Article Text

This article has a correction. Please see:

Download PDFPDF

Deficiency of natural anticoagulant proteins C, S, and antithrombin in portal vein thrombosis: a secondary phenomenon?
  1. N C Fishera,
  2. J T Wildeb,
  3. J Roperb,
  4. E Eliasa
  1. aLiver Unit, Queen Elizabeth Hospital, Birmingham, UK, bDepartment of Haematology, Queen Elizabeth Hospital
  1. Dr N C Fisher, Department of Gastroenterology, Wordsley Hospital, West Midlands DY8 5QX, UK


BACKGROUND Hereditary deficiencies of natural anticoagulant proteins are implicated in the pathogenesis of portal vein thrombosis (PVT). Secondary deficiencies of these proteins have also been reported in PVT, making interpretation of concentrations difficult.

AIMS To characterise the coagulation profiles in adult patients with PVT and to investigate the possible mechanisms of natural anticoagulant protein deficiency.

PATIENTS Twenty nine adult patients with portal hypertension caused by PVT, and normal biochemical liver function tests.

METHODS Routine coagulation profiles and concentrations of proteins C, S, and antithrombin were measured; where indicated, corresponding concentrations in parents were also measured. Synchronous peripheral and hepatic or splenic vein concentrations were compared in seven patients undergoing interventional procedures, as were peripheral concentrations before and after shunt surgery in three patients.

RESULTS Deficiencies of one or more of the natural anticoagulant proteins occurred in 18 patients (62%), with six patients having combined deficiency of all three proteins. There were strong correlations between prothrombin and partial thromboplastin time ratios and concentrations of natural anticoagulant proteins. Family studies in nine cases of anticoagulant protein deficiency revealed possible hereditary deficiency in only three cases, and significantly lower concentrations of anticoagulant proteins in all PVT cases compared with parents. Levels of anticoagulant proteins tended to be lower in hepatic veins but higher in splenic veins compared with peripheral vein concentrations. Peripheral concentrations decreased after shunt surgery.

CONCLUSIONS Deficiency of natural anticoagulant proteins is common in PVT and is probably a secondary phenomenon in most cases, occurring as part of a global disturbance of coagulation variables. The mechanism for this remains unclear but may result from a combination of reduced hepatic blood flow and portosystemic shunting itself.

  • portal vein thrombosis
  • extrahepatic portal hypertension
  • natural anticoagulant protein
  • protein C
  • protein S
  • antithrombin

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Portal hypertension owing to portal vein thrombosis (PVT) usually presents as unexpected variceal haemorrhage or splenomegaly with peripheral blood cytopenias. There may be a history of abdominal inflammatory disease with or without a prothrombotic condition such as pregnancy or oral contraceptive usage. Regardless of whether or not there is an associated precipitant, patients presenting with PVT should also be investigated for an underlying thrombophilic condition such as a myeloproliferative disorder or a hereditary thrombophilic state. Hereditary thrombophilias that are known to predispose to PVT include certain mutations of the prothrombin or factor V genes, or deficiency of one of the natural anticoagulant proteins C, S, or antithrombin.1-5 However, diagnosis of hereditary natural anticoagulant protein deficiency is difficult if there is impaired liver function. Furthermore, reduction in circulating concentrations of natural anticoagulant proteins has been reported in patients with PVT, even in the presence of biochemically normal liver function and it has been suggested that in many patients these deficiencies may be acquired as a consequence of the PVT.6 7 Thus individuals with reduced concentrations of natural anticoagulant proteins in this setting cannot be assumed to have a hereditary deficiency state. The aims of this study were to characterise the coagulation profiles and concentrations of natural anticoagulant proteins in a cohort of adult patients with PVT; to determine whether deficiencies, if detected, are primary or secondary; and to elucidate the possible mechanism of acquired natural anticoagulant protein deficiencies.

Patients and methods

Case records of 42 adult patients with radiologically proved extrahepatic PVT or non-cirrhotic portal hypertension with intrahepatic portal sclerosis (NCPHT), a disease which probably has a similar pathogenesis to extrahepatic PVT, were analysed.8 9 The following patients were excluded: eight patients with abnormal liver function as defined by persistently abnormal serum albumin or aspartate aminotransferase (AST) concentrations (normal ranges: at least 36 g/l, and 40 U/l or less, respectively); five patients who had undergone portosystemic shunts prior to investigation, as we later discovered that therapeutic portosystemic shunting influences concentrations of coagulation proteins.

Twenty nine patients were therefore included in the study, of whom 26 had extrahepatic PVT and three had NCPHT. All patients had negative serological tests for viral or autoimmune hepatitis, genetic haemochromatosis, and Wilson's disease; alcoholic liver disease was excluded by careful history and liver histology. The patients were further categorised as follows: group 1—childhood presentation and/or history of umbilical vein sepsis (n=5); group 2—adult presentation without any history of abdominal surgery or inflammatory disease (n=13, including two patients with NCPHT); and group 3—adult presentation with known history of pancreatitis and/or upper abdominal surgery (n=11, including one patient with NCPHT). The median age at presentation in these groups was: group 1, 13 years (range 1–24); group 2, 35 years (range 17–56); group 3, 41 years (range 23–63).

Two patients in group 2 were using oral contraceptives and two patients in group 3 had known myelofibrosis prior to presentation. The modes of presentation were variceal haemorrhage (n=21), splenomegaly and/or cytopenia (n=5), ascites (n=2), or transient jaundice owing to choledochal varices (n=1). All patients were over 16 years of age at the time of this study; the median delay between first presentation and laboratory investigation was five years (range 0–26).

Contrast angiography was performed to confirm and document extent of portal vein thrombosis in most patients; in a minority of cases the diagnosis was made by non-invasive imaging with colour Doppler ultrasonography and/or magnetic resonance angiography. The extent of thrombosis of portal, mesenteric, and splenic veins was documented. Liver histology was obtained in 21 patients and was reported normal or near normal in all patients apart from three; in two of these there was fibrous expansion of portal tracts and in the third patient (who had jaundice owing to choledochal varices) there was histological evidence of cholestasis but no other abnormality. Patients were managed by variceal sclerotherapy or banding alone (n=12), surgical splenorenal shunting (n=3), splenorenal or portocaval shunting with splenectomy (n=1 each), splenectomy alone (n=3), transjugular intrahepatic portosystemic shunt (TIPSS) insertion (n=2), and medical treatment alone or no intervention (n=7).


The following coagulation variables were measured for this study (normal ranges given in parentheses where applicable): full blood count including platelet count (150–400 × 109/l), prothrombin (PT) ratio (1.2 or less), activated partial thromboplastin (APTT) ratio (0.8–1.2), protein C activity (660–1220 U/dl), free protein S antigen (680–1460 U/dl), antithrombin activity (750–1400 U/dl), activated protein C (APC) resistance ratio (more than 2.6), and lupus anticoagulant screen using the dilute Russell's viper venom (DRVV) assay ratio (less than 1.1). Molecular analysis for the factor V Leiden mutation was performed in 26 patients, and for the prothrombin gene (G20210A) mutation in 10 patients (representing one, four, and five patients from groups 1, 2, and 3 respectively). Myeloproliferative disease was sought by bone marrow biopsy if suspected by peripheral blood count and blood film analysis. In nine patients with increased PT ratios, concentrations of factor VII (normal range 500–1500 U/dl), fibrinogen (1.5–4.0 g/l), and D-dimers (less than 250 ng/ml) were measured, in order to determine whether this was owing to procoagulant protein deficiency or disseminated intravascular coagulation.

In the majority of cases two measurements of each variable were made at different timepoints, with a mean taken of the two values obtained. Protein C and antithrombin were measured by chromogenic assay (Immuno Ltd, Austria); protein S was measured by enzyme linked immunosorbent assay (ELISA; Dako Ltd, UK). APC resistance was measured by the Coatest assay (first generation assay, Chromogenix AB, Sweden). Factor V and prothrombin gene analyses were performed using in house polymerase chain reaction (PCR) techniques. Fibrinogen was measured by Clauss (Immuno Ltd), factor VII by a one stage clotting based technique, and D-dimers by a latex-agglutination serial dilution assay (Accuclot, Sigma Ltd, UK).


If a deficiency of any of the natural anticoagulant proteins C, S, or antithrombin was found, blood samples were obtained from both parents of the index cases where possible, to determine whether the deficiency was hereditary.


In selected patients who underwent transjugular liver biopsy (n=3), TIPSS insertion (n=1), or splenorenal shunts (n=3), synchronous blood samples were taken from peripheral, hepatic, and splenic veins where possible for measurement of coagulation profiles in order to determine whether differences could be detected in different splanchnic venous beds in comparison with peripheral veins. For comparative purposes, synchronous peripheral and hepatic vein blood samples were also analysed from 15 patients with established parenchymal liver disease (mostly because of alcoholic cirrhosis) with portal hypertension undergoing transjugular liver biopsy or TIPSS insertion.


Three patients underwent elective splenorenal shunt surgery (with preservation of the spleen) for definitive management of varices during the study period. In all cases, there was no active variceal haemorrhage at the time of surgery. In these patients coagulation profiles and natural anticoagulant protein concentrations were measured before and at least two months after surgery in order to determine whether portosytemic shunting influenced these values.


Statistical analysis was done with SPSS statistical software, using the Mann-Whitney U test, Wilcoxon rank sum, Kruskal-Wallis, and Spearman correlation tests. A p value of less than 0.05 was considered significant. Informed consent for blood sampling was sought and given by all patients for the purposes of this study.



A total of 18 (62%) patients had a deficiency of one or more of the natural anticoagulant proteins as defined by our normal ranges. Furthermore, values of anticoagulant proteins were towards the lower end of the normal range in most patients (fig 1). There were deficiencies of protein C and antithrombin in 12 patients (41%) each and protein S in 11 patients (38%), with overlapping of deficiencies in several patients. Thus there were eight cases (28%) of combined proteins C and S deficiency, nine cases (31%) of protein C and antithrombin deficiency, seven cases (24%) of protein S and antithrombin deficiency, and six cases (21%) of combined deficiency of all three proteins. An increased PT ratio of at least 1.3 was present in 19 patients (66%) and an increased APTT ratio of at least 1.2 was present in eight (28%). There were significant positive correlations between concentrations of the different anticoagulant proteins in individual patients, and significant negative correlations with corresponding PT and APTT ratios, but no significant correlations with albumin concentrations. Table 1 summarises these data.

Figure 1

Scatterplot of protein C, S, and antithrombin concentrations in 29 patients with portal vein thrombosis. The box outlines indicate normal ranges (protein C: 660–1220 U/dl; protein S: 680–1460 U/dl; and antithrombin: 750–1400 U/dl).

Table 1

Spearman correlation coefficients (r values) for anticoagulant proteins, prothrombin time (PT) and acitivated partial thromboplastin time (APTT) ratios, and albumin in patients with portal vein thrombosis

When natural anticoagulant concentrations in the different patient groups were compared, there were no consistently significant differences found between groups. However, there was a trend towards lower concentrations of each protein in group 2 compared with group 3; median protein C concentrations were 660 and 880 U/dl respectively (p=0.02, Mann-Whitney U test), median protein S concentrations 660 and 750 U/dl respectively (p=0.12), and median antithrombin concentrations 690 and 840 U/dl respectively (p=0.06). When anticoagulant concentrations were analysed according to the presence or absence of additional mesenteric and/or splenic vein thrombosis in patients with extrahepatic PVT, no consistent differences or trends were found (data not shown).

In patients without overt myeloproliferative disease (n=25) platelet counts were low in 19/25 cases (median level 103 × 109/l, range 20–322). Factor VII concentrations were measured in nine patients with increased PT ratios and were low in five cases (median level 490 U/dl, range 350–750). Fibrinogen concentrations were normal in all nine of these patients (median 2.9 g/l, range 1.8–4.2) and D-dimer concentrations were normal in all except two (concentrations 500 and 1000 ng/ml respectively). Lupus anticoagulant and the factor V Leiden mutation were detected in one patient each and one further patient had APC resistance in the absence of the factor V Leiden mutation. The prothrombin gene mutation was not found in any of the 10 patients tested. Two new cases of myeloproliferative disease (both essential thrombocythaemia) were diagnosed by bone marrow analysis at the time of presentation.


Complete family studies of proteins C, S, and antithrombin concentrations for both parents of an index case of PVT with a deficiency of at least one of these proteins were available in nine cases. Figure 2 illustrates these data. All index cases had deficiency of at least one natural anticoagulant protein and some had combined deficiencies; in total there were six cases of protein C deficiency, three of protein S deficiency, and five of antithrombin deficiency. In contrast only three cases of natural anticoagulant protein deficiency were identified in corresponding parents: one had protein C deficiency and two had antithrombin deficiency. All were parents of different cases and all were associated with corresponding deficiency of the same protein as the index case (the family with protein C deficiency had a history of thromboembolic disease in other siblings).

Figure 2

Scatterplot of protein C, S, and antithrombin concentrations in nine patients with portal vein thrombosis (index cases) and corresponding concentrations in parents. Comparison of concentrations in the index cases with the lower of the corresponding values in parents showed significantly lower concentrations in index cases (protein C, p<0.01; protein S, p<0.05; and antithrombin, p<0.05, Wilcoxon rank sum).

When concentrations of natural anticoagulant proteins in all index cases and parents were compared, these were significantly lower in the index cases (p<0.001 for each of proteins C, S, and antithrombin, Mann-Whitney U test). As deficiency of natural anticoagulant proteins is usually inherited heterozygously,10 concentrations in index cases were also compared with whichever was the lower of the two corresponding values in parents. For protein C the median value in index cases expressed as a percentage of the lower of the corresponding values in parents was 55% (range 38–80%, p<0.01, Wilcoxon rank sum); for protein S the corresponding percentage was 72% (range 63–122%, p<0.05); and for antithrombin the corresponding percentage was 78% (range 39–119%, p<0.05).


Table 2 shows individual concentrations of proteins C, S, and antithrombin in hepatic and splenic veins, expressed as a percentage of synchronous peripheral vein concentrations in seven patients, together with mean concentrations from 15 patients with portal hypertension caused by parenchymal liver disease. Results were not uniformly consistent but hepatic vein concentrations of each of the natural anticoagulant proteins tended to be lower compared with peripheral vein concentrations (p=0.27, Wilcoxon rank sum). In contrast, splenic vein concentrations tended to be higher than peripheral vein concentrations (p=0.14, Wilcoxon rank sum). In patients with parenchymal liver disease, hepatic vein concentrations were lower than peripheral vein concentrations and in these patients more observations were available (p<0.01, Wilcoxon rank sum).

Table 2

Concentrations of proteins C, S, and antithrombin in different vascular beds


There were significant reductions in the concentrations of proteins C, S, and antithrombin postoperatively in each of the three patients who underwent splenorenal shunting (p<0.05, Wilcoxon rank sum) and corresponding significant increases in the PT and APTT ratios (p<0.05). In contrast the serum albumin level fell slightly in only one patient. There were no clinical complications postoperatively in any patient and none had been treated with anticoagulants. There was a trend towards subsequent “recovery” of these values in one patient in whom serial postoperative values were available. Figure 3illustrates these cases.

Figure 3

Levels of proteins C (PC), S (PS), antithrombin (AT), and albumin, with prothrombin time (PT) and activated thromboplastin time (APTT) ratios, before and after surgery in three patients undergoing splenorenal shunting (with preservation of the spleen) for portal vein thrombosis. (A) Patient 1, (B) patient 2, (C) patient 3.


This study confirms that single or combined deficiencies of natural anticoagulant proteins are a common finding in PVT and suggests that the majority of deficiencies are acquired, presumably as a consequence of PVT, and not because of a hereditary genetic defect. However, a minority of cases of PVT may have a true underlying hereditary anticoagulant protein deficiency and this can only be confirmed by careful investigation of family members, preferably including both parents. In this series a diagnosis of probable hereditary anticoagulant protein deficiency was made in three of nine cases investigated. An alternative way to make this diagnosis where parental studies are not possible might be by screening of siblings, which could be used for both diagnostic and counselling purposes. Lastly, recent usage of gene sequencing in the elucidation of anticoagulant protein gene mutations may now provide the potential to determine whether such anticoagulant deficiencies in PVT are truly primary or not.11

Many cases of PVT are associated with increased PT and APTT ratios, and the low factor VII concentrations we found in these patients indicates that these abnormalities are likely to be caused by true reductions in procoagulant proteins. The strong correlation of PT and APTT ratios with natural anticoagulant protein concentrations suggests that a similar mechanism may underlie both abnormalities, involving reduction in both procoagulant and anticoagulant proteins. This mechanism remains unclear but is probably multifactorial. Based on our own observations and those of others we propose two possible complementary mechanisms which are summarised in fig 4. Firstly, reduction in liver blood flow following PVT probably leads to a degree of hepatic atrophy; we have noted that many patients with PVT have small livers on imaging. This would probably lead to reduced hepatic protein synthesis, perhaps more selectively for coagulation factors compared with albumin in most cases (in support of this we have also managed several patients with PVT excluded from this study who had reduced concentrations of albumin in addition to coagulation factors). Secondly, shunting of blood from the liver may in itself lead to reduced peripheral concentrations of coagulation proteins, possibly owing to clearance or consumption. While we and others did not find definitive evidence for disseminated intravascular coagulation in patients with PVT,12 others have nevertheless reported changes consistent with a mild and compensated consumption coagulopathy in PVT and in experimental models of portosystemic shunting.6 13 Portosystemic shunting will clearly occur in most or all cases of PVT and will be increased following any subsequent surgical shunt procedure. Thus the further reduction in anticoagulant proteins that we and others have observed following shunt surgery may in part be a consequence of therapeutic shunting per se, although this may also reflect further reduction in hepatic blood flow.7

Figure 4

Proposed mechanism for reduction in concentrations of procoagulant and anticoagulant proteins in patients with portal vein thrombosis.

Our finding of a trend towards lower concentrations of anticoagulant proteins in hepatic veins compared with peripheral veins in patients with PVT or intrinsic liver disease with portal hypertension is contrary to the assumption that these proteins are produced exclusively within the liver. This finding also supports the possibility that these proteins may be cleared or consumed within the portal circulation in portal hypertension (whether caused by PVT or intrinsic liver disease), although this is difficult to ascertain in the absence of comparative data from healthy individuals. Furthermore, compensatory synthesis of these proteins occurs in other vascular beds including the spleen.14 15

Our case series also highlights some of the dilemmas in managing patients with PVT who have presented with variceal haemorrhage, and who may have an underlying thrombophilia state. Our usual practice is to eradicate oesophageal varices if these have bled, using adjunctive propranolol therapy to lower portal pressure. In cases of severe oesophageal or gastric variceal haemorrhage a portosytemic shunt may be chosen. Good long term results from either of these approaches have been reported by others.16 17 A potential beneficial role for warfarin therapy in PVT remains debatable. We do not routinely warfarinise patients with PVT because of potential facilitation of variceal haemorrhage, although in our series some patients were warfarinised after shunt surgery because of associated myeloproliferative disease or previous shunt thrombosis in one case. In our experience it is rare for patients with PVT to develop clinically obvious extension of splanchnic thrombosis or any other thromboembolic disease. It is possible that the balance of coagulation abnormalities that frequently occur following PVT may favour reduction in any pre-existing thrombotic tendency. A recent retrospective case series did not show any significant alteration in outcome with warfarin therapy in PVT and further study is required before definite conclusions on the role of anticoagulation can be made.18


We are very grateful to fellow physicians and surgeons at the Liver Unit for allowing us to study patients under their care. We are also very grateful to Dr S P Olliff, consultant radiologist, and to Mr J A C Buckels, consultant surgeon, for help in obtaining splanchnic vein blood samples.



  • Abbreviations used in this paper:
    non-cirrhotic portal hypertension
    prothrombin time
    partial thromboplastin time
    portal vein thrombosis

Linked Articles

  • Correction
    BMJ Publishing Group Ltd and British Society of Gastroenterology