recanalization appears to be inversely related to the extent of thrombosis. The most recent and large study identified peritoneal fluid detectable at imaging as the only independent predictive factor.19 A panel of international experts has recommended that in patients with acute PVT, anticoagulation be given for at least 3 months, and permanent anticoagulation therapy be considered for patients with permanent prothrombotic conditions.20 In patients with deep vein thrombosis, a lack of complete recanalization indicates a high risk of recurrence after cessation of anticoagulation therapy. It remains to be assessed whether this holds true for patients with acute PVT.12

The experience with other treatment modalities [surgical thrombectomy, systemic or in situ thrombolysis, or transjugular intrahepatic portosystemic shunt (TIPS)] is likewise limited. Available data suggest that these procedures achieve recanalization at a similar rate as anticoagulation alone, but that the risk of major complication is much higher.21

Treatment of intestinal infarction requires emergency laparotomy for resection of the necrotic parts of the gut. The challenge here is to limit the extent of intestinal resection. Anticoagulation therapy appears to improve the survival of patients who undergo surgery.8

Chronic Portal Vein Thrombosis: Outcome and Role of Anticoagulation.

Chronic PVT is also know as portal cavernoma because the occluded portal vein is replaced by a network of hepatopetal collateral veins connecting the patent portion of the vein upstream from the thrombus to the patent portion downstream. Complete occlusion of the portal vein trunk or of its 2 main branches is virtually always associated with portal hypertension. As reviewed elsewhere,8, 12 at present, diagnosis is commonly made after a fortuitous finding of hypersplenism or portal hypertension. Repeated bleeding from portal hypertension is the most common complication, followed by recurrent thrombosis at splanchnic or extrasplanchnic sites, and portal cholangiopathy.8, 12, 13

The challenge in the management of chronic PVT is to cope simultaneously with the risk of gastrointestinal bleeding and the risk of recurrent thrombosis. There have been no controlled studies of beta-adrenergic blockers or endoscopic therapy in such patients. Nevertheless, screening for gastroesophageal varices, and beta-adrenergic blockers or endoscopic therapy for patients with large varices might be carried out for patients with portal cavernoma, as is done for patients with cirrhosis.20 Indeed, retrospective multivariate analyses found that their application reduced the risk of bleeding22 or improved survival,23 regardless of whether anticoagulation was administered. Endoscopic ligation proved superior to sclerotherapy in children, but data are lacking in adults. For primary prophylaxis of variceal bleeding, there are insufficient data on whether beta-blockers or endoscopic therapy should be preferred. Contrasting findings have been reported regarding the feasibility and the outcome of surgical portosystemic shunting, which suggests differences in patient selection or referral.24 Data on splenectomy and devascularization are limited. There are anecdotal reports of successful TIPS insertion in patients with a portal cavernoma in the absence of cirrhosis.25 Superior mesenteric to left portal vein bypass has shown encouraging results in children.26 Its place in adults has to be evaluated.

Randomized controlled trials of anticoagulation therapy for the prevention of recurrent thrombosis are lacking. Retrospective multivariate analysis in cohort studies found that anticoagulation therapy significantly decreased the risk of recurrent thrombosis without increasing the risk of gastrointestinal bleeding.8, 12, 23 Actually, the risk of bleeding tended to be lower on anticoagulation therapy. The severity of bleeding was reported to be similar with or without anticoagulation therapy, and there were no bleeding-related deaths in patients receiving anticoagulation. Moreover, warfarin administration independently improved survival of patients with chronic portomesenteric venous thrombosis.23

Taken together, these limited but consistent data weigh for anticoagulation therapy in patients with chronic PVT. They indicate that large varices or previous gastrointestinal bleeding related to portal hypertension do not contraindicate anticoagulation, provided an appropriate prophylaxis for bleeding has been instituted. Still, further studies should address the issue of the benefit:risk ratio according to the presence of a permanent underlying prothrombotic disorder, a history of bleeding, and the involvement of the superior mesenteric vein.

Current outcome for treated patients with chronic PVT is good. In patients followed for 5 years, less than 5% died of the classical complications of PVT (intestinal infarction or gastrointestinal bleeding).22 In the medium term, mortality is related mainly to age and cause of PVT.8, 18, 22 Long-term outcome in patients with portomesenteric venous obstruction was poorer than in those with obstruction limited to the portal vein.28 Mortality related to MPD, which is negligible within 5 years of PVT diagnosis, progressively increases with time.3

Intrahepatic Portal Vein Thrombosis

There is evidence that obliteration of small intrahepatic portal veins induces regenerative changes in the lobule, sinusoidal dilatation, a multiplication of abnormal vascular channels within or at the periphery of the portal tracts, or fibrous expansion of the portal tracts.29 When nodular regeneration is isolated, the corresponding entity is denominated nodular regenerative hyperplasia. When portal tract changes are present, the corresponding entity has been called hepatoportal sclerosis in the west, noncirrhotic portal fibrosis in India, and idiopathic portal hypertension in Japan.

Thrombosis has been incriminated as a cause for obliterative portal venopathy. Indeed, a prothrombotic disorder has been found in approximately half the patients investigated in recent surveys.34 However, this is a lower prevalence than in PVT or Budd-Chiari syndrome (BCS) patients. Familial aggregation, as well as similar hepatic changes in patients with Turner's syndrome or Adams-Oliver syndrome, in the absence of prothrombotic conditions, suggest that genetically determined nonthrombotic mechanisms also may be involved.31, 32 A direct effect on endothelial cells without thrombosis might be implicated in arsenic or vinyl chloride intoxication, and in exposure to thorium sulfate, where similar hepatic lesions occur.33 Thus, the plausible link of nodular regenerative hyperplasia or hepatoportal sclerosis with intrahepatic PVT cannot be regarded as fully established yet. Extrahepatic PVT is found in up to 25% of patients at the time of diagnosis, and occurs during follow-up in 40%, a much higher incidence than in patients with cirrhosis.30 Extrahepatic thrombosis might be secondary to primary alterations of the venous wall, to secondary blood stasis and portal hypertension, or to both of these mechanisms.

Gastrointestinal bleeding is the main complication of nodular regenerative hyperplasia and hepatoportal sclerosis or related entities.30 Cases of end-stage liver disease are increasingly reported.30, 34 Diagnosis is based on histological features. Besides an adequate specimen, the pathologist's experience and awareness are crucial. Many patients coming to transplantation are misdiagnosed as having cirrhosis of other origin before examination of the explanted liver.34

There has been no reported trial of anticoagulation therapy for nodular regenerative hyperplasia, or hepatoportal sclerosis and similar entities. A high bleeding risk and the uncertain pathogenesis stand against anticoagulation therapy. Conversely, a high risk of extrahepatic PVT, a possible need for liver transplantation later in the course of the disease, and an occasional association with prothrombotic conditions weigh in favor of anticoagulation trials in selected patients.

Budd-Chiari Syndrome

Primary BCS can be related to pure hepatic vein thrombosis, to pure thrombosis of suprahepatic IVC, or to a combination thereof. With time, scarring and remodeling of the thrombosed portion may yield a fibrous stenosis involving the entire length of the vein, or a short length of it, being sometimes so thin as to simulate a membrane or a web. Because of extremely diverse presentations, diagnosis should be considered in any patient with acute or chronic liver disease. Noninvasive imaging of the hepatic venous outflow tract rarely misses the diagnosis when the operator is experienced and aware of the clinical suspicion. X-ray venography is needed only for planning interventional therapy.35

Prothrombotic Disorders.

The local factors that trigger thrombosis within the hepatic venous outflow tract remain unidentified in over 95% of patients. The general prothrombotic factors consistently found across surveys are presented in Table 2. The prevalence of MPD (about 50%) is even higher than in patients with PVT.11, 13–15 Because the accumulated incidence of BCS or PVT in MPD patients is less than 10%,37 other factors are obviously involved. The association of factor V Leiden with BCS is as strong as with deep vein thrombosis of the lower limbs and much stronger than with PVT.9, 11, 13 The reverse holds true for G20210A prothrombin gene mutation.9, 11, 13–15 Further understanding of these apparent site specificities for prothrombotic disorders would shed light on the interactions between blood components and liver endothelia.

External factors play an important role in BCS. In Nepal, pure suprahepatic IVC thrombosis, which is the commonest cause of hospital admission for liver disease, is strongly related to extreme poverty.38 The odds ratio for BCS in oral contraceptive users (approximately 2.4) is similar to that for venous thromboses at other sites.9, 39 Pregnancy also appears to be a risk factor for BCS, based on the temporal association between both conditions.40 BCS related to oral contraceptives or pregnancy is almost always related to pure hepatic vein thrombosis.40 Therefore, within the hepatic venous outflow tract itself, there might be site specificity for thrombosis according to these external factors.41

The role of a primary deficiency in protein C, protein S, or antithrombin, and of hyperhomocysteinemia or antiphospholipid syndrome, could not be fully ascertained because of the changes in plasma levels occurring as a consequence of liver disease. Paroxysmal nocturnal hemoglobinuria42 and Behcet's disease43 are rare diseases that seem to be overrepresented in BCS patients.

Overall, an underlying prothrombotic condition is found in up to 87% of BCS patients.11 A combination of several conditions can be demonstrated in approximately 25% of patients, a proportion much higher than expected in a healthy population.9, 11 A combination with another prothrombotic condition is particularly common in patients with heterozygous factor V Leiden44 or in oral contraceptive users or pregnant women.39 A requirement for concurrent prothrombotic disorders would explain why BCS is so rare, and also why only a minority of patients with any given prothrombotic disorder will develop BCS.


Symptomatic BCS is a spontaneously lethal disease in the short term. It has been estimated that 90% of untreated patients die within 3 years of diagnosis.45 Causes of deaths are massive, intractable ascites with renal failure and emaciation, gastrointestinal bleeding related to portal hypertension, or liver failure. Pathological and radiological studies have indicated that in most patients, thrombosis occurs and progresses independently in the major hepatic veins, at a speed that seems to vary according to patients.35 Collaterals to obstructed hepatic veins or IVC likely play a major compensatory role. The deleterious impact of superimposed intrahepatic and extrahepatic PVT has been well established.46

Survival increased steadily in the last 4 decades, as illustrated in Fig. 1. It is difficult to clarify the parts played in this improvement by earlier diagnosis in patients with little symptoms, introduction of specific treatment means, and general improvement of care for patients with liver disease.

Figure 1.

Schematic representation of survival curves according to the period of enrollment in series of patients with BCS not selected on the basis of treatment received. Surgical portosystemic shunts were in wide used from the early 1970s to the late 1990s. Transplantation and percutaneous angioplasty have been used since the late 1970s. Routine anticoagulation has been proposed since 1985. TIPS has progressively replaced surgical shunts since the late 1980s. Data are from Tavill et al.45 for the period 1960–1970, from Zeitoun et al.47 for the period 1970–1985, from Murad et al.48 for 1987–2002, and from Plessier et al.49 for 1997–2004.

The Place of Anticoagulation in Therapy.

Permanent anticoagulation therapy appears logical because BCS patients usually have underlying prothrombotic disorders that are not curable. There have been no randomized controlled trials in this area. Two retrospective studies with multivariate analysis have attempted to evaluate the impact of anticoagulation on mortality from BCS. In a multicenter French study, 120 patients admitted from 1970 to 1992 were enrolled.47 Permanent anticoagulation was systematically administered to patients admitted from 1985 on. Analysis by year of admission disclosed a sharp increase in survival starting in 1985. No other change in referral or management pattern taking place in 1985 could be identified. In a recently reported international study, 237 patients were enrolled.48 Overall, 171 patients (72%) were treated with anticoagulants. The use of anticoagulants did not yield a significant beneficial effect on survival in the total population [relative risk, 1.05; 95% confidence interval (CI), 0.62–1.76]. Separate analyses of the effect of anticoagulation on survival for 3 classes of prognosis suggested improved survival for patients with a good prognosis (relative risk, 0.14; 95% CI, 0.02–1.21), but not for those with an intermediate (relative risk, 0.88; 95% CI, 0.39–2.01) and poor prognosis (relative risk, 1.3; 95% CI, 0.50–3.04). Neither of these 2 studies included the presence of an underlying risk factor for thrombosis in the analysis.

There has been no report of bleeding-related death in BCS patients on anticoagulation, but there have been few studies on this particular issue. A recent study disclosed a high rate of anticoagulation-related complications in patients undergoing transhepatic interventional therapy. Many patients had heparin-induced thrombocytopenia (14%), with recurrent thrombosis in two thirds of those afflicted.49

Some data on anticoagulation are derived from the experience in recurrent thrombosis after interventions. After percutaneous hepatic vein or IVC angioplasty, univariate analysis disclosed that lack of anticoagulation therapy for at least 6 months was associated with reobstruction.50 In a recent European survey on 248 patients undergoing transplantation for BCS, 85% of patients received anticoagulation after transplantation.51 Within a median follow-up of 48 months, thrombosis recurred in 27 anticoagulated patients (11%), at the level of the portal vein in 17, or at the hepatic veins or IVC in 12. Mortality was 40.7% in patients with recurrence. Hemorrhage attributed to anticoagulants was observed in 27 patients (11%). The mortality attributed to anticoagulants was 1%.51 Thus, recurrent BCS after transplantation is uncommon with anticoagulation therapy. However, morbidity and mortality from anticoagulation are significant.

Despite severe limitations in interpreting the above data, panels of experts have invariably recommended immediate and permanent anticoagulation therapy for hepatic vein thrombosis.20, 52 Permanent anticoagulation alone is associated with complete control of BCS in up to 20% of patients. In the other patients, radiological decompressive procedures are needed.49, 53 Hepatic vein or IVC angioplasty (with or without stenting) will be followed by complete control of BCS in approximately 20% of patients; TIPS, in a further 65%. In up to 20% of patients failure of these procedures will require liver transplantation, although this proportion decreases with increased experience in radiological interventions.49, 53

Thrombosis, Prothrombotic Factors, and Anticoagulation in Parenchymal Liver Disease

These topics have been examined in 3 main areas: development of extrahepatic PVT; development of intrahepatic portal or hepatic venous thromboses; and progression of fibrosis in patients with chronic noncirrhotic liver disease.

Extrahepatic Portal Vein Thrombosis in Patients with Cirrhosis

The prevalence of PVT increases with the severity of the cirrhosis, being less than 1% in patients with compensated cirrhosis54 but 8% to 25% in candidates for liver transplantation.55 A differential diagnosis of malignant obstruction should always be considered. In patients with cirrhosis, PVT is often accompanied by gastrointestinal bleeding, ascites, or encephalopathy.56 In many patients, the thrombus is partial, and it changes in aspect and location at follow-up imaging. When the thrombus extends to the superior mesenteric vein, the risk of intestinal infarction is high.56 A major issue that remains unanswered is whether development of PVT is the chicken or the egg of advanced liver disease.

Even in patients with well-compensated cirrhosis, an underlying prothrombotic condition is difficult to detect because of the secondary changes in markers of many prothrombotic disorders. Patients with PVT, compared with those without PVT, have an increased prevalence of factor V Leiden, methylene tetrahydrofolate reductase C677T polymorphism, and prothrombin gene mutation, the latter being particularly frequent.56 Splenectomy and surgical portosystemic shunting and endoscopic therapy of esophageal varices are associated with an increased risk of PVT in patients with cirrhosis. However, these interventions could be mere indicators of severe portal hypertension, which per se might be a risk factor for PVT.56

With respect to anticoagulation therapy, the scarce available data are limited to patients who develop PVT while waiting for liver transplantation. Among 19 such patients given anticoagulation, 10 had recanalization, whereas among 10 historical controls, none had recanalization.55 Posttransplantation prognosis is strongly compromised by persistent complete PVT.55 These data are too limited to advocate anything else than further trials in selected patients from specialized centers. In our current view, for patients with decompensated disease, such a selection might be made on the fulfillment of either 1 of the following criteria: (1) the patient is a candidate for liver transplantation; (2) thrombosis extends into the superior mesenteric vein and intestinal ischemia was suspected; or (3) a prothrombotic disorder has been demonstrated without any ambiguity. By contrast, we believe that in patients with well-compensated liver disease, the occurrence of PVT should not be considered the mere consequence of cirrhosis. Therefore, these patients should be investigated for a cause and treated like PVT patients without cirrhosis.

PVT is a rare but serious complication of orthotopic liver transplantation.57 In this setting, the incidence has been reported to be 1.16% to 2.7%. Risk factors include pathological changes of the vessel wall (mainly in relationship to preexisting PVT); splenectomy during transplantation; and donor/recipient portal vein diameter mismatch. Possible consequences of PVT are graft dysfunction, graft loss, portal hypertension, and death. Graft loss is associated with complete PVT in the early posttransplantation course. Treatment for PVT has usually been regarded as necessary when additional complications such as arterial occlusion or bile duct injuries have occurred. Posttransplantation PVT has been managed in various ways that are discussed elsewhere.57 Routine postoperative screening with Doppler sonography and multidisciplinary approaches to treatment have considerably reduced graft loss and patient mortality. Available data do not allow for making a clear idea of the benefit:risk ratio of anticoagulation in this particular context.

Thrombosis of Intrahepatic Vessels in Patients with Cirrhosis

In a study of 61 cirrhotic livers removed at transplantation, intimal fibrosis was found at the level of hepatic veins and portal veins in 70% and 36% of livers, respectively.58 The distribution of hepatic venous lesions was patchy and largely confined to veins of medium caliber. Portal venous lesions were more uniform throughout the liver. Hepatic venous lesions were associated with regions of confluent fibrosis, so-called focal parenchymal extinction. Portal venous lesions were associated with regional variation in the size of cirrhotic nodules and bleeding related to portal hypertension. The interpretation of these findings has been that (1) thrombosis likely explains intimal fibrosis, and is of frequent occurrence in cirrhosis; (2) the hepatic and portal veins are affected in a different pattern; and (3) these thromboses could cause progression of cirrhosis. These findings are in line with those made in different settings, including congestive heart failure,59 BCS,46, 60 and hepatoportal sclerosis.29 Before this interpretation can be fully endorsed, it should be ascertained that intimal fibrosis is strongly related to thrombosis. This issue is of importance because anticoagulation would then be a logical proposal to prevent thrombosis and progression of cirrhosis.

Prothrombotic Disorders in the Progression of Fibrosis Related to Chronic Hepatitis

Experimental data show intricate relationships between hepatic fibrosis and thrombin activation. These data have stimulated interest in prothrombotic disorders as a factor for progression in chronic parenchymal diseases. In mice homozygous for factor V Leiden mutation, the only noticeable lesion has been occasional hepatic fibrosis.61 In the rat or mouse models of CCl4 intoxication, early deposition of fibrin/fibrinogen/fibronectin occurs in the vessels of the necrotic areas and in the fibrous septa.62 In the murine hepatitis virus 3 infection model, disease severity is strongly related to microvascular thrombi likely elicited by the induction of fibrinogen-like protein 2 (FgI2)/fibroleukin prothrombinase in activated endothelial cells and macrophages.63 Hepatic stellate cells are activated by thrombin through the induction and activation of protease activated receptors (PARs). Activation of PAR1 in cultured stellate cells causes their proliferation, contraction, and secretion of collagen 1 and inhibits their migration (see for example Gillibert-Duplantier et al.64). In the CCl4 rat and mouse models, and in the bile duct ligation rat model, inhibition of thrombin, or PAR inhibition or inactivation, attenuated stellate cell activation, decreased fibrosis area, and decreased collagen 1 expression at the protein and messenger RNA levels (for example Fiorucci et al.65). Furthermore, decreased fibrosis has also been found with administration of low-molecular-weight heparin in the rat with CCl4 or thioacetamide intoxication, or with bile duct ligation (see for example Abe et al.66). In these models, inhibition of thrombin, of its generation, or of its receptor did not attenuate the severity of hepatic cell necrosis and of cholestasis, which suggests a selective effect on fibrogenesis.

Corresponding data in patients with liver disease are somewhat limited, but consistent. Thrombin stimulates proliferation of cultured human stellate cells.67 Up-regulation of thrombin receptors has been shown to occur in hepatitis C virus and hepatitis B virus–related acute or chronic liver disease.68 In patients with chronic active hepatitis B, there is marked expression of FgI2/fibroleukin prothrombinase correlating with fibrin deposition.63

In patients with chronic hepatitis C or nonalcoholic steatohepatitis,69 decreased plasma levels of antithrombin, protein C or protein S, or increased factor VIII levels have been consistently associated with more severe fibrosis stage (see for example Poujol-Robert et al.70). Again, it remains difficult to ascertain whether these prothrombotic alterations are a cause for fibrosis progression, a mere marker for a higher stage of the disease, or both. However, cross-sectional multivariate analyses on large cohorts of patients showed an association of factor V Leiden with an accelerated progression to cirrhosis (odds ratio approximately 4), whereas no such association was found for prothrombin gene mutation.71 In line with these considerations, preliminary uncontrolled data in a small sample of patients with chronic hepatitis B have suggested a benefit from low-molecular-weight heparin.72


It is now well established that hepatic and portal vein thromboses are strongly associated with several specific prothrombotic factors. The prominent role of MPD is striking although still unexplained. Recent advances in molecular tests for MPD, factor V Leiden, and G20210A prothrombin gene mutation have considerably facilitated the identification of these causes. These advances contrast with the persistent difficulties in diagnosing underlying antiphospholipid syndrome, hyperhomocysteinemia, or deficiencies in protein C, protein S, or antithrombin. There is indirect but consistent evidence that many patients with primary vascular liver diseases benefit from anticoagulation therapy. However, the constraints of permanent oral anticoagulation and the potential severity of its complications cannot be neglected. Further studies are required to identify the subgroups of patients in which the benefit:risk ratio of permanent anticoagulation is most favorable, and those in which it is null or unfavorable. In addition, safe oral anticoagulant agents, requiring a minimal monitoring that would not be affected by liver dysfunction, are eagerly awaited.

Further studies are also needed in the field of chronic parenchymal diseases. An important issue is whether intrahepatic or extrahepatic, portal or hepatic venous thromboses are a mere reflection of liver disease severity or an aggravating factor. A further issue is whether anticoagulation therapy for recanalization of nonmalignant PVT in patients with cirrhosis is efficient, beneficial, and harmless. Still another issue is whether primary prothrombotic alterations or secondary changes mimicking primary alterations actually translate into increased thrombin generation and accelerated fibrosis in patients with primary parenchymal liver disease. The answers to these questions will allow for considering whether it is justified to trial anticoagulation therapy as a mean to prevent aggravation of otherwise incurable, rapidly progressive, or advanced chronic liver disease. 2

Figure 2.

Schematic representation of the two pathways of thrombin actions in animal models of liver diseases. The first pathway is through its serine protease activity, leading to fibrin deposition in hepatic microvessels. The second pathway is though engagement with PARs (mainly PAR1) present on stellate cells, resulting in the activation of these cells. Theoretically, prothrombotic disorders can be expected to increase parenchymal damage or fibrogenesis by increasing thrombin generation. Limiting thrombin generation by anticoagulation, direct thrombin inhibition, and inhibition of PAR1 have been shown to decrease fibrogenesis in various animal models of liver injury.


The author is indebted to Pierre-Emmanuel Rautou, Bertrand Condat, and Aurélie Plessier for fruitful discussion of the manuscript.