Clinical characteristics and treatment outcomes in a cohort of patients with pyogenic and amoebic liver abscess

Background We describe the clinical features of a cohort of patients with liver abscesses and investigate relationships between clinical, radiological and microbiological findings and mortality. Methods Retrospective review of pyogenic (PLA) or amoebic liver abscesses (ALA) diagnosed and treated at a major infectious diseases department in London over 9 years. Results One hundred forty-one patient records were identified; 132 (93.6%) had PLA and 9 (6.4%) ALA. No organism was identified in 38.6% (51/132); a single bacterial species was isolated in 47.0% (62/132) of PLA, ≥ 2 in 14.4% (19/132). There was weak evidence of variation in abscess size by type of microorganism, with streptococcal PLA typically larger (p = 0.03 for Streptococcus milleri group, p = 0.05 for non-milleri streptococci). Patients with ALA were younger (median 41, IQR 37–51 years) than those with PLA (median 68, IQR 50.5–78 years) (p = 0.003) and all were male (9/9, 100%, (p = 0.03)), with a history of recent travel in the majority (6/9, 66.7% (p = 0.003)). C-reactive protein was higher in ALA than in PLA (p = 0.06). In the entire cohort, loculation (HR = 2.51 (95% CI 1.00–6.32), p = 0.04) and baseline ALP (HR = 4.78 (95% CI 1.19–19.2) per log10 increase, p = 0.03) were associated with mortality. 16S ribosomal RNA (rRNA) analysis was used in a subset of culture-negative cases and increased the diagnostic yield by 13%. Conclusions Clinical or radiological features cannot be used to distinguish between PLA and ALA, or help identify the bacterial cause of PLA. However, ALA is more common in young, male patients with a history of travel. 16S rRNA analysis of abscess fluid has a role in improving microbiological diagnosis in culture-negative cases. Electronic supplementary material The online version of this article (10.1186/s12879-019-4127-8) contains supplementary material, which is available to authorized users.


Background
The incidence of liver abscess, a rare but potentially lifethreatening infection, appears to be increasing worldwide [1][2][3][4]. The use of antibiotics, imaging and less invasive procedures for source control have improved treatment outcomes over the last century, but mortality remains high [1,5,6] At present, there are no national or international clinical guidelines, and limited evidence to guide decisions about care.
A wide variety of bacteria have been described as causes of pyogenic liver abscess (PLA), but published data on associations of bacterial species with clinical presentation, radiological findings and prognosis are sparse. Entamoeba histolytica, a protozoan parasite, is also a well-recognized non-bacterial cause of liver abscess. In Europe, cases of amoebic liver abscess (ALA) are rare and usually imported, whereas in some highly endemic areas ALA can be more common than PLA [7]. There is good evidence that ALA can be treated successfully with shorter courses of antimicrobial therapy than PLA, and that drainage is generally not required [8]. However, there is limited evidence as to how PLA and ALA can be distinguished clinically or radiologically.
In this study we describe a cohort of liver abscess patients treated at a large infectious diseases department over a 9-year period and report the clinical characteristics and treatment outcomes of PLA and ALA.

Setting and patients
We conducted a retrospective case note review of all adult liver abscess patients treated at our hospital (Northwick Park Hospital, Middlesex, UK) between May 2008 and June 2017. Patients presenting with liver abscess were identified using a prospectively recorded database of infectious disease inpatients, as well as a retrospective search of clinical coding data (search term "liver abscess", ICD 10 code = K750). Data collected from the hospital electronic patient information systems included patient characteristics (age, sex, ethnicity), medical background and comorbidities, the results of standard biochemical and haematological tests on admission (full blood count, C-Reactive Protein (CRP), alanine transaminase (ALT), alkaline phosphatase (ALP) and bilirubin), the characteristics of the abscess (isolated microorganism(s) and resistance patterns, number of abscesses, maximum diameter and loculation), and treatment outcome (antibiotic use, use of aspiration, discharge, death).

Management of liver abscess
In the absence of consensus guidelines, patients with liver abscesses were investigated and managed at the discretion of the attending physician. Abscesses were identified by ultrasound or computed tomography. Samples submitted to the microbiology lab (abscess pus, blood cultures) were processed according to national guidelines (British Society of Antimicrobial Chemotherapy (BSAC), until February 2016, then European Committee on Antimicrobial Susceptibility Testing (EUCAST)) in a CPA (Clinical Pathology Accreditation) accredited laboratory. The choice of antimicrobial agents was guided by infection specialists. In a subset of cases, aspirates were sent for 16S ribosomal RNA (rRNA) sequencing (Great Ormond Street Hospital, in-house assay). Amoebic serology was performed when there was clinical suspicion of E. histolytica infection (in-house immunofluorescent antibody test (IFAT) and Cellulose Acetate Precipitin (CAP), Department of Parasitology, Hospital for Tropical Diseases, London).

Statistical analysis
Characteristics of pyogenic and amoebic abscesses were compared using Fisher's exact and Kruskal-Wallis tests for categorical and continuous variables, respectively (α = 0.05). Differences in clinical characteristics and treatment outcome (death within 30 days, and death within 6 months) by type of isolated microorganism were tested using a Wald-type test (F statistic) of Somers' D parameters for continuous variables (adjusted for withinpatient clustering) or Fisher's exact tests for categorical variables. Associations of type of microorganism with all-cause mortality (within 30 days and within 6 months of admission) were investigated by fitting random-effects parametric survival-time models with patient as the panel variable and standard errors adjusted for within-patient clustering. Risk factors for 6month all-cause mortality were investigated using Cox regression models. Variables for which there was evidence (p ≤ 0.1) suggestive of an association in univariate models were carried forward to a multivariable model, from which variables were eliminated by mutual adjustment to obtain a final model of independently associated (p ≤ 0.05) risk factors.

Clinical characteristics by type of microorganism
There was weak evidence for overall variation in size (p = 0.04) and loculation (p = 0.09) of abscesses by type of microorganism (Table 2). Abscesses caused by streptococci (S. milleri group and non-milleri streptococci) tended to be larger than those caused by other microorganisms (Somers' D p = 0.03 for S. milleri, p = 0.05 for non-milleri streptococci). Abscesses with no isolated microorganism tended to be smaller (p = 0.04). There were no overall associations between specific microorganisms and baseline biochemical measurements ( Table 2), but compared with all other microorganism: Enterobacteriaceae (including E. coli) and S. milleri were associated with lower ALT (Somers' D p = 0.03 and p = 0.05, respectively), and Klebsiella with higher ALT (p < 0.001). ALA tended to be associated with higher CRP (p = 0.06).  Table 2).

Associations of type of microorganism and other factors with death
There were no deaths in the ALA group, compared to 5.3% (7/132) 30-day mortality and 15.2% (20/132) 6month mortality in the PLA group. Two of the seven 30-day deaths and 4/20 6-month deaths were patients with biliary malignancy. There were no associations between type of microorganism and all-cause mortality within 30 days or 6 months (Additional file 1:  Table S4) were independently associated with mortality when mutually adjusted.

Discussion
In our cohort of patients with liver abscess we found no large differences in clinical parameters between pyogenic liver abscess (PLA) and amoebic liver abscess (ALA), other than weak evidence of higher baseline C-Reactive Protein (CRP) for ALA compared with PLA. PLA patients had a 6-month mortality of 15.2% (20/132), whereas no ALA patients died. We found that streptococci were associated with larger abscesses than other bacteria. Baseline ALP and loculation, but not type of microorganism, were associated with increased risk of 6month all-cause mortality in the PLA group. Previous studies comparing pyogenic and amoebic abscesses have attempted to identify features that distinguish the two, other than travel history [7,[9][10][11][12]. Many reports have shown a preponderance of ALA in male patients, who also tend to be younger than patients with PLA [4,6,[9][10][11][12][13][14][15][16][17][18]. Others have documented differences in shape, position and echogenicity on ultrasonography that suggest ALA over PLA [10,12,19,20]. Our results are consistent with the epidemiology described in these earlier papers, as all of our ALA patients were male and had a median age 30 years younger than the PLA patient group. As in other studies, all 9 ALAs in our cohort involved the right lobe (2 also involved the left). We also found that CRP was higher in ALA than in PLA, although this result is unlikely to assist in differentiating these entities.
Half (47%) of PLAs in our study were monomicrobial, and the most common bacterial isolates (found in   roughly equal proportions) were E. coli, S. milleri group and Klebsiella species, consistent with results from a number of other studies [1,16,17,21,22]. Culture of fluid obtained by cyst aspiration provided a microbiological diagnosis in 34% of cases, whereas blood culture was only positive in 25%. Given the importance of Enterobacteriaceae as etiologic agents in PLA, and the associated problems of antibiotic resistance, aspiration for diagnostic purposes would appear to be a crucial tool in optimising management. There is also likely to be a role for routine testing of culture-negative aspirates with newer molecular techniques such as PCR amplification and sequencing of the 16S ribosomal RNA (rRNA) gene. This has been shown to be more sensitive than traditional culture methods [23,24], particularly when antibiotics have been administered [25]. In our cohort, 16S rRNA PCR was the only method resulting in a microbiological diagnosis in 9/14 (64.2%) cases in which it was used. Overall, this technique increased pathogen detection by 13% and only 3/17 (17.6%) samples were negative. Although 16S rRNA analysis does not provide antibiotic susceptibility data, the risk of resistance can be inferred from the identity of the pathogen. Associations between radiological features and type of organism were weak, although we did find that abscesses caused by streptococci tended to be larger than abscesses caused by other bacterial species. Biliary abnormalities are estimated to be responsible for 30-50% of cases of PLA and are more common than those related to portal pyaemia, or haematogenous or direct spread from an adjacent viscus [5,12,[26][27][28]. In our cohort, a biliary source was demonstrated in only 18/132 (13.6%). The reason biliary sources were less common in our cohort than in other series is not clear. Diabetes appears to be a major risk factor for developing PLA, and is associated with severe disease [29][30][31]. About 30% of our patients with both ALA and PLA were diabetic and it is possible that cases in this group are over-represented in our cohort. The source of infection remains unclear in significant number of PLAs [1,15,16] and we were unable to demonstrate a source of infection in almost 40%. A source of particular concern is GI malignancy, as these can easily be missed on initial imaging and reports have documented occult colonic malignancies in approximately 24% of cryptogenic PLAs, suggesting that luminal imaging or colonoscopy should be seriously considered in individuals with PLA without an obvious source [32].
Along with antibiotic therapy, options for treatment of PLA include percutaneous aspiration, percutaneous drainage and open surgical drainage. There have been no trials that have compared these different treatment strategies. In our group, 60% of cases were managed with aspiration or drainage in addition to antibiotic therapy (76.3% of large (> 5 cm) abscesses vs. 25% of small (< 5 cm) abscesses). Seven (5.3%) of our PLAs required surgical drainage. It is unclear how percutaneous or surgical drainage affected outcomes. Antibiotic treatment was altered at least once in 79.5% of PLA patients (more than twice in 34.1%), some due to antibiotic resistance, with 34.1% of patients growing a resistant organisms, including 12.1% resistance to one of the most commonly used empiric antibiotic, co-amoxiclav, but others will have been altered empirically due to deterioration of the patient. This observation underlines the importance of early pathogen identification in order to optimise management.

Conclusions
There is little quality evidence to guide optimal management of PLA. Using a well-characterised cohort of liver abscess patients, we found some correlations between organism, clinical features and outcomes of treatment. The heterogeneity of this cohort of liver abscesses, as with others in the literature, means that associations may have been missed and that those identified, such as higher CRP in ALAs and larger size in PLAs caused by streptococci, are too subtle to be of use to clinicians. These findings emphasise the need for a microbiological diagnosis, which is most effectively obtained by culture of aspirate fluid. We found 16 s rRNA gene PCR useful in the identification of bacteria in culture negative samples. Newer molecular technologies such highthroughput sequencing may, in the future, further improve the sensitivity of nucleic acid amplification tests and may also provide antibiotic susceptibility data and this is likely to lead to improved outcomes for individuals with these infections.

Additional file
Additional file 1: 4 tables showing additional supporting statistical analysis of data described in main manuscript. (DOCX 30 kb)