Bacterascites: A study of clinical features, microbiological findings, and clinical significance

2.1 Study design and data collection

All consecutive ascites cultures performed in patients with advanced chronic liver disease between January 2003 and August 2016 at Erasmus MC, University Medical Center, were retrospectively reviewed to identify patients fulfilling the diagnostic criteria for bacterascites. The Medical Ethics Committee Erasmus MC, approved the study protocol on February 27th, 2017 and stated that written informed patient consent was not necessary considering the design of the study.

All ascites samples obtained during the study period were reviewed, implying that, if applicable, multiple bacterascites episodes per patient were taken into account. The lowercase letter n was used to indicate the number of patients and the capital letter N for the number of bacterascites episodes.

Paracentesis was performed in patients with new-onset ascites, clinical deterioration, and large-volume removal in refractory ascites.9 White blood cell (WBC) and PMN count in ascites were automatically determined and aerobic and anaerobic blood culture bottles (Bactec^®) were used for bacterial cultures. Blood cultures taken within two hours before or after ascites cultures were considered to be concomitant.

Demographic, clinical, biochemical, and survival data from patient hospital records were collected.

To determine the prognostic impact of bacterascites in comparison with reported outcomes of SBP, a control cohort was established with patients from our centre with SBP, performed as described in a previous publication.10 In order to create homogenous groups for survival analyses, patients with both episodes of bacterascites and SBP were categorized as SBP when the first ascites infection was SBP or bacterascites developed within 48 h to SBP. Patients with bacterascites developing SBP after 48 h, but within 30 days were excluded from survival analysis. In addition, the MELD score-dependent relation of the prognosis of bacterascites patients was studied and compared with SBP patients.11, 12

Furthermore, a PubMed search was performed on December 1st, 2017 with the following search terms: spontaneous bacterial peritonitis (ALL) AND (outcome (ALL) OR mortality (ALL)) AND prognos* (ALL). The studies were reviewed and included when the following criteria were met: (1) observational studies, (b) study population consisted of patients with SBP defined as a PMN count of 250/μL or greater in ascites, (c) minimum study population of 50 adult patients, (d) reporting survival analysis and 1-month or in-hospital mortality rate, and (e) written in English. Interventional studies (eg, randomized controlled trials), studies with a selected population (eg, HIV patients), and abstracts were excluded. Study and clinical characteristics were collected from the included studies.

2.2 Definitions

Bacterascites was defined as an ascitic fluid sample with a PMN count below 250/μL and a positive bacterial culture, in the absence of evidence for an intra-abdominal source of infection.1 Infection acquisition was categorized as nosocomial (infection was detected after 48 h after hospital admission), healthcare-associated (<48 h after hospital admission in patients with any 90-day prior healthcare contact), or community-acquired (within 48 h after hospital admission in patients without any 90-day prior healthcare contact).13 Ascites was graded as diuretic-responsive or diuretic-refractory, and hepatic encephalopathy (HE) as none, or West Haven grade 1-4. Patients were categorized as symptomatic, when one or more of the following symptoms, shown to be frequently present in patients with SBP, were recorded: abdominal discomfort (feeling of fullness), abdominal pain/tenderness, fever, and a change in mental status (recorded as HE grade).14 Recent gastrointestinal (GI) bleeding was defined as a diagnosed upper GI bleeding in the 72 h prior to index paracentesis. During the study period, the standard primary antibiotic treatment in patients with variceal bleeding was oral norfloxacin 400 mg twice daily or intravenous ceftriaxone 1 g daily during five days. The secondary prophylaxis for SBP was norfloxacin 400 mg once daily.8, 9

2.3 Statistical analysis

Continuous variables were reported as mean with standard deviation (SD), after visual confirmation of approximate normality, and compared using the Student’s t-test. Categorical variables were reported as count with proportion and compared using the Chi-square test. A two-sided P-value <0.05 was considered significant.

Transplantation-free survival was analysed using Kaplan-Meier survival analysis. Follow-up started at the time of the first ascitic fluid analysis. A multivariable logistic regression analysis was carried out to identify predictors for treatment of bacterascites, a multivariable logistic regression analysis in the untreated patient group to identify risk factors for worse outcome (ie, liver-related death before culture results were known, SBP development, and persisting bacterascites), and a multivariable Cox's proportional hazard analysis to identify independent predictors for 3-month mortality. These analyses were performed using the candidate predictor variables: age, gender, aetiology of liver disease, MELD score, hepatocellular carcinoma, gastrointestinal bleeding, HE, grade of ascites, symptoms of infections, immunosuppressive medication use, antibiotic prophylaxis use, Staphylococci cultured, and PMN count in ascites, with the addition of initiation of antibiotic treatment for bacterascites in the Cox's regression. The regression models were employed using the backward stepwise selection method with removal testing based on the probability of the likelihood-ratio statistic. Statistical analyses were performed using SPSS Statistics for Windows, Version 21.0. (IBM Corp., Armonk, NY, USA).

3.1 Patients

Between January 2003 and August 2016, 142 episodes of bacterascites were diagnosed in 123 patients. The demographic, clinical, and laboratory data are summarized in Table 1. Patients with bacterascites were mainly male with a mean age of 63 years (±14) and a median MELD score of 20 (IQR 14-25).

3.2 Bacterascites

The infection was in 11% of the bacterascites episodes community acquired, in 55% healthcare-associated, and in 34% nosocomial acquired.

One or more clinical symptoms of infection were present in the majority (78%) of patients with bacterascites. Sole abdominal discomfort was reported by 18%, HE by 16%, abdominal pain by 9%, and fevers or chills by 3%. A combination of these symptoms were present in 32%: 13% had HE and abdominal pain, 11% HE, fever and abdominal pain, 5% fever and abdominal pain, and 3% HE and fever.

Symptomatic patients had a similar age (mean 64 vs 63 years; P = 0.907), MELD score (median 20 vs 19 points; P = 0.313), serum creatinine (median 106 vs 104 mmol/L; P = 0.606) and PMN count in ascites (median 23 vs 19 cells/μL; P = 0.576) compared to asymptomatic patients. Table 2 shows additional characteristics in relation to the presence of symptoms. Monomicrobial bacterascites was just as likely to be symptomatic as polymicrobial bacterascites (P = 0.660). Seventeen percent of patients with bacterascites were diagnosed with concomitant infections of the respiratory tract (6%), urinary tract (5%), or skin (6%). Concomitant blood cultures were obtained in 42% and 17% of symptomatic and asymptomatic cases respectively. The proportion of positive blood cultures did not statistically differ between the groups.

Fourteen patients were diagnosed with more than one episode of bacterascites. In five patients, the second episode was diagnosed within 5 days, and in nine patients after a median time of 31 days. Of these 14 patients, nine patients died, four patients received a liver transplant, and one patient was lost to-follow-up. The median time till one of the endpoints was reached was 74 days. When patients with a single episode were compared with patients with multiple episodes, there were no statistical differences in age, gender, MELD score, PMN count in ascites, presence of symptoms, antibiotic prophylaxis use, or presence of a GI bleed.

In 36 of 142 bacterascites episodes (25.4%), an admission to the intensive care unit (ICU) was necessary; 14 cases (9.9%) of bacterascites were diagnosed on the ICU and in 22 cases (15.5%) patients were admitted to the ICU after organ failure development following bacterascites diagnosis. ICU admission was less than 7 days in 21 cases, between 7 and 14 days in six cases, between 14 and 30 days in four cases, longer than 30 days in five cases.

3.3 Microbiology

Monomicrobial bacterascites was found in 81% of all episodes, consequently multiple pathogens were cultured in 19% of all episodes. In polymicrobial bacterascites, two or three different organisms were cultured. Gram-positive bacteria were predominantly cultured in monomicrobial bacterascites. The 177 species cultured in monomicrobial and polymicrobial bacterascites are listed in Figure 1. Staphylococci were most often isolated (43%), followed by Streptococci (17%), Enterococci (14%), and Escherichia (11%). The cultured species of these four most common found genera are subtyped in Figure S1.

Multidrug antibiotic-resistant (MDR) organisms were isolated in 25% of all episodes. Methicillin-resistant Staphylococci were the most frequently found MDR organism (N = 26), followed by extended-spectrum beta-lactamase (ESBL) producing bacteria (N = 8). A similar proportion of patients using primary antibiotic prophylaxis were infected with a MDR organism (29.0%) compared to patients without prophylaxis (23.1%) (P = 0.506).

3.4 Antibiotic therapy

In 96 (68%) of the total 142 episodes of bacterascites antibiotic treatment was initiated: in 49 episodes after paracentesis and before culture results became available, in 47 episodes after culture results were known. In 16 cases, the antibiotic treatment was modified based on culture results. Amoxicillin-clavulanic acid combination was most often prescribed (30%), followed by cephalosporin (14%) and vancomycin (10%). Symptomatic patients more often received treatment compared to asymptomatic patients (72% vs 52%; P = 0.031). Patients with higher MELD score (HR 1.156 per point, 95%CI 1.060-1.260, P = 0.001), higher PMN count in ascites (HR 1.017 per point, 95%CI 1.005-1.030, P = 0.007), an infection with another bacterial genus than Staphylococci (HR 3.512, 95%CI 1.333-9.253, P = 0.011), and a female gender (HR 2.837, 95%CI 1.066-7.547, P = 0.037) were more likely to receive antibiotic treatment for bacterascites.

In 46 episodes of bacterascites, antibiotic treatment was not initiated. In 31/46 episodes (67%), patients had signs or symptoms of infection. A total of 111 episodes of bacterascites were symptomatic. The ascitic PMN count in 80 patients who were treated with antibiotics was significantly higher (median 31, range 0-235) than the PMN count (median 13, range 0-71) in those patients (n = 31) who did not receive antibiotic treatment (P = 0.002). Figure 2 shows a flow chart of the clinical course of the untreated episodes of bacterascites. Of these 46 bacterascites episodes, the patient died before culture results were known in eight cases, in all these cases of decompensating liver disease. These eight patients had a median MELD score of 29 (IQR 18-30), median serum creatinine of 147 μmol/L (IQR 100-250), and 5/8 patients were admitted at the ICU; either before paracentesis (two patients), or after paracentesis (three patients). SBP developed in six cases, and bacterascites persisted in five cases. The latter group and those diagnosed with SBP were immediately treated with antibiotics. A logistic regression analysis was performed in the untreated patients to identify risk factors for liver-related death before culture results were known, SBP development, and persisting bacterascites (19/46) compared to clinical remission (27/46). We found that MELD score (HR 1.286 per point, 95%CI 1.071-1.546, P = 0.007) and age (HR 1.113 per year, 95%CI 1.027-1.205, P = 0.009) were independent risk factors.

3.5 Clinical course and outcome

The survival analyses included 114 patients with bacterascites and 88 patients with SBP, after the exclusion of patients with both SBP and bacterascites. The median follow-up time in 114 patients was 38 days (IQR 15-272). In this study cohort, 27 patients were alive or lost to follow-up, 16 patients received a liver transplant, and 71 patients died. The causes of death were in 36 patients (50.7%) liver disease-related, in 29 patients (40.9%) unknown, and in six patients (8.4%) a combination of liver disease-related and nonliver disease-related.

The cumulative mortality rates in bacterascites patients (1-month: 36%; 3-month: 56%; 6-month: 62%; 1-year: 66%) are statistically comparable to that reported for SBP patients (1-month: 34%; 3-month: 54%; 6-month: 67%; 1-year: 77%) (log-rank test P = 0.397) (Figure 3).

The most important predictive factors for 3-month mortality after bacterascites diagnosis were: MELD score and the presence of hepatic encephalopathy (Table 3). Figure 4 shows the MELD score-dependent relation of survival in 114 patients with bacterascites and 88 patients with SBP.

3.6 Bacterascites in comparison with SBP in the literature

Our literature search for relevant studies of SBP identified 17 publications (Table 4).10, 15-30 The reported baseline clinical characteristics including age, gender distribution, and liver disease severity scores of SBP were comparable to those identified in our patient population with bacterascites. In addition, the cumulative mortality rate in our bacterascites cohort (1-month: 36%; 6-month: 62%; 1-year: 66%) also appears comparable to that reported for SBP (1-month: 13%-49%; 6-month: 52%-59%: 1-year: 49%-70%).