Correspondence Oren Shibolet, MD, Liver Unit, Division of Medicine, Hadassah-Hebrew University Medical Center, PO Box 12000, Jerusalem 91120, Israel Tel: 972-2-6778253 Fax: 972-2-6420338 e-mail: firstname.lastname@example.org
Transjugular intrahepatic portosystemic shunt (TIPS) is in widespread use for the decompression of portal pressure. The entity of persistent TIPS infection, also known as ‘endotipsitis’ is a rare but serious complication of TIPS insertion. The exact definition of ‘endotipsitis’ is still debated, but involves persistent bacteremia and fever together with either shunt occlusion, or vegetation, or bacteremia in the presence of a patent shunt, when other sources of bacteremia have been ruled out. To date, approximately 40 cases of ‘endotipsitis’ have been described, with predominance for male and alcoholic hepatitis patients. The clinical course is variable, but fever and chills are a constant feature. Bacteremia, can either occur early (<120 days) or late (>120 days) after stent insertion, with some cases occurring many years after the procedure. Although no predominant bacterial species have been described in ‘endotipsitis’, staphylococci and other Gram-positive bacteria are more commonly seen in early infection. The diagnosis of ‘endotipsitis’ is difficult and requires a high index of suspicion. A rigorous imaging work-up to rule out other sources of endovascular infection is usually required including ultrasonography, computed tomography and echocardiography. Because removal of the infected stent is impractical, treatment is empirical and based on a prolonged course of antibiotics. If eligible, some patients may be referred for liver transplantation. The use of prophylactic antibiotics during the initial TIPS procedure is controversial, and despite the lack of evidence, prophylaxis is the common practice. The aim of this review was to describe the definition, clinical course, diagnosis, pathogenesis, microbiology, treatment and outcome of endotipsitis.
Transjugular intrahepatic portosystemic shunting (TIPS) is commonly used for the decompression of portal pressure. TIPS was first described as a research technique in 1969 and gained clinical acceptance in the early 1990s (1, 2). It is widely used to treat variceal bleeding (3–5), and refractory ascites (6–8), but the list of indications continues to expand and includes, Budd–Chiari syndrome (9, 10), hepatorenal syndrome (11, 12), veno-occlusive disease (13), hepatic hydrothorax (14), non-variceal bleeding and pre-operative portal decompression (15). As endoscopic, mechanical and pharmacological treatments provide adequate control in the majority of patients with complications of portal hypertension, TIPS is currently indicated as a ‘rescue’ measure after other procedures have failed to control bleeding and ascites (16, 17). The expansion in the use of TIPS led to the recognition of multiple complications. These include systemic procedure-related complications such as hepatic failure, encephalopathy, sepsis and death. Local procedure-related complications include cardiac tamponade, laceration of vessels with bleeding and pneumothorax, as well as those associated with the endoprosthesis itself such as migration, misplacement and occlusion (18–22).
Sustained bacteremia associated with the endovascular infection of TIPS stents is a rare and serious complication of the procedure. The term endotipsitis has been proposed to define this entity by Sanyal and Reddy (23).
The aim of this review was to describe the definition, clinical course, diagnosis, pathogenesis, microbiology, treatment and outcome of endotipsitis.
We performed a MEDLINE database search from January 1965 to January 2009, using the terms ‘bacteremia’ or ‘infection’ or ‘endotipsitis’ or ‘endovascular infection’ combined with the term transjugular intrahepatic portosystemic shunt or TIPS, and also checked the Cochrane library and references for reviews about TIPS and endotipsitis. Reference lists from all available manuscripts were manually searched for additional reports.
The exact definition of endotipsitis is still being debated. The initial definition was suggested by Sanyal and Reddy (23) and included:
Definite infection: Clinically significant continuous bacteremia (fever and multiple positive blood cultures), with vegetations or thrombus inside the TIPS.
Probable infection: Sustained bacteremia and unremitting fever in a patient with an apparently patent TIPS and no other obvious source of infection.
Blood cultures were considered to represent clinically significant bacteremia when they were positive from two isolated, separate venipunctures. Stent thrombosis was defined as occlusion or near occlusion of the stent, whereas vegetations were defined as smaller non-occlusive lesions adherent to the stent wall.
DeSimone et al. (24) later used stricter criteria to define sustained bacteremia as two blood cultures drawn >12 h apart that were positive for the same organism; or, a majority of four or more blood cultures that were positive for the same organism (first and last cultures drawn at least 1 h apart) without another identifiable source of bacteremia.
In a later report, Armstrong and MacLeod described three patients with endotipsitis and prolonged bacteremia. In an effort to create a more specific standardized definition, the authors suggested that sustained bacteremia be defined as two or more blood cultures positive for the same organism, the first and last being separated by 7 days or more. However, this definition has been contested by others (25).
As TIPS is used to decompress portal pressure, it is almost invariably used in patients with advanced liver disease. Our search identified 36 patients with endotipsitis. Twenty-eight were males and eight females. The age of the patients was reported in 29 cases with a mean of 49 years (range 43–69 years). Underlying liver diseases included alcoholic cirrhosis (n=23), hepatitis C virus (n=9), cryptogenic cirrhosis (n=7) and hepatitis B virus (n=2), primary sclerosing cholangitis (n=1) and diagnosis not reported (n=1). Several patients had multiple underlying liver diseases. It is unclear whether the male and alcoholic cirrhosis predominance mentioned above represented a true association or is because of selection and publication bias.
The indication for TIPS was reported in 23 patients (64%). Twenty-one patients suffered from upper gastrointestinal bleeding, and five from ascites. Three patients had both bleeding and ascites. The outcome was reported in 35 patients. In 25 patients the infection resolved, whereas 10 patients died of disease complications. In one case the outcome was not reported.
The population of patients in our review came from several case reports and case series, and therefore could not be used to predict the risk of endotipsitis in the general population of patients undergoing TIPS. Furthermore, although the majority of patients with endotipsitis underwent TIPS for gastrointestinal bleeding and had underlying alcoholic cirrhosis, we were unable to determine the effects of the indication for TIPS and the underlying liver disease on the incidence of endotipsitis, as this was not alluded to in the literature.
Of interest is the question of whether the use of covered stents alters the prevalence, course or susceptibility to endotipsitis. The use of polytetrafluoroethylene-covered stents was first reported in 1995 (26); however, their widespread use has only been recent. We were unable to discern their impact on the incidence of endotipsitis as their use was not reported in the early case series of endotipsitis. We found one case report in which a covered stent was used to seal a TIPS/biliary fistula. The patient was reported to be in a stable condition on antibiotic treatment at the time of publication (27). Covered stents have been shown to improve TIPS patency, as well as reduce the rate of encephalopathy and the need for re-intervention (28, 29). It is, therefore, possible that they lower susceptibility to the development of endotipsitis.
While fever and transient bacteremia have been described in 2–25% of the patients after a TIPS procedure, the overall incidence of endotipsitis is unknown because of its rarity, the lack of a uniform definition and the lack of data concerning the overall number of TIPS procedures performed (30, 31). We identified 36 patients with endotipsitis in the literature. Most were individual case reports; however, there were four case series that reported the overall number of TIPS procedures with an incidence of approximately 1.5% (range 0.6–5.5%) (23–25, 32). The time to the initial diagnosis of sustained bacteremia after TIPS insertion was described as variable and ranged from a few days to a few years. In the two largest series in the literature, the average time span from TIPS placement to the onset of symptoms was 284 and 210 days respectively (23, 33). In a review by Armstrong and MacLeod, the onset was as early as 6 or as late as 1065 days after TIPS insertion (25). On the basis of the temporal variation in disease onset, it was suggested that endotipsitis could be classified in a similar fashion as other endovascular infections, i.e. early infection if occurring within 120 days of the TIPS procedure or late infection if occurring after this period. Although the importance of this classification is unclear, it appeared that the two entities were clinically distinct both in terms of the causative agents and pathogenesis.
The clinical presentation of endotipsitis was variable. However, fever was a constant feature and usually ranged from 37.3 to 39.5 °C (23–25). The fever pattern described varied from intermittent to continuous and could be ‘erratic’. Other systemic manifestations included chills, malaise, anorexia, diarrhoea, anaemia and shock.
Increasing jaundice attributed to sepsis and liver dysfunction was described in the majority of patients. Several case reports described the formation of a biliary venous fistula following TIPS insertion as a cause of jaundice (27, 34, 35). Hepatomegaly, increasing ascites and recurrent variceal bleeding were also described and attributed to TIPS occlusion (36, 37). Other signs of bacteremia and endovascular infection have been reported and include purpuric lesions, back pain and limb weakness secondary to vertebral osteomyelitis and spondylodiscitis, and signs and symptoms associated with septic pulmonary emboli (23, 25, 38).
Laboratory findings included leukocytosis with a ‘left shift’, anaemia, increased C-reactive protein and erythrocyte sedimentation rate. Hepatic injury manifested as hyperbilirubinaemia, elevated aminotransferases and disturbed synthetic function including hypoalbuminaemia and coagulopathy.
A ‘gold standard’ definite diagnosis of endotipsitis requires removal of the stent and identifying it as the source of infection. This method is impractical as TIPS removal is only possible during liver transplantation or at autopsy. Therefore, the diagnosis was based on the above-mentioned criteria of sustained bacteremia together with an extensive work-up to rule out other sources of infection.
Several findings may support the diagnosis of endotipsitis. Quantitative differences in colony-forming units of bacteria, demonstrated in blood cultures obtained from the portal circulation near the TIPS with those from peripheral veins, strongly suggested that the TIPS was the source of infection (32, 39). TIPS stenosis or thrombosis was another strong risk factor for endotipsitis. In 17 of the 36 (47%) reported cases, thrombi or vegetations in the stent were demonstrated by ultrasound, Doppler-imaging or other methods. As mentioned above, definitive diagnosis can be obtained from culturing thrombotic material obtained from the stent removed from the liver after liver transplantation or at postmortem. Tissue cultures could also be obtained from direct endothelial biopsy performed from within the TIPS. This procedure was reported in three cases from a single centre, all of whom had in-stent vegetations (33, 40). Endothelial biopsy is not widely available and is not required for the definitive diagnosis of endotipsitis.
A strong emphasis should be placed on ruling out other sources of infection and especially other sources of endovascular and deep-seated infection such as endocarditis, central line infection, abdominal or hepatic abscess, empyema, cholangitis or osteomyelitis.
When endotipsitis is suspected, the patient work-up should begin with colour Doppler ultrasound to assess TIPS patency. Further efforts to identify the source of sustained bacteremia vary and may include urine, central venous tip and ascetic-fluid cultures. Imaging modalities should be used to rule out deep-seated infections and include: chest radiographs, computed tomography and magnetic resonance imaging. Transthoracic and/or transoesophageal echocardiography could be utilized to rule out infective endocarditis, while gallium-67 and technetium-99m bone scans could rule out abscesses and osteomyelitis respectively. Finally, upper and lower endoscopy and endoscopic retrograde cholangiopancreatography will identify TIPS/biliary fistula. Angiograms may be performed to document TIPS occlusion and may demonstrate TIPS-biliary fistula. Because of the diversity and complexity of the patients with endotipsitis, it is impractical to recommend a single diagnostic algorithm. Emphasis should be placed on a case-based diagnostic approach. A suggested algorithm is presented in Figure 1.
No distinct organism was identified as the predominant aetiological agent in endotipsitis. The causative organisms in endotipsitis from all published papers were: Enterobacteriaceae spp., in 10 patients, [Escherichia coli (n=4), Klebsiella pneumonia (n=2), Pseudomonas aeruginosa (n=2), Klebsiella oxytoca (n=1), Serratia marcescens (n=1)]; Enterococcus spp., in nine patients [Enterococcus faecalis (n=9)]; Staphylococcus spp., in seven patients [Staphylococcus aureus (n=6), Staphylococcus epidermidis (n=1)]; Streptococcus spp., in four patients; [Streptococcus sanguis (n=2), Streptococcus bovis (n=1), Streptococcus salivarius (n=1)]; anaerobic spp., in two patients [Gemella morbillorum (n=1), Bacteroides fragilis (n=1)]; Lactobacilli spp. in two patients [Lactobacillus acidophilus (n=2)]; there was a case each of Citrobacter amalonaticus, Acinetobacter calcoaceticus and Hemophilus parainfluenza. There were three cases of fungal infection [Torulopsis glabrata (n=2) and Candida albicans (n=1)] and three patients had polymicrobial infection (Table 2).
Table 2. Aetiological agents and prevalence in endotipsitis
Subtype (number of cases)
Total cases (%)
Escherichia coli (4)
Klebsiella pneumonia (2)
Pseudomonas aeruginosa (2)
Klebsiella oxytoca (1)
Serratia marcescens (1)
Enterococcus faecalis (9)
Staphylococcus aureus (6)
Staphylococcus epidermidis (1)
Streptococcus sanguis (2)
Streptococcus bovis (1)
Streptococcus salivarius (1)
Gemella morbillorum (1)
Bacteroides fragilis (1)
Lactobacillus acidophilus (2)
Torulopsis glabrata (1)
Candida albicans (1)
Citrobacter amalonaticus (1)
Acinetobacter calcoaceticus (1)
Hemophilus parainfluenza (1)
Polymicrobial infection (3)
Although no clear pattern of infection emerged from these data, no fungal infection occurred in early endotipsitis. Also, Enterobacteriaceae were uncommon causative agents in early endotipsitis with only one described case of P. aeruginosa and one case of S. marcescens occurring 35 and 30 days after TIPS respectively Thus, the most common organisms in early endotipsitis were Staphylococci and other Gram-positive bacteria (isolated in 72% of cases) and enterococci spp. (33, 36, 38, 41–43).
Transjugular intrahepatic portosystemic shunt infection is a rare entity and its pathogenesis is not well understood. It has been hypothesized that early infection is the result of bacterial seeding during the initial procedure or hospitalization. It can also be the result of technical complications of stent insertion such as early occlusion, creation of a portal vein to bile duct fistula, or the extra-hepatic passage of the needle into infected peritoneal fluid.
Late infection occurs months to years after TIPS insertion. It may occur following TIPS revision including balloon dilatation of stent stenosis. Similar to other endovascular infections, once inside the blood stream, bacteria can adhere to exposed foreign material where they are protected from antimicrobial peptides. They proliferate and cause pro-coagulant effects. Fibrin deposition and platelet aggregation cause local vegetation (25).
The histological changes occurring within the stents have not been extensively studied. Both animal and human studies have suggested that a thin layer of contiguous endothelium and fibrocellular neo-intima are formed on endovascular stents within a few weeks of insertion (44–47). In TIPS, a layer of endothelial-like cells covers the luminal surface of the stent with a layer of fibrocellular tissue forming between the stent wires and the endothelium.
It is unclear what leads to TIPS stenosis or occlusion; however, processes that appear to be involved are either neo-intimal hyperplasia, or in cases that occur following re-canalization and dilatation, shearing and embolization of neo-intima or thrombus. An inflammatory reaction has been seen near the surface of the shunt which may be the result of macrophage and platelet aggregation. In most reported cases, approximately 10–15% of the shunt surface was covered by thrombus, which may cause propagation of neo-intimal hyperplasia. Bile duct injury during TIPS creation may cause bile leak, which interferes with endothelialization leading to an acute inflammatory response and thrombus formation. There are conflicting reports concerning the site of stenosis, which may occur either in the parenchymal portion or at the hepatic vein insertion site (44, 45, 47).
Histopathological description of the external side of TIPS stents has not been adequately reported. However, it is important to remember that in areas where the stent passes through liver parenchyma, the external surface of the stent is exposed to sinusoidal blood (both venous and arterial) and probably does not undergo endothelialization.
The process by which bacteria adhere to endothelialized stents is unclear. Microbial surface components recognizing adhesion matrix molecules (MSCRAMMs) mediate bacterial adhesion to vascular surfaces (48). Many pathogenic bacteria use fibronectin matrices in the host for their adherence. Several fibronectin-binding MSCRAMMs have been identified and their binding characteristics analysed (46, 48). These included protein F from Streptococcus pyogenes (49), and FnbpA and FnbpB from S. aureus, both of which have been implicated in the pathogenic process of infective endocarditis (50–52). Although not specifically studied in endotipsitis, it stands to reason that bacteria use similar mechanisms to adhere to the endothelialized stent surfaces inside the TIPS.
Treatment and prophylaxis
The TIPS stent is irremovable after implantation without liver transplantation, and therefore treatment of endotipsitis consists of medical therapy with antimicrobial agents. Treatment is empirical with broad-spectrum antibiotics to cover Gram-positive and -negative bacteria. Most patients initially receive two different types of antibiotics including: vancomycin and aminoglycosides, or cephalosporins or β-lactam antibiotics together with vancomycin. Single-agent regimens include: vancomycin, aminoglycosides, cephalosporins and β-lactam agents, trimethoprim–sulphamethoxazole, rifampin and quinupristin/dalfopristin. Treatments are adjusted according to the causative pathogen once it was identified in cultures. In cases of suspected fungemia, patients were treated with amphotericin B and fluconazole or flucytosine. Despite their widespread use in the reported cases, it would seem prudent to avoid the use of aminoglycosides in patients with cirrhosis who were prone to renal injury, if other treatment options were available.
A literature search of all reported cases of endotipsitis, showed that 17 patients required one course of antibiotic treatment (11 with definitive and six with probable disease), two patients required two antibiotic courses (both with probable infection) and one patient with definitive endotipsitis required three courses. The antimicrobial agents used were vancomycin in 12 patients, third-generation cephalosporins in five and other agents in 14 (Table 1). Three patients with fungemia received amphotericin B and fluconazole or flucytosine.
The mean length of antibiotic therapy was 30 days (range 5–1460 days). Twenty-five patients had a resolution of their infection. They were treated for an average of 18 weeks (range 2–101, median 6); there was no correlation between the length of treatment and success rates. Similar to other endovascular infections, treatment should be administered intravenously for a prolonged period. Although there is no clear recommendation a 6-week course seems reasonable.
Blood cultures should be taken a few days after the initiation of treatment to ensure early negativity. It was also suggested that blood cultures be taken from patients after each TIPS manipulation because of the high likelihood of bacteremia after these procedures (33).
Medical treatment was ineffective in the eradication of infection in 10 patients (seven with definite endotipsitis and three with probable endotipsitis). All these patients died.
In patients with continued bacteremia, despite appropriate antibiotic treatment, removal of the TIPS via liver transplantation should be considered, (20, 25). Transplantation can be curative, but is usually contra-indicated in the presence of ‘ongoing infection’. However, if infection can be brought under partial control by antibiotics, transplantation may be successful, similar to patients with sclerosing cholangitis and Caroli's disease who are often transplanted in the presence of infection (23). In the literature, we found one case report of a patient with P. aeurginosa endotipsitis and recurrent bacteremia who underwent a successful liver transplantation when blood cultures were negative after a 6-month course of antimicrobial therapy (35).
The prognosis of endotipsitis is unknown. Overall, 8/36 (24%) patients died directly from infectious complications related to endotipsitis and three patients died of unrelated causes (9%). Only three autopsies were reported. In each case, they confirmed TIPS thrombosis and the aetiological agent previously isolated in blood cultures (25, 30, 53).
It has been suggested that all patients who undergo a TIPS procedure should receive antibiotic prophylaxis, mainly to prevent early infection. In a follow-up study of 64 patients who underwent a TIPS procedure, 10.7% developed infection, despite prophylactic antibiotics. The authors concluded that prophylactic antibiotics were not useful in preventing early bacteremia in patients undergoing TIPS insertion (30). Despite these data, it would appear prudent to use antibiotics in a technically complicated TIPS implantation or in patients with an occluded stent who undergo TIPS revision. The role of prophylactic antibiotics during the initial TIPS procedure or during procedures that could cause bacteremia such as those involving the oral cavity, genitourinary or intestinal tracts is unclear. However, in view of the low incidence of endotipsitis and the early epithelialization of the stent surface, such caution would seem unnecessary.
Endotipsitis is an emerging infectious disease that should be suspected in patients with TIPS and a persistent bloodstream infection that is not clearly attributable to another source. The disease can appear as an early infection within days or weeks after insertion of the stent or as late infection, appearing months to years after the initial procedure. Multiple causative agents have been described without a single bacteria emerging as the predominant pathogen.
Currently, professional societies need to agree on a definition for endotipsitis which will help to establish a uniform approach to diagnosis while ruling out other sources of infection. We think that the definition suggested by DeSimone and colleagues provides an appropriate framework for the diagnosis of endotipsitis. A uniform approach will facilitate a more accurate registry of patients with this rare entity. More studies are needed in order to answer questions regarding pathogenesis, prevention/prophylaxis and treatment. Prolonged courses of antibiotics are the mainstay of treatment and can be curative. Although reports vary, a minimum of 6 weeks of treatment is recommended. Patients who fail medical therapy should be considered for liver transplantation if their infection can be brought under control (Fig. 1).
Disclosure: All the authors have no financial interest in a business or commercial entity, and have no conflict of interest, that relates to the manuscript above.