Introduction

Klebsiella liver abscess (KLA) is common in Asia, representing a distinct syndrome caused by strains of Klebsiella pneumoniae bearing the hypermucoviscous phenotype mediated by the K1 or K2 capsule type [1]. KLA differs from other forms of pyogenic liver abscess in that it affects individuals without biliary disease and is strongly associated with diabetes mellitus [2]. The syndrome is notable for an increased incidence of metastatic infection to lung, eye or brain [1]. There are distinct radiologic features [3–7]. Whereas other pyogenic liver abscesses are often multiple and unilocular, KLA are more frequently single and multilocular. KLA may be associated with thrombophlebitis of hepatic or portal veins, although reported incidence varies. We have previously reported regional thrombophlebitis in up to one third of cases of KLA, as compared to only 5% in other causes of liver abscess [7]. Another report identified thrombophlebitis rates of 42% in pyogenic liver abscess but did not differentiate by causative organism [6]. Others reviewing radiologic appearances of KLA have not reported thrombophlebitis rates [3–5]. Septic thrombophlebitis has been identified as a risk factor for haematogenous spread of infection [7]. However optimal management is undefined. Anticoagulation of portal vein thrombosis secondary to liver abscess is controversial. A recent study showed spontaneous resolution of pylephlebitis in two thirds of patients with KLA treated with antibiotics alone [8]. To date no study has addressed management of hepatic vein thrombosis in KLA.

Materials and Methods

Results

169 patients were identified. The median age was 58 years. 71% were male, 55% were diabetic, and 22.5% required ICU admission. 66.3% were bacteraemic with K. pneumonia. 66.9% of liver abscesses were in the right lobe, 26.6% in the left lobe and 2.4% in the caudate lobe. The remainder were in more than one lobe. The mean abscess maximal diameter was 6.7cm. 63.3% were percutaneously radiologically drained, 13.6% surgically drained, and 23.1% required no drainage. Appropriate antibiotics were given in all cases.

Thrombophlebitis was identified on CT in 53 patients (31.4%), of which 3 were affecting portal vein, 49 hepatic vein and one inferior vena cava (IVC) (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Location of thrombosis.

https://doi.org/10.1371/journal.pone.0140129.t001

Three subjects had clots extending to the IVC. Thrombosis was unrelated to age, gender, diabetes, immobility (wheelchair or bedbound), presence of bacteraemia, abscess size, or whether the abscess was multiple or loculated (Table 2). Nine patients had concomitant pneumonia which may represent septic metastatic infection.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Risk factors for thrombosis.

https://doi.org/10.1371/journal.pone.0140129.t002

41 patients with thrombosis had a follow up CT scan for comparison.The mean interval time between scans was 44 days. 30 (73.2%) revealed improvement or complete recanalization, 9 (22.0%) showed interval stability and 2 (4.9%) showed extension of the thromboses (Table 3). The two who showed extension had no complications and were cured clinically. Recanalization of the affected vein was strongly associated with complete abscess resolution. Of the 30 patients with recanalization or reduction in size of thrombus, 11 (36.7%) showed complete abscess resolution, compared to none in the group that did not show improvement of the thrombus (p = 0.02). Diabetes, immobility, location of thrombus (hepatic versus portal) and drainage of the abscess were not associated with thrombus resolution. Scans were assessed for radiologic evidence of thrombotic complications such as ascites, splenomegaly, collaterals and cavernous transformation. Only 3 cases of ascites were identified, all in patients with hepatic vein thrombosis. At follow up scan the ascites had resolved in two cases and improved in the other, without anticoagulation. No other radiologic complications were identified.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Radiologic outcomes for those individuals with a follow up scan available.

https://doi.org/10.1371/journal.pone.0140129.t003

In terms of clinical outcomes, there was no significant difference between rates of intensive care unit (ICU) admission, mean length of hospitalization or mortality between those with and without thrombosis. Six patients in the study died (3.6%). In all cases the cause of death was sepsis. Of these, three had hepatic vein thrombosis. Two had a follow up scan of which one showed no recanalization and one showed improvement.

Of the 53 patients with septic thrombophlebitis, only one received therapeutic anticoagulation. This patient had a 13.3cm right lobe liver abscess with middle and left hepatic vein thrombosis extending to the IVC. The abscess was radiologically drained and K. pneumoniae was isolated from blood and abscess fluid. The patient developed metastatic infection to the lungs with pneumonia requiring intubation, but no pulmonary embolus was identified. Following 28 days of therapeutic anticoagulation with low molecular weight heparin (LMWH), the patient was well and an ultrasound showed complete abscess resolution. LMWH therapy was complicated by hematuria. It is worth noting that of the three patients with pylephlebitis none were anticoagulated. Two had follow up scans available, of which one showed complete recanalization and one showed improvement. All three were clinically cured without complications. Incidentally there were two patients without septic thrombophlebitis who received therapeutic anticoagulation for other indications; one for lower limb deep vein thrombosis (DVT) and one for pulmonary embolism (PE). The subject with DVT developed a gastrointestinal bleed as a complication of anticoagulation. 6 patients in the total cohort received prophylactic anticoagulation with LMWH, 3 in the group with thrombosis and 3 in the group without thrombosis.

Discussion

This study identified thrombophlebitis in nearly one third of patients with KLA. The natural history of these thrombosis has not been described previously and the role of anticoagulation has not been defined in this population.We found little clinical consequence of septic thromboses of portal vein or hepatic veins. The one patient with PE had no evidence of thrombophlebitis in the liver.

It would appear that in KLA the vast majority of hepatic vein septic thromboses resolve spontaneously without anticoagulation. The thromboses improved as the abscess resolved, suggesting management should be targeted at the underlying abscess. Although too few to draw any real conclusions it is worth noting that the 3 portal vein thromboses in this series resolved spontaneously, which is in keeping with the recent observation of high rates of spontaneous recanalization of portal vein thromboses in KLA [8].

Whilst septic thrombophlebitis at presentation has been associated with metastatic infection, most of this metastasis occurred prior to presentation [7]. Targeting the thrombus once antibiotics have been initiated does not appear to be necessary, and furthermore we have demonstrated that therapeutic anticoagulation is not without risk. While we cannot rule out a role for anticoagulation in a subset of patients, the priority should be on early diagnosis with prompt drainage of the abscess and initiation of appropriate antibiotics.

Supporting Information