Systematic review: portal vein thrombosis in cirrhosis

Authors

  • E. A. TSOCHATZIS,

    1. The Royal Free Sheila Sherlock Liver Centre and Division of Surgery, Royal Free Hospital, Hampstead, London, UK
    Search for more papers by this author
  • M. SENZOLO,

    1. The Royal Free Sheila Sherlock Liver Centre and Division of Surgery, Royal Free Hospital, Hampstead, London, UK
    2. Gastroenterology and Liver Transplantation Unit, Department of Surgical and Gastroenterological Sciences, University Hospital of Padua, Padua, Italy
    Search for more papers by this author
  • G. GERMANI,

    1. The Royal Free Sheila Sherlock Liver Centre and Division of Surgery, Royal Free Hospital, Hampstead, London, UK
    Search for more papers by this author
  • A. GATT,

    1. Haemophilia Centre & Thrombosis Unit, Royal Free Hospital, Hampstead, London, UK
    Search for more papers by this author
  • A. K. BURROUGHS

    1. The Royal Free Sheila Sherlock Liver Centre and Division of Surgery, Royal Free Hospital, Hampstead, London, UK
    Search for more papers by this author

Prof. A. K. Burroughs, The Royal Free Sheila Sherlock Liver Centre and Division of Surgery, Royal Free Hospital, Pond Street, Hampstead, London NW3 2QG, UK.
E-mail: andrew.burroughs@royalfree.nhs.uk

Abstract

Aliment Pharmacol Ther 31, 366-374

Summary

Background  As current imaging techniques in cirrhosis allow detection of asymptomatic portal vein thrombosis during routine ultrasonography, more patients with cirrhosis are diagnosed with portal vein thrombosis. Although a consensus on noncirrhotic extra-hepatic portal vein thrombosis has been published, no such consensus exists for portal vein thrombosis with cirrhosis.

Aim  To perform a systematic review of nonmalignant portal vein thrombosis in cirrhosis in terms of prevalence, pathogenesis, diagnosis, clinical course and management.

Methods  Studies were identified by a search strategy using MEDLINE and EMBASE.

Results  Portal vein thrombosis is encountered in 10–25% of cirrhotics. In terms of pathophysiology, cirrhosis is no longer considered a hypocoagulable state; rather than a bleeding risk in cirrhosis, various clinical studies support a thrombotic potential. Clinical findings of portal vein thrombosis in cirrhosis vary from asymptomatic disease to a life-threatening condition at first presentation. Optimal management of portal vein thrombosis in cirrhosis is currently not addressed in any consensus publication. Treatment strategies most often include the use of anticoagulation, while thrombectomy and transjugular intrahepatic portosystemic shunts are considered second-line options.

Conclusions  Portal vein thrombosis in cirrhosis has many unresolved issues, which are often the critical problems clinicians encounter in their everyday practice. We propose a possible research agenda to address these unresolved issues.

Introduction

Portal vein thrombosis (PVT) is encountered in a variety of clinical settings such as myeloproliferative diseases, cirrhosis, cancer or infection.1–4 Its clinical presentation, prognosis and management vary substantially according to aetiology. It can be classified as acute or chronic, extra- or intrahepatic, occlusive or non-occlusive and progressive or self-resolving. PVT in cirrhosis has rarely been the subject of detailed evaluation and thus many questions remain unanswered regarding its clinical course, prognosis and optimal management. Indeed, although a consensus concerning noncirrhotic extra-hepatic PVT has been published,5 no such consensus exists for PVT with cirrhosis. An evolving consensus on portal hypertension as well as a recent consensus on vascular disorders of the liver also failed to address the matter adequately.6, 7

As current imaging techniques allow detection of asymptomatic PVT during routine ultrasonographic examination, more patients with cirrhosis are diagnosed with PVT. Furthermore, while liver transplantation has altered the prognosis of patients with cirrhosis, the presence of PVT can exclude a patient from transplant listing, or negatively impact on post-transplantation survival.8 Therefore, there is a growing need for optimal, evidence-based management of PVT in cirrhosis. In the present systematic review, we explore the different aspects of nonmalignant PVT in cirrhosis in terms of prevalence, pathogenesis, clinical course, prognosis and management. We focus in particular on the unresolved issues, which are often the critical problems clinicians encounter in their everyday practice. We try to provide answers to these problems on the basis of available literature. Finally, we propose a possible research agenda to address these unresolved issues.

Methods

Studies were identified by a search strategy using MEDLINE and EMBASE. We also searched the references of selected studies to identify further relevant studies. For the MEDLINE search, we used the following terms: (‘liver cirrhosis’ [MeSH Terms] OR ‘cirrhosis’ [All Fields] OR ‘cirrhoses’ [All Fields]) AND (‘portal vein’ [MeSH Terms] OR ‘portal vein’ [All Fields]) AND (‘Thrombosis’ [MeSH Terms] OR ‘venous thrombosis’ [MeSH Terms] OR ‘Embolism and thrombosis’ [MeSH Terms] OR ‘Thrombosis’ [All Fields]. For the EMBASE, we used the following terms: (cirrhosis OR exp liver cirrhosis/) AND (exp thrombosis/OR exp vein thrombosis/OR exp deep vein thrombosis/OR exp thrombosis prevention/OR exp portal vein thrombosis/OR exp liver vein thrombosis/OR exp mesenteric vein thrombosis/OR thrombosis). We identified 4375 search results in MEDLINE and 1567 in Embase. Studies were eligible if they referred to any aspect of prevalence, pathophysiology, clinical presentation, diagnosis and management or therapy of PVT in cirrhosis. We excluded studies that referred to noncirrhotic or hepatocellular carcinoma-related PVT. We only included case reports or case series, if no cohort studies were available. We thus identified 79 relevant studies.

Prevalence

Amongst all cases of PVT, cirrhosis is the underlying cause in 22–28%.9, 10 The reported prevalence of PVT in cirrhosis largely depends on the diagnostic tool used and the inclusion or exclusion of patients with hepatocellular carcinoma (HCC). When HCC patients are excluded, studies based on ultrasound have reported a prevalence of 10–25%,8, 11–13 with the prevalence being 11% in the largest study comprising 701 patients.14 Operator experience or different patient and/or disease characteristics may explain these differences. When different means of diagnosis are used, such as angiography or surgery, the prevalence has ranged from as little as 0.6% to 16%.15–18 Aetiology of liver disease has an influence on prevalence according to a study of 885 patients who underwent liver transplantation, being 3.6% in primary sclerosing cholangitis, 8% in primary biliary cirrhosis, 16% in alcoholic and HBV-related cirrhosis and mounting to 35% in HCC.19 The true prevalence of PVT in cirrhosis might be greater, as intrahepatic thrombosis is more difficult to diagnose and may not be looked for unless the main portal vein shows thrombosis. Moreover, in an autopsy study, portal vein intimal fibrosis was found in 36% of the livers examined, but most of this portal vein disease would not have been clinically detectable with available ultrasound techniques.20 Finally, the risk of PVT is independently associated with the severity of cirrhosis,14 i.e. with worsening indices of coagulation, there is more thrombosis.

While there are many studies on prevalence, this is not the case for the incidence of PVT in cirrhosis. In a cohort of 251 patients with cirrhosis listed for liver transplantation, 17 (7.4%) developed de novo PVT after the initial evaluation during a mean of 12 months of follow-up,8 while a similar incidence of 16% (12/73 patients) was reported in a recent Italian study with 12 months prospective follow-up.21

Pathophysiology

Recently, our understanding of coagulopathy in cirrhosis has changed, and cirrhosis is no longer considered to be a hypocoagulable state. What happens is that in the setting of hepatic synthetic impairment, both pro- and anticoagulant proteins are reduced to a similar degree. The net result is a haemostatic balance that in normal circumstances is compensated, with no tendency for bleeding or thrombosis.22 Indeed, thrombin generation in stable patients with cirrhosis was found to be normal in an in vitro study, when the natural anticoagulant C pathway was activated by the addition of thrombomodulin.23 The same group reported that thrombin generation correlated with platelet number, but only when thrombocytopenia was severe was thrombin generation impaired.24 The international normalized ratio (INR) in liver disease probably overestimates the bleeding risk as the international sensitivity index used is determined by means of plasma from patients on vitamin K antagonists.25 The same is true of the rest of conventional coagulation tests in patients with cirrhosis, as they do not take into consideration the reduction in anticoagulant proteins.26 This might explain the paradox of the poor prediction of bleeding with the prolonged global coagulation tests. However, the haemostatic balance in cirrhosis can be easily deregulated such as with an infection.27

Moreover, rather than a bleeding risk (excluding portal hypertension-related bleeding) in cirrhosis, various clinical studies as well as a recent in vitro one28 support a thrombotic potential. Three recent independent studies have been published specifically evaluating the prevalence of deep vein thrombosis (DVT) and pulmonary embolism (PE) in patients with cirrhosis, two of which were case-control studies29, 30 and one, a retrospective study.31 Overall, 24,037 cirrhotic patients and 12,518 controls (113 cirrhotics without DVT) were studied and the incidence of DVT or PE ranged from 0.5% to 1.87% across the different studies. In particular, in the largest population (which was the one with the lower incidence), 21% of patients were receiving antithrombotic prophylaxis with drugs (7%) or compression devices (14%) at the time they experienced the thrombotic event.30 Thus, even patients with cirrhosis and a prolonged PT receiving antithrombotic drugs can develop a venous thrombosis. Moreover, in a Danish case-control study, patients with cirrhotic or noncirrhotic liver disease had significantly higher relative risk for venous thromboembolism compared with that in controls (from 1.74 to 1.87 in patients with cirrhotic and noncirrhotic liver disease respectively).32 In two studies,29, 30 decreased hepatic synthetic function as reflected by albumin levels was associated with the risk of DVT/PE, suggesting a greater tendency to thrombosis the more severe the liver disease. Indeed, one study also confirms that the risk of PVT is worse with more severe cirrhosis.14 Recently, in vitro evidence supporting the above epidemiological findings was published evaluating patients with cirrhosis. The ratio of the two most powerful pro- and anti-coagulants operating in plasma, factor VIII and protein C respectively, showed a balance strongly in favour of factor VIII, which indicates hypercoagulability.28 Importantly, concentrations of factor VIII, a potent pro-coagulant involved in thrombin generation, increased progressively with worsening Child-Pugh grade and score (from Child class A to C).28

As mentioned above, the prevalence of PVT in cirrhotics is 10–25%. Thrombosis risk is substantial when any of the components of Virchow’s triad i.e. venous stasis, endothelial injury and hypercoagulability is present. In cirrhosis, stasis in the portal vein is favoured by the relative splanchnic vasodilatation and is further aggravated by the liver architectural derangement that decreases portal flow. In one prospective study, a reduced portal flow velocity was the only predictive variable for the development of PVT in a cohort of 73 patients with cirrhosis.21 Local factors like chronic pancreatitis may further aggravate stasis. Moreover, a thrombophilic genotype has been reported in 69.5% of patients with cirrhosis and PVT; mutations TT677 of methylenetetrahydrofolate reductase and G20210A in the prothrombin gene were significantly more frequent in this group compared with that in controls.11, 14 A recent study reported that high levels of factor VIII were independently associated with both noncirrhotic and cirrhotic PVT, with the odds ratio for thrombosis for levels above 129 IU/dL being 6.0 for cirrhosis.33 The role of anticardiolipin antibodies remains controversial, as available studies have reported conflicting results.14, 34–36 Finally, in a small study of 11 patients, endotoxaemia emerged as a potential mechanism for facilitating the clotting cascade and favouring PVT.37 Indeed, endotoxaemia is common with worsening liver disease and may be its surrogate.38, 39 The role of sclerotherapy as a potential trigger factor for PVT is controversial40, 41 and any reported association could be as a result of selection bias of patients with more severe portal hypertension. The theoretical risk for cyanoacrylate glues is higher than sclerotherapy; however, existing evidence comes from case reports or small case series.42–44

Diagnosis

Ultrasound and ultrasound Doppler are almost always sufficient for diagnosis.45–47 Contrast-enhanced sonography was shown to be superior to sonography and colour Doppler sonography for the detection and characterization of PVT.48 The gold standard of invasive angiography such as portal venography or superior mesenteric arteriography is rarely necessary. Retrograde carbon-dioxide portography can make PVT more evident than conventional CT/MRI, especially when an important hepatofugal flow is present.49 However, CT and MRI are better for determining the extent of thrombosis.46, 50–52 Contrast-enhanced sonography or CT imaging can help differentiate benign from malignant PVT.53, 54 After diagnosis, further evaluation with upper gastrointestinal endoscopy is warranted to assess the presence and degree of oesophageal varices.

Clinical findings and prognosis

Clinical findings of PVT in cirrhosis vary from asymptomatic disease to a life-threatening condition at first presentation. In a study of 79 patients with PVT and cirrhosis at diagnosis, 43% were asymptomatic, 39% had gastrointestinal bleeding because of varices or portal hypertensive gastropathy and 18% had acute abdominal pain of which 70% had intestinal infarction (10 of 79).14 Intestinal infarction occurs as a consequence of the extension of the thrombus in the portal vein, to the superior mesenteric vein; it is a fatal complication if not rapidly recognized and treated. As bleeding risk appears higher at 39% in PVT superimposed on cirrhosis than cirrhosis alone (27%),55 there is even more reason to schedule endoscopy to screen for oesophageal varices and to institute appropriate management for prophylaxis against bleeding.

Factors influencing recanalization or extension of thrombus have not been reported to date and therefore the natural history of PVT in cirrhosis is not clear, including how survival is affected. Although in a single study of 37 consecutive patients with nonmalignant cirrhotic PVT, a 5-year survival of only 45% was reported,9 there was no direct comparison with a matched control group and hence the question of what effect PVT has on survival in cirrhosis remains. The thrombus in the portal vein can spontaneously resolve or further extend distally into intrahepatic branches or proximally into the superior mesenteric vein. In an autopsy study, Wanless et al.20 found that thrombotic events began in a large vein with propagation to smaller veins, followed by recanalization in the larger veins. However, as PVT usually occurs in Child B or C patients with cirrhosis, it is not always easy to attribute specific symptoms or deteriorating clinical condition to either worsening of the cirrhosis or to the de novo occurrence of PVT. Indeed, chronic intrahepatic microthrombosis, termed as ‘parenchymal extinction’ has been suggested as a pathogenetic mechanism in cirrhosis, per se aggravating fibrosis and inducing atrophy.20, 56 A ‘porto-centric’ pattern of cirrhosis nodules is characteristic in patients with PVT, with enlargement of cirrhotic nodules adjacent to major portal systems.20

Management

Optimal management of PVT in cirrhosis is currently not addressed in any consensus publication, including the recent practice guidelines on vascular disorders of the liver.7 The advent of liver transplantation has altered the natural history of cirrhosis and made this issue clinically relevant. PVT and especially thrombus extension to the superior mesenteric vein may adversely affect the outcome of liver transplantation and may even preclude patients from being listed because of technical difficulties in intra-operative anastomoses.57, 58 While there have been r reports, all retrospective, suggesting that PVT does not affect post-transplant mortality, none has evaluated the effect of PVT on the patients’ eligibility for liver transplantation nor reported on exclusions from listing because of PVT.59–62 Moreover, as already discussed above, even asymptomatic PVT might negatively impact on survival, but current allocation systems and ‘exceptions’ do not take PVT into consideration. In the absence of randomized controlled trials, all existing evidence concerning treatment comes from case series and is therefore of low quality.

Treatment strategies that are published most often include the use of anticoagulation, while thrombectomy and transjugular intrahepatic portosystemic shunts (TIPS) are considered as second-line options. Regarding anticoagulation, its goal is not necessarily recanalization, but prevention of further thrombosis with extension to the superior mesenteric vein (SMV). Consensus on noncirrhotic PVT and the recent American Association for the Study of Liver Disease (AASLD) practice guidelines on liver vascular disorders advocate anticoagulation in acute PVT, as recanalization may occur in up to 80% of cases.5, 7 Moreover, in 84 patients with nonmalignant noncirrhotic PVT who received anticoagulation, a significant reduction in the risk of new thrombotic events was noted with no aggravation in the risk of bleeding.63 However, anticoagulation is more complex in the setting of cirrhosis. First, anticoagulation in patients with oesophageal varices, especially when the risk of rupture risk is high, could further exacerbate a variceal bleeding should it occur. Thus, although individual case reports of successful use of anticoagulation have been published,64 in everyday clinical practice, the risk to benefit ratio of anticoagulation usually proves unfavourable in cirrhosis if there are medium or large oesophageal varices. To date, only one study evaluated anticoagulation in PVT in 29 patients with cirrhosis listed for transplantation.8 The proportion of patients with partial or complete recanalization was significantly higher in those who received anticoagulation (8/19) compared with those who did not (0/10). No anticoagulation-related complications were noted and no patient had to stop treatment. It is noteworthy that patients with complete PVT at the time of liver transplantation had significantly lower survival after transplant. This suggests that at least in patients with cirrhosis and PVT listed for liver transplantation, anticoagulation should be used.

As the INR is a basic component of the MELD score, currently used in the US for prioritizing patients for liver transplantation, and the use of coumarin anticoagulants increase the INR and ‘artificially’ increase the score,65 an alternative score that excludes INR (MELD-XI) has been proposed for patients on anticoagulation listed for transplantation.66 An attractive alternative to oral anticoagulants could be the use of low-molecular weight heparin, the dosing of which is weight-based and screening therefore is not necessary. A 50% portal vein recanalization was recently reported with low-molecular weight heparin in 38 patients with cirrhosis with only one episode of nonsevere variceal bleeding.67 A second study reported complete response in 75% of patients by using enoxaparin with no serious side effects. However, 64% of the cohort had only partial occlusion at presentation.68 It is noteworthy that enoxaparin was given for a prolonged period of 6–17 months.68

If anticoagulation is to be used, there is no consensus on whether nonselective beta-blockers are sufficient for bleeding prophylaxis if varices are present or whether varices should be eradicated by ligation and any ulcers healed before anticoagulation is started. A theoretical advantage of endoscopic therapy over nonselective beta-blockers could be argued for, as the latter might further reduce splanchnic blood flow and thus cause further portal vein stasis and thrombus extension.69 Eradication of varices through ligation might be a safer choice, but iatrogenic bleeding is possible with ligation.70 In the absence of randomized or even observational studies, no recommendation can be made on the optimal first choice or indeed whether combination therapy is the best option.

Another important issue is that monitoring of anticoagulation by use of the INR in cirrhosis is likely to be suboptimal. The INR is prothrombin time (PT) ratio introduced in 1983 in an attempt to harmonize PT results among laboratories.71 It is calculated by dividing patients’ PT by a control mean normal PT and the result is then raised to the power of the international sensitivity index (ISI) for thromboplastin. However, in cirrhosis, a 29% variation in mean INR was documented when three different thromboplastin reagents were used.72 Moreover, significant interlaboratory variation in INR measurement reaching 100% has been documented, which also affected MELD score calculations.73 This occurs because the ISI is determined by using plasma from patients on warfarin therapy. Indeed, the INR system has only been validated in patients with normal liver function on stable anticoagulation.74 This problem could be solved if plasma from patients with liver disease were used for plasma calibration and indeed, two independent groups have recently reported on such a calibration, thus introducing the term ‘INR liver’.75, 76 However, as yet there is no such standardized measurement, as it would require two different INR formulae in individual laboratories, one for liver patients and another for warfarin monitoring and most probably it would not be cost-effective.25

Radiological interventions can be used in PVT to re-permeate the vessel even percutaneously in acute forms through the chest wall or by using TIPS when concomitant portal hypertensive complications or cirrhosis is present. The use of TIPS is a feasible option, provided that the intrahepatic portal branches are patent.77, 78 The Royal Free Hospital experience of 28 patients with PVT undergoing TIPS, of whom all had bled (12 had cirrhosis), reported a success rate of 73% (19/28).79 In all patients, TIPS placement was followed by anticoagulation with warfarin. However, two of the 12 patients with cirrhosis died of haemorrhagic complications 2 days after TIPS placement, but three were successfully transplanted in the months following the procedure.79 Interestingly, the ‘age’ of the thrombus or its extent were not predictors of a successful TIPS placement; only the patency of intrahepatic portal branches emerged as a significant factor. Other groups have also reported favourable outcomes of TIPS in the setting of PVT and cirrhosis,80, 81 thus providing a bridge to liver transplantation. Cavernomatous transformation does not preclude attempt at TIPS placement.82 Finally, successful combined mechanical thrombectomy with intraportal thrombolysis by radiological means has been reported,83 but as for TIPS, it requires specialized units to which these patients should be referred.

As part of a therapeutic algorithm currently used in one of our centres (Padova, Italy), when PVT is first diagnosed, an attempt at evaluating the time interval from appearance is made. Endoscopic screening for varices takes place. Anticoagulation using standard therapeutic dose is considered in all patients with acute onset of PVT and in patients with recent onset (<6 months) in whom repermeation is more likely. High-risk oesophageal varices are banded before initiation of anticoagulation and reduction in coagulation dose is advocated when platelet count is very low. After the first 6 months of anticoagulation, even if re-permeation is not accomplished, prophylactic anticoagulation is continued in patients with underlying thrombophilic conditions or in patients who are likely candidates for liver transplantation in the future, to avoid thrombosis extension in the splanchnic vessels. When PVT is longstanding and there is cavernous transformation of the portal vein, prophylactic anticoagulation is reserved only in patients who have thrombophilic conditions and/or high risk of thrombus extension into the SMV. Radiological interventions (i.e. TIPS and thrombectomy) are used in patients who have already had bleeding or intractable ascites or when thrombosis has extended despite adequate anticoagulation.

Conclusions: future directions

Portal vein thrombosis is fairly common in cirrhosis, with a prevalence of 10–25% and its diagnosis is probably more frequent because of the routine use of Doppler ultrasonography or CT. Important questions on optimal management have arisen particularly because of liver transplantation. These are summarized in Table 1. The first question concerns the prognostic significance of diagnosing PVT. The second is which patient should be treated: all patients, only those listed for liver transplantation or those with evidence of thrombus extension or its imminent occurrence? Finally, what is the optimal therapeutic algorithm? Is it oral anticoagulation for all or only for some patients? How should it be monitored? How soon should the clinician proceed to second-line treatment options such as TIPS if recanalization does not occur with standard treatment?

Table 1.   Important unanswered questions in cirrhotic portal vein thrombosis (PVT)
Does occurrence of PVT alter the natural history of cirrhosis and therefore should asymptomatic patients be treated with the goal of recanalization or prevention of further thrombus extension?
Should all patients with cirrhosis and PVT be aggressively anticoagulated?
Should this apply only to patients on transplantation waiting list?
If recanalization does not occur should patients be offered second-line treatment with transjugular intrahepatic portosystemic shunts?
How long should the interval be whilst being anticoagulated before considering therapy to have failed?
How should patients be monitored?
Is oral warfarin better than low-molecular weight heparin?

The answers to these questions should come from prospective cohort studies and randomized trials if appropriate and if applicable, as the answers should translate into important survival benefits for individual patients. Until more definitive answers are available and in line with existing evidence, it seems reasonable to use oral anticoagulation in patients with cirrhosis and PVT listed for liver transplantation, after banding and/or beta-blocker prophylaxis for existing oesophageal varices. In all other cases, decisions should be individualized according to the risk of anticoagulation versus extension of portal thrombosis and the likelihood or not of liver transplantation. TIPS therapy should be considered in patients who have bled or have intractable ascites and/or if there is thrombus extension.

Acknowledgements

Declaration of personal interests: E. Tsochatzis has received an educational grant from the Hellenic Association for the Study of the Liver. Declaration of funding interests: None.

Ancillary