Clinical and bacteriological features and prognosis of ascitic fluid infection in Chinese patients with cirrhosis

This study was conducted at Beijing 302 Hospital, which is the largest liver disease hospital in China. The hospital’s database holds records of clinical histories, disease manifestations, physical and laboratory findings, and treatments of admitted patients. Bacteriology laboratory files and patient characteristics were reviewed to identify all cases with positive ascitic fluid cultures in cirrhotic patients hospitalized in our institution from January 1, 2012 to December 31, 2015. Patients with secondary peritonitis were excluded from the study. Patients with a positive culture for common skin contaminants (coagulase-negative staphylococci, corynebacteria, propionibacteria, and Bacillus spp.) were also excluded.

The diagnosis of cirrhosis was based on clinical, biochemical, histological and/or radiological findings. A diagnostic paracentesis was performed in all patients as recommended by European Association for the Study of the Liver (EASL) [2]. SBP was diagnosed when (a) ascitic fluid polymorphonuclear leucocytes (PMN) count ≥250 cells/μL, (b) ascitic fluid culture was positive; (c) there was no evident intra-abdominal surgically treatable source for infection [2, 14]. The diagnosis of BA was made when (a) the ascitic fluid PMN count < 250 cells/μL, (b) ascitic fluid culture was positive; (c) there was no evident intra-abdominal surgically treatable source for infection [2, 15]. Fever, chills, abdominal pain and abdominal tenderness were considered symptoms of peritonitis [15]. Severity of cirrhosis was assessed at the time of the SBP or BA diagnosis using the Model for End-Stage Liver Disease (MELD) score [16]. Nosocomial infection was defined as an infection that occurred > 48 h after admission to the hospital [17].

Ascitic fluid samples were extracted and inoculated into bottles at the patient’s bedside by aseptic manipulation and cultured with BacT/Alert 240 automated blood culture system (BioMérieux France). White blood cell (WBC) and PMN were counted by Sysmex automatic cell analyzer XT-4000 (SYSMEX, Japan). Microorganism identification was performed using the VITEK-II auto microbe system (BioMérieux France). Isolated pathogens were tested for antimicrobial susceptibility using the disk-diffusion method and minimum inhibitory concentration testing. Antimicrobial susceptibility was judged according to guidelines of the Clinical and Laboratory Standards Institute (CLSI) [18].

Third-generation cephalosporins (TGC), such as cefotaxime, were used as empirical therapy antibiotic in our hospital. Empirical antibiotic therapy initiated immediately on all patients with ascitic PMN count ≥250 cells/μL, without the results of ascitic fluid culture. Bacterascites patients exhibit signs of systemic inflammation are also treated with cefotaxime. Otherwise, the bacterascites patient should undergo a second paracentesis when culture results come back positive. Patients in whom the repeat PMN count is ≥250/μL would be treated for SBP. Carbapenems were used in all severe patients with septic shock. Paracentesis was performed repeatedly 2–3 days after initiation of antibiotic therapy to determine leukocyte and PMN counts in ascitic fluid.

Quantitative variables were given as the mean ± standard deviation or the median (interquartile range). Student’s t test or Mann–Whitney U-test was used to compare continuous variables, and the x² test or Fisher’s exact test was used to compare categorical variables. Statistical differences among several groups were analyzed using one-way ANOVA or the nonparametric Kruskal-Wallis test, as appropriate. Risk factors for SBP development were determined by multivariate binary logistic regression including significant univariate predictors (p < 0.05) using stepwise backward elimination. The life-table method was used to compare survival probability for patients according to different variables. A Cox proportional hazard model analysis was carried out to identify independent predictors of 30-day mortality for every episode, defining death as the main event. Variables with p < 0.05 in the univariate analysis were candidates for multivariate analysis and non-significant factors were removed by a backward selection process. All tests were two-tailed and p < 0.05 were considered to be statistically significant. IBM SPSS Statistics 19.0 was used for statistical analyses.

During the study period (January 1, 2012 to December 31, 2015), a total of 8365 patients were subjected to diagnostic paracentesis. Ascitic fluid culture in blood culture bottles was positive in 13.0% (n = 1088) of patients. The flowchart for patient enrollment was shown in Fig. 1, and only 600 ascitic fluid infection episodes (408 SBP and 192 BA) were identified in patients with cirrhosis and enrolled in this study (Fig. 1). Laboratory and clinical features at the time of diagnostic paracentesis were given in Table 1. Serum laboratory data (alanine transaminase/ALT, aspartate aminotransferase/AST, prealbumin, BUN, creatinine, total protenin, total bilirubin, erythrocytes/WBC, PMN and prothrombin time) and ascitic features (WBC and PMN) of SBP were significantly higher than that of BA (p < 0.01). In additional, septic shock was more common in patients with SBP (p < 0.01). There were no significant differences in the percentage of nosocomial episodes between SBP and BA.

Among the 600 patients, 554 patients had monobacterial infection and the other 46 had polybacterial infection. Thus, a total of 656 pathogens were isolated from these patients and shown in Table 2. Gram-negative bacteria, accounting for 68.1% of pathogens in total, were more common than gram-positive bacteria. The major pathogens identified were Escherichia coli (n = 267, 40.7%), Streptococcus spp. (n = 110, 16.8%), Klebsiella spp. (n = 87, 13.3%), Enterococcus spp. (n = 66, 10.1%), Coagulase-positive staphylococci (n = 25, 3.8%), Enterobacter spp. (n = 20, 3.0%) and Acinetobacter spp. (n = 18, 2.7%). In total, these 7 types of pathogens account for 90.4% (n = 593) of the causative bacteria. The antimicrobial susceptibility of these bacteria was shown in Fig. 2a. Nearly half of Escherichia coli (49.4%) were resistant to TGC, on the contrary, most of Klebsiella spp. isolates were sensitive to TGC. Vancomycin was a reliable agent for treating gram-positive pathogen infections because of the low resistant rates. Nine Acinetobacter baumannii strains were multidrug resistant. Multidrug resistant pathogens were associated with high 30-day mortality rate. Our data showed that patients infected with multidrug resistant Acinetobacter spp. had significantly higher 30-day mortality compared to patients infected with susceptible Acinetobacter spp. (100% vs. 11.1%, respectively, p < 0.001, Fig. 2b). Survival curves shows that patients infected with TGC-resistant bacteria had a significantly lower survival probability than those with TGC-susceptible bacteria (p = 0.001, Fig. 2c). Also, there was a significantly lower survival probability for patients infected with carbapenem–resistant organisms than it was in carbapenem-susceptible cases (p < 0.001, Fig. 2d). Levofloxacin-resistant organisms were not associated with greater mortality than levofloxacin-susceptible ones (p = 0.092, Fig. 2e).

For 187 of 237 cases with ascitic fluid PMN counts < 250 cells/μL at the first paracentesis, a follow-up paracentesis was performed when culture results come back positive. In our hospital, if BA patient exhibited symptoms of peritonitis, the patient would be treated with empirical antibiotic immediately. The development of BA between the first and second paracentesis (usually 2–3 days) was of concern. Thus, potential risk factors for BA progressing to SBP were identified (Table 3). Results showed that development of BA was associated with the severity degree of liver diseases (MELD score, p < 0.001), serum PMN (p = 0.002), ascitic PMN (p = 0.002), presentation of fever (p = 0.036) and serious complications (hepatocellular carcinoma and septic shock, p = 0.035 and 0.014, respectively). In multivariate analysis, only the presentation of hepatocellular carcinoma (OR = 3.047, 95% CI:1.161–7.997, p = 0.024) and higher ascitic PMN count (OR = 1.007, 95% CI: 1.001–1.013, p = 0.023) remained independent predictors for BA progressing to SBP.

The 30-day mortality of patients with SBP was significantly higher than that of BA (25.7% vs. 13.5%, p = 0.001) (Table 1). We studied independent risk factors of 30-day mortality in BA and SBP respectively, and found that independent risk factors were different between them. Multivariate analysis, including the variables associated with mortality by univariate analysis, showed that the concentration of BUN was independent risk factor for 30-day mortality of BA patients (Table 4). For patients with SBP, the independent risk factors for 30-day mortality were age, MELD score, septic shock and hepatocellular carcinoma (Table 5).

To study the impact of isolate type on clinical characteristics and mortality, we limited the analysis to patients with monobacterial ascitic fluid infections. Clinical characteristics of patients infected with different types of bacteria were shown in (Additional file 1): Table S1. There were significant differences in clinical characteristics among the seven types of bacterial infection, for example the percentage of peritonitis symptoms, several clinical complications, type of peritonitis (SBP), nosocomial infection, serum and ascitic features, MELD score and days between admission and onset of infection (p < 0.01). To investigate the impact of isolate type on the 30-day mortality, Cox proportional hazard model was used to control confounding variables, such as MELD score, defining isolate type as dummy variable (Table 6 and Additional file 1: Table S2). In univariate analysis, patients infected with Acinetobacter spp., Klebsiella spp. and Enterococcus spp. had higher hazard ratios of 30-day mortality compared to those infected with Escherichia coli. Multivariate analysis showed that only patients infected with Klebsiella spp. had higher hazard ratio of 30-day mortality compared to those with Escherichia coli (Table 6).