It remains unclear when anticoagulant therapy should be given in patients with non-cirrhotic portal vein thrombosis (PVT). The aim of this study was to assess the effect of anticoagulation on recurrent thrombotic events and gastrointestinal bleeding in non-cirrhotic PVT patients.
Retrospective study of all patients with non-cirrhotic PVT (n =120), seen at our hospital from 1985 to 2009. Data were collected by systematic chart review.
Sixty-six of the 120 patients were treated with anticoagulants. Twenty-two recurrent thrombotic events occurred in 19 patients. The overall thrombotic risk at 1, 5 and 10 years was 4%, 8% and 27%, respectively. Seventy-four percent of all recurrent thrombotic events occurred in patients with a prothrombotic disorder. Anticoagulant therapy tended to lower the risk of recurrent thrombosis (hazard ratio [HR] 0.2, P = 0.1), yet the only significant predictor of recurrent thrombotic events was the presence of a prothrombotic disorder (HR 3.1, P = 0.03). In 37 patients, 83 gastrointestinal bleeding events occurred. The re-bleeding risk at 1, 5 and 10 years was 19%, 46% and 49%, respectively. Anticoagulation therapy (HR 2.0, P ≤ 0.01) was a significant predictor of (re)bleeding. Anticoagulation therapy had no effect on the severity of gastrointestinal bleeding. Poor survival was associated with recurrent thrombotic events (HR 3.1 P = 0.02), whereas bleeding (HR 1.6 P = 0.2) and anticoagulant treatment (HR 0.5 P = 0.2) had no significant effect on survival.
In non-cirrhotic PVT patients recurrent thrombotic events are mainly observed in patients with underlying prothrombotic disorders. Anticoagulation therapy tends to prevent recurrent thrombosis but also significantly increases the risk of gastrointestinal bleeding.
Portal vein thrombosis (PVT) is an important cause of portal hypertension. It is frequently seen as a complication of (advanced) liver cirrhosis or hepatobiliary malignancies but may also occur in the absence of associated liver disease [1-3]. Both local inflammatory factors and systemic procoagulant factors have been identified as risk factors for PVT in patients without cirrhosis or malignancy [4-6]. Myeloproliferative neoplasms (MPN) are especially recognized as one of the principal causes of PVT in non-cirrhotic patients [7-9]. Moreover, it is now clear that, like venous thrombosis of the lower extremities, development of PVT is often a multifactorial process involving more than one risk factor for thrombosis [10, 11].
PVT can be classified as either recent or chronic depending on clinical presentation and findings of imaging studies. Ascites and moreover bleeding from gastroesophageal or ectopic varices are major complications in patients with long-standing PVT [3, 4, 12, 13]. Patients with recent thrombosis are usually treated with (low-molecular-weight) heparin followed by vitamin K antagonists [9, 14, 15]. The duration of this treatment is still a matter of debate and guidelines are predominantly based on expert opinion. It is suggested, however, that in patients with PVT and an underlying prothrombotic disorder, anticoagulant therapy should be continued lifelong [16-18]. Nevertheless, data on the risk and benefit of anticoagulant therapy are limited. The benefit of treatment was reported by a few cohort studies of patients with PVT in the absence of cirrhosis. Some of these previous studies addressed anticoagulation therapy in patients with recent PVT only [9, 15, 19, 20].
The aim of this study was to assess the effect of anticoagulant therapy on thrombotic events and gastro-intestinal (re)bleeding during long-term follow-up of a large cohort of patients with non-cirrhotic PVT treated within one institution.
Patients and methods
Patients were identified by means of a search in the computerized patient registration system of our clinic. All adult patients identified between January 1985 and December 2008 were enrolled if PVT was documented and cancer, cirrhosis, liver transplantation or combined Budd–Chiari syndrome were absent. For all patients a standardized clinical record form for detailed clinical data was completed with data obtained from the medical charts.
Follow-up started at the time of diagnosis, defined as the date of the first radiological imaging documenting PVT and was continued to either December 2009 or death, whichever came first. Patients lost to follow-up were censored at the last visit.
Diagnostic criteria for PVT were partial or complete obstruction of the extrahepatic portal vein, as documented by radiological imaging, such as Doppler ultrasonography, computed tomography, magnetic resonance imaging and/or venography. Risk factors for the development of PVT were divided into inherited and acquired risk factors. Inherited risk factors were factor V Leiden mutation, prothrombin gene mutation, protein C deficiency, protein S deficiency and antithrombin deficiency. Inherited deficiencies of the latter three proteins were considered to be present if patients were known with this deficiency prior to PVT or a low concentration of these proteins was confirmed at least 6 weeks after the onset of PVT or, in case anticoagulation therapy was given, at least 6 weeks after discontinuation, all in the absence of hepatic dysfunction. Acquired risk factors were MPN, paroxysmal nocturnal hemoglobinuria, antiphospholipid syndrome, abdominal surgery, intra-abdominal infection, inflammatory bowel disease (IBD), the use of oral contraceptives or trauma. Abdominal surgery, trauma or abdominal infections were regarded as an underlying cause if these events had occurred within 3 months prior to the development of PVT. A patient was defined as having an underlying prothrombotic disorder if at least one of the following underlying disorders was present: inherited thrombophilia, MPN or antiphospholipid syndrome.
PVT was defined recent in the absence of a portal cavernoma and gastrointestinal varices. All other cases were classified as chronic PVT. Gastroesophageal varices were graded as: grade I, varices flattened by insufflation; grade 2, varices not flattened by insufflation; grade 3, confluence of varices not flattened by insufflation; and grade 4, grade 3 with red marks . Anticoagulant therapy included treatment with vitamin K antagonists, therapeutic use of low-molecular-weight heparin or unfractionated heparin. To compare the severity of bleeding between patients with and without anticoagulant therapy, we assessed the number of bleeding events per patient, the number of transfused red blood cells units, the number of Intensive Care Unit admissions and hemoglobin level at admission for each bleeding episode in both groups.
The Kaplan–Meier method was used to calculate overall survival and survival free of recurrent thrombosis and of (re)bleeding. The log-rank test was used for comparing groups. Predictors of additional thrombotic events and (re)bleeding were estimated as a hazard ratio (HR) and corresponding 95% confidence interval (CI) using Cox regression analysis adjusted for age and gender. The effects of thrombotic events, bleeding episodes and anticoagulant use on survival were estimated as time-dependent covariates using Cox regression analysis. Owing to the small number of thrombotic events, univariate Cox regression models, stratified for number of thrombosis, were used to determine significant predictors of thrombosis. As patients could stop and restart anticoagulation therapy during follow-up, we correlated recurrent thrombotic and bleeding events to the therapy used at that time. In case of oral anticoagulation use, a period of 3 days was added to the stop date and considered as using anticoagulation therapy. The use of anticoagulation therapy was analyzed as a time-dependent factor. Sensitivity analysis was performed for subgroups. A P-value of < 0.05 was considered statistically significant. Data were expressed as median values with the accompanying range and interquartile range (IQR) where appropriate. All statistical analyzes were performed with the Statistical Package for Social Sciences for Windows, version 16.0 (SPSS, Chicago, IL, USA).
From a total of 241 patients with PVT 120 patients, 43 males and 77 females with a median age of 44 years (range 16–87 years), were enrolled. Patients were excluded because of the following reasons: 63 patients had cirrhosis, 49 cases had underlying malignancies, seven patients had developed PVT after liver transplantation and in two cases there was concomitant Budd–Chiari syndrome. Baseline characteristics of the 120 analyzed patients are summarized in Table 1. In 53% of the patients thrombosis was confined to the portal vein. In the other cases thrombosis was more extensive, involving the splenic and/or mesenteric veins. Two patients presented with concomitant thrombosis outside the portal venous system: deep vein thrombosis of the lower limb and sinus sagittalis thrombosis, respectively. PVT was diagnosed by ultrasound (n = 69), computed tomography (n = 45), venography (n = 2), magnetic resonance imaging (n = 3) or per-operative (n = 1). The median duration of follow-up was 5.5 years (range 0.1–32.5 years; IQR was 2.0–10.4 years). An underlying prothrombotic state was found in 69 patients (58%). Thirty-nine patients (33%) had evidence of a MPN as determined by either bone marrow biopsy and/or presence of the JAK2 mutation . Gastro-intestinal blood loss at baseline was caused by variceal bleeding in all, except two patients where angiomatosis of the stomach and arteriovenous malformation of the rectum were present.
Table 1. Characteristics of 120 patients with portal vein thrombosis
Variable at baseline
No. of patients
Available data (%)
IBD, inflammatory bowel disease.
Median age (range)
History of thrombosis
Site of thrombosis
Portal and splenic veins
Portal and mesenterial veins
Portal, splenic and mesenterial veins
Protein C deficiency
Protein S deficiency
Factor V Leiden
Prothrombin gene mutation
Oral contraceptive use
Varices grade I/II/III/IV
Sixty-six patients (55%) were treated with anticoagulant therapy, 29 of 40 cases with recent PVT, 33 of 71 cases with chronic PVT and in 4 of 9 cases where onset of thrombosis could not be determined. In 59 patients (49%) treatment was instituted immediately after the diagnosis of PVT was established. In the other seven cases anticoagulant therapy was started during follow-up after a median period of 3.4 years (range 1.2–15.4 years). Anticoagulant therapy was used for a median period of 1.9 years (range 0–15.8 years; IQR 0.7–5.1 years). Forty-two patients (35%) still used anticoagulants at the end of follow-up. Anticoagulants were stopped in 24 of 66 patients for the following reasons: completed intended time of treatment (n = 16), iron deficiency anemia caused by menorrhagia (n = 2), epistaxis (n = 1), the presence of varices grade ≥ 2 or variceal bleeding (n = 2) or unknown (n = 3).
Of the 120 patients, 16 cases (13%) were treated with anticoagulant therapy and antiplatelet drugs, 13 (11%) were on antiplatelet treatment only and 41 patients (34%) were neither on anticoagulant and antiplatelet therapy. The number of patients on antiplatelet drugs only was too small to assess the effects on bleeding, recurrent thrombosis or survival.
Recurrent thrombotic events
A total of 22 recurrent thrombotic events occurred in 19 patients (venous thrombosis n = 15, arterial thrombosis n = 7). In 9 of the 15 venous thrombotic events no treatment was used. Four venous thrombotic events occurred during anticoagulant therapy and two during Ascal use. Regarding those on anticoagulation, it was documented that the International Normalized Ratio was sub-therapeutic in two of the four patients. Arterial thrombosis occurred during anticoagulant therapy in five out of seven events. One event occurred during Ascal use and in another event no therapy was used. Sites of recurrent thrombosis were: pulmonary embolism (n = 3), superior mesenteric vein (n = 2), lower limb (n = 4), upper limb (n = 4), sagittal sinus (n = 1), mesocaval shunt (n = 2), intestinal ischemia (n = 2) and ischemic stroke (n = 4). The median time between diagnosis of PVT and occurrence of a recurrent thrombotic event was 5.7 years (range 0.1–19.0 years; IQR of 1.8–8.6 years). The overall risk of a recurrent thrombotic event was 3% (95% CI 0–7) at 1 year, 8% (95% CI 3–14) at 5 years and 24% (95% CI 13–36) at 10 years (Fig. 1). Seventy-four percent of the recurrent thrombotic events occurred in patients with a prothrombotic disorder. Sixteen recurrent thrombotic events were found in 69 patients (23%) with a prothrombotic disorder versus six events in 51 patients (11%) without a prothrombotic disorder (P = 0.03). The use of anticoagulants tended to reduce the occurrence of a recurrent venous thrombotic event (HR 0.2, P = 0.1). Predictors of a recurrent thrombotic event are shown in Table 2.
Table 2. Univariate analysis of variables associated with bleeding, recurrent thrombosis and survival
HR (CI 95%)
HR (CI 95%)
HR (CI 95%)
*Statistical significant. †10-year increment. ‡According the Paquet Classification . §Analyzed as a time-dependent factor. ¶There were not enough bleeding events observed to assess the effect of IBD. HR, hazard ratio; CI, confidence interval; MNP, myeloproliferative neoplasms; IBD, inflammatory bowel disease.
A total of 83 bleeding episodes (variceal bleeding n = 52 and other gastro-intestinal bleeding n = 31) occurred in 37 patients during follow-up. Fifty bleeding episodes occurred during anticoagulation therapy. Causes of other gastro-intestinal bleeding events were: gastric erosion (n = 7), hemorrhoid (n = 7), angiodysplasia (n = 4) duodenal and jejunal ulcer (n = 2), esophageal ulcer due to sclerotherapy (n = 2), arteriovenous malformation (n = 2), Mallory Weiss lesion (n = 2), other (n = 2) and unknown in the absence of varices (n = 3). Sixteen patients had a single episode of bleeding, 11 patients experienced two episodes and 10 patients had three or more bleeding episodes (range 3–9 events). The median time until the first bleeding event occurred was 7 months after PVT diagnosis was established (range 0.0–11.5 years; IQR 0.2–3.0 years). The overall risk of gastro-intestinal bleeding was 33% (95% CI 24–41) at 1 year, 43% (95% CI 33–53) at 5 years and 46% (95% CI 36–56) at 10 years (Fig. 2). The median time until a re-bleeding event occurred was 4 months after the initial bleeding (range 0.01–8.6 years; IQR 0.9–8.0 months). The overall risk of re-bleeding was 46% (95% CI 36–56) at 1 year, 63% at 5 years (95% CI 52–74) and 69% at 10 years (95% CI 59–82) (Fig. 2). Predictors of (re)bleeding in the univariate analysis are shown in Table 2. In the multivariate analysis gastrointestinal bleeding at baseline (HR 2.1, P < 0.01), ascites at baseline (HR 2.0, P = 0.01) and the use of anticoagulant therapy (HR 2.0, P < 0.01) were significant predictors of gastrointestinal (re)bleeding. Patients with anticoagulant therapy (n = 66) experienced 58 bleeding episode vs. 25 in patients without anticoagulant therapy (n = 54). There was no significant relation between the severity of the gastrointestinal bleeding and the use of anticoagulants. The median hemoglobin level at admission was 5.5 mmol L−1 (range 1.8–8.7 mmol L−1) in patients on anticoagulant therapy vs. 5.8 mmol L−1 (range 3.0–8.5 mmol L−1) in patients without anticoagulant therapy. The median transfused packed red blood cells per bleeding episode was four (range 2–20) in both groups. In total there were seven admissions to the ICU in the group of patients on anticoagulation therapy vs. five in the other group.
Twenty-nine patients (24%) died during follow-up. The median age at death was 64.2 years (range 30.4–95.3 years). Overall survival was 90% (95% CI 84–96) and 70% (95% CI 58–82) at 5 and 10 years, respectively (Fig. 3). The predictors of survival in the univariate analysis are presented in Table 2. In the multivariate analysis, increased age (HR 1.1, P < 0.01) and ascites at diagnosis (HR 4.0, P < 0.01) were the only significant factors associated with increased mortality. In the time-dependent covariate analysis a recurrent thrombotic event was significantly associated with poor survival (HR 3.1, P = 0.02). Gastrointestinal bleeding (HR 1.6, P = 0.2) and anticoagulant use both had an effect (HR 0.5, P = 0.2) on survival, but was not statistically significant. Five patients died as a result of a bleeding event: cerebellar hematoma (n = 1), variceal bleeding (n = 2) and other upper gastro-intestinal bleeding (n = 2). The last two patients were both on anticoagulant therapy. Three patients died owing to a recurrent thrombotic event (necrotic colon due to massive arterial ischemia, ischemic cerebral vascular accident and sagittal sinus thrombosis). The last two patients were on anticoagulant therapy when the thrombotic events occurred. Other causes of death were progressive MPN (n = 6), infection (n = 3), other causes (n = 7) and unknown (n = 5).
In this study, we investigated the role of anticoagulant therapy and other factors associated with recurrent thrombotic events, gastrointestinal bleeding and survival in patients with non-cirrhotic non-malignant PVT. In approximately 20% of the patients with an underlying prothrombotic disorder a recurrent thrombotic event occurred, which was significantly associated with decreased survival. Anticoagulant therapy showed a tendency to prevent recurrent thrombotic events; however, it also significantly increased the risk of gastrointestinal bleeding.
The duration of anticoagulant therapy in patients with PVT is still a matter of debate. Although recent consensus guidelines state that prolonged anticoagulation should be considered in patients with an underlying prothrombotic state, these recommendations are based on limited data . The aim of anticoagulant treatment in PVT is to minimize thrombosis extension, prevent recurrent thrombotic events and, if possible, induce recanalization. In the present study, we focused on recurrent thrombotic events and bleeding complications only, as recanalization was not systematically investigated in our study population. However, the question regarding recanalization has been recently addressed in a prospective European study. In a cohort of 102 patients with recent PVT, it was demonstrated that recanalization occurred in one-third of the cases receiving early anticoagulation . These results confirm findings from earlier, retrospective series showing a benefit of anticoagulant therapy with respect to recanalization in cases with recent PVT [14, 15, 20].
In this study, we found that patients with PVT and a prothrombotic disorder had a three-fold increased risk of developing recurrent thrombotic events, either in or outside the splanchnic vascular bed. Extension of thrombosis into the splanchnic veins occurred only in patients with an underlying prothrombotic disorder. Particularly patients with an underlying MPN appear to have a high risk of recurrent thrombotic events. In a previous study, 27% of cases with PVT and concomitant MPN experienced recurrent thrombosis . Our results show that treatment with anticoagulation tended to have an overall beneficial effect on decreasing recurrent thrombotic events and even somewhat on improving survival. Although the effect of antiplatelet drugs could not be systematically investigated in this study, there was no recurrent thrombosis in patients with PVT and MPN treated with aspirin, highlighting that in this subgroup of patients, treatment with antiplatelet drugs could be beneficial . On the other hand, the use of anticoagulant therapy significantly increased the risk of gastrointestinal bleeding. A total of four patients (3%) died as a result of gastrointestinal bleeding, two of which were on anticoagulation therapy.
Previous studies have shown that the prevalence of gastrointestinal (re)bleeding is mainly determined by size of varices and initial presentation with gastrointestinal bleeding [12, 19, 24]. In addition to these factors, we found that both extension of thrombosis into the splanchnic veins and the presence of ascites at baseline were significant factors predicting (re)bleeding. These new findings may help to identify patients at a high risk of gastrointestinal (re)bleeding and also may influence whether to start anticoagulant therapy. In an earlier study, Condat et al.  concluded that the benefit-to-risk ratio favors anticoagulant therapy for most patients with PVT. This conclusion was based on two major findings. First, the risk of recurrent thrombotic events was profoundly reduced by anticoagulant therapy. This reduction mainly involved thrombotic recurrence or extension within the portal venous system. Second, there was no significant correlation between the risk and severity of gastro-intestinal bleeding and the use of anticoagulant therapy in their cohort. With our study we support the current guidelines which state that anticoagulation therapy should be considered in patients with PVT and an underlying prothrombotic disorder [16, 17]. Otherwise these patients may be at high risk of developing local recurrence or extension of thrombosis. However, our findings demonstrate that anticoagulant therapy is not without risk and increases the risk of gastrointestinal bleeding. This is in line with a recent study analyzing a large cohort of patients with splanchnic vein thrombosis that showed that warfarin therapy was an independent predictor of bleeding .
Our study reports on one of the largest cohorts of patients with non-cirrhotic PVT followed-up over a long period of time. This has allowed us to specifically study the long-term effects of anticoagulant therapy in this patient group. Nevertheless, there are also several limitations, in part as a result of the retrospective nature of our study. Decisions concerning anticoagulant treatment, performing radiological imaging and the choice of prophylaxis treatment were made by the treating physician. Endoscopic therapy was the preferred method for secondary prophylaxis . Patients with an extensive thrombosis are probably more likely to be prescribed anticoagulant therapy than others, in particular patients presenting with a gastro-intestinal bleeding. Both factors were carefully taken into account in the uni- and multivariate analysis. Furthermore, during the long study period new insights and diagnostic tests have been developed. This may have led to an underrepresentation and undertreatment of patients with prothrombotic disorders enrolled at the beginning of our study. In parallel, most of these patients were initially not on anticoagulant therapy. As patients could stop and restart anticoagulant therapy during the course we analyzed each bleeding and thrombotic event according to the anticoagulant therapy used at that time.
In conclusion, in patients with non-cirrhotic PVT recurrent thrombotic events occur primarily in patients with an underlying prothrombotic disorder and are significantly associated with decreased survival. Anticoagulant therapy can lead to the prevention of recurrent thrombosis, but also significantly increases the risk of gastrointestinal bleeding. These findings imply that anticoagulant therapy on one hand is warranted in patients with PVT and an underlying prothrombotic disorder, whereas on the other hand it should be given with caution to those with high risk of bleeding.
Study concept and design: M. C. W. Spaander, H. L. A. Janssen; acquisition of data: M. C. W. Spaander; analysis and interpretation of data: M. C. W. Spaander, J. Hoekstra, B. E. Hansen, F. W. G. Leebeek, H. L. A. Janssen; drafting of the manuscript: M. C. W. Spaander; critical revision of the manuscript for important intellectual content: J. Hoekstra, H. R. Buuren, F. W. G. Leebeek, H. L. A. Janssen; statistical analysis: M. C. W. Spaander, B. E. Hansen; technical, or material support: H. L. A. Janssen; study supervision: H. L. A. Janssen; has approved the final draft submitted: M. C. W. Spaander, J. Hoekstra, B. E. Hansen, H. R. Buuren, F. W. G. Leebeek, H. L. A. Janssen; obtained funding: No.
Disclosure of Conflict of Interests
The authors state that they have no conflict of interest.