Study design

This was a retrospective, observational, multicenter study that analyzed hospital discharge of patients with liver cirrhosis and a primary or secondary diagnosis of SBP from January 2008 to December 2013. Patients were followed up until two years after diagnosis. The study sample consisted of all consecutive cases from three Catalonian hospitals, Hospital Universitari Parc Taulí in Sabadell, Hospital Universitari Joan XXIII in Tarragona and Hospital Universitari Sant Joan in Reus. The total reference population was approximately 1,200,000 people. The study was approved by the ethics committee of the Hospital Universitari Parc TaulÍ (CEIC 2013/587).

Patient selection

The study population was patients consecutively admitted with a diagnosis of liver cirrhosis based on clinical, laboratory or ecographic criteria or a liver biopsy. The study inclusion criteria included cases with a biochemical study of ascitic fluid (AF) consistent with SBP (presence of more than 250 polymorphonuclears [PMNs] in AF and positive microbiological culture) or with culture negative neutrocytic ascitis (more than 250 PMNs and negative culture). In clinical practice it is not usual to differentiate between SBP and culture negative neutrocytic ascites; the manuscript will generically refer to SBP and the two conditions will only be distinguished when necessary.

The study exclusion criteria included: a) patients with suspected secondary bacterial peritonitis based on one of the following: a positive polymicrobial culture, glucose levels < 50 mg/dl or suggestive radiographic signs; b) patients with bacterial ascites (pathological AF culture but less than 250 PMNs/ml); and c) patients with a prior episode of SBP.

Study variables

The following variables were collected: demographic data (age, sex), clinical variables (unit of hospitalization, etiology of cirrhosis, CPT, associated comorbidities, CCI, drug treatment at diagnosis, blood pressure, temperature and heart rate, associated complications such as hepatic encephalopathy, upper gastrointestinal bleeding and acute kidney injury), laboratory variables (white blood cells [WBCs] in plasma, creatinine, blood urea nitrogen [BUN], sodium, albumin, bilirubin and prothrombin time); microbiological data (AF and blood cultures resistance profile of isolated germs and community-acquired/nosocomial infections) and therapeutic data (antibiotic administered, treatment duration, albumin dose administered and changes in treatments).

In-patients who suffered an infection between two weeks to 48 hours before the SBP episode were considered as nosocomial. Acute kidney injury (AKI) was defined using the AKIN (Acute Kidney Injury Network) criteria ¹¹. CPT, MELD and CCI were calculated at the time of SBP diagnosis. CPT indicates the degree of hepatic failure on a scale from 5 to 15 based on laboratory and clinical variables ¹².

The MELD index is a mathematical model for the prediction of survival of patients with liver disease ¹³. Finally, the CCI is a prognostic index of mortality that considers patient age and the presence and severity of 19 comorbidity conditions including diseases such as cancer, diabetes, cirrhosis and ischemic heart disease ¹⁴. The presence of bacterial resistance or a decrease lower than 25% in PMN count at 48 hours were considered as treatment failure in analysis of the impact of treatment on mortality.

According to the main clinical guidelines ¹⁵, patients with hyperbilirubinemia (conjugated bilirubin > 4 mg/dl) or impaired kidney function (serum creatinine > 1 mg/dl) were considered as candidates for treatment with serum albumin.

Statistical analysis

Continuous variables were described as mean ± standard deviation (SD) and were compared using the Student's t test. Continuous variables with an associated non-linear risk were categorized by risk groups. Categorical variables were presented as percentages and were analyzed using a Chi-squared test.

Both short-term (at 30 and 90 days) and long-term (at one and two years) inhospital mortality was analyzed. Overall survival was analyzed using KaplanMeier curves, the dependent variable was death or liver transplantation, and the follow-up time considered, as up to occurrence of the event or two years after diagnosis of SBP. The relapse rate of SBP in the same follow-up time was recorded.

For each time interval (30 and 90 days/one and two years), logistic regression univariate analysis was performed. Variables with values of p < 0.05 were considered as statistically significant. Variables with values of p < 0.2 in the univariate study were subsequently compared using a log-rank test in a multivariate study. Independent short-term (30 and 90 days) and long-term (one and two years) prognostic factors were sought and values of p < 0.05 were considered as statistically significant. Logistic regression and Kaplan-Meier analyses were subsequently performed.

The short- and long-term prognostic power of CPT, MELD and CCI was analyzed. For CPT, the classification of Child A, B and C was maintained. MELD and CCI were categorized in homogeneous groups with a comparable number of patients. The following groups were created for MELD: MELD < 15, MELD 15-21 and MELD ≥ 22. CCI groups were categorized as follows: CCI 1-4, CCI 5-7 and CCI ≥ 8. A ROC curve was constructed with the cut-off points for sensitivity and specificity with the best discrimination of survival for all three indices. Data were analyzed using the SPSS version 21 statistical software (SPSS Inc, Chicago Illinois, USA).

Baseline characteristics

A total of 159 patients who met all the inclusion criteria and none of the exclusion criteria were enrolled into the study. Table 1 shows patient characteristics. Most patients were male (71.7%) with a mean age of 63.5 years (SD ± 13.3).

Upon admission, 21% of patients were receiving primary prophylaxis for SBP with norfloxacin (7.5%) or were taking rifaximin due to prior episodes of hepatic encephalopathy (13.5%). Paracentesis was performed within 24 h of attendance to the Emergency Room or day hospital in all patients with community-acquired SBP. The study design did not allow for the assessment of a delay in diagnosis of inpatients.

Fifty-six per cent of patients had some sort of complication at diagnosis. The most common complication was hepatic encephalopathy, followed by AKI. According to the AKIN classification, 54.5% had AKIN 1, 30.9% had AKIN 2 and 14.5%, AKIN 3.

SA was administered for volume expansion in 86.8% (138) of patients. Of these, 102 had creatinine levels of 1 mg/dl or higher and/or bilirubin levels greater than 4 mg/dl at diagnosis. Twenty-one patients did not receive SA and volume expansion was indicated in 12 of these patients.

The microbiological study of AF was positive in 59 patients (37%) and the remaining patients had a negative culture of neutrocytic ascites. Eighteen patients (11%) had bacteremia (Table 1).

Antibiotic therapy was started within 24 hours of diagnosis in 156 patients (98%) and the mean treatment duration was 8.3 days (SD ± 4.2). The failure rate of antibiotic treatment was 35%. No significant biochemical improvement was seen at 48 hours in 32 patients and coverage was adjusted based on susceptibility testing in 23 patients.

Table 1 Clinical and laboratory characteristics of the study patients 

Summary of results of the descriptive study of demographic, laboratory, clinical and microbiological variables. OH: alcohol; HCV: hepatitis C virus; HBV: hepatitis B virus; DM: diabetes mellitus; CKF: chronic kidney failure; HCC: hepatocellular carcinoma; HIV: human immunodeficiency virus; PPI: proton pump inhibitor; HE: hepatic encephalopathy; UGB: upper gastrointestinal bleeding; AKI: acute kidney injury; MBP: mean blood pressure; WBC: white blood cells.

Analysis of mortality

In-hospital mortality was 23%. Mortality rates at 30 days, 90 days, one year and two years were 21%, 31%, 55% and 68.8%, respectively. Mean survival after two years of follow-up was 368 days (SD ± 23.8) (Fig. 1). Four patients underwent transplants during follow-up. Even though all patients were discharged with antibiotic prophylaxis, the relapse rate was 22%.

Fig. 1 Cumulative overall survival of the patient cohort after two years of follow-up. 

Analysis of short-term mortality (30 and 90 days)

Table 2 summarizes the prognostic factors in the univariate study. Chronic renal failure, unit of hospitalization and development of hepatic encephalopathy or AKI were short-term prognostic factors for mortality. Bacteremia increased mortality at 30 days (p = 0.007) but not at 90 days.

Of all the laboratory parameters examined, WBCs in plasma were significantly related with 30-day mortality (p = 0.012) but did not follow a linear distribution. The continuous variable was categorized into three groups: a) WBCs in the normal range; b) leukopenia (WBCs < 4,500/mm3); and c) leukocytosis (> 11,000/mm3). At 30 days, leukopenia and leukocytosis decreased survival (leukocytosis OR 4.78, 95% CI 1.8-12.24, p = 0.002; leukopenia OR 3.44, 95% CI 1.8-12.6, p = 0.026) and no differences in mortality were found between patients with leukocytosis or leukopenia (p = 0.57). No significant differences were found with regard to short-term mortality between patients with or without volume expansion with SA (at 30 days p = 0.43, and at 90 days p = 0.84).

Table 2 Univariate study of prognostic factors of mortality 

Summary of the results of the univariate study of short- and long-term prognostic factors of mortality in demographic, laboratory, clinical and microbiological variables. NS: not significant; DM: diabetes mellitus; HIV: human immunodeficiency virus; CKF: chronic kidney failure; HCC: hepatocellular carcinoma; PPI: proton pump inhibitor; HE: hepatic encephalopathy; UGB: upper gastrointestinal bleeding; AKI: acute kidney injury; AF: ascitic fluid; WBCs: white blood cells; BUN: blood urea nitrogen; PT: prothrombin time.

With regard to the sub-analysis based on SA treatment indication, mortality was significantly lower in the group of patients who received volume expansion without an indication than in patients who received SA based on a clinical indication or who were not given SA (at 30 days, OR 0.08, 95% CI 0.01-0.63, p = 0.003; at 90 days, OR 0.15, 95% CI 0.04-0.51, p = 0.001). Mean survival at 90 days was 65.2 days (SD 3.1) versus 86.6 (SD 2.5 p = 0.001). Short-term mortality was higher when antibiotic therapy failed (at 30 days, OR 3.4, 95% CI 1.54-7.48, p = 0.002; at 90 days, OR 4.4, 95% CI 2.15-9, p = 0.001), regardless of the reason for the change of antibiotic. Failure of antibiotic treatment was significantly related to risk of bacteremia (OR 3.4, 95% CI 1.23-9.71, p = 0.01).

In the multivariate analysis (Table 3), the presence of leukopenia or leukocytosis, hepatic encephalopathy, elevated BUN and unit of hospitalization were independent risk factors for 30-day mortality. At 90 days, BUN and the unit of hospitalization continued to be independent factors, as well as age, temperature > 37.5 °C, creatinine and serum bilirubin.

Table 3 Multivariate study of short- and long-term mortality 

HCC: hepatocellular carcinoma; PPI: proton pump inhibitor; AKI: acute kidney injury; HE: hepatic encephalopathy; PT: prothrombin time.

Analysis of long-term mortality (one and two years)

In the univariate study of long-term mortality (Table 2), patients treated with proton pump inhibitors (PPIs) had a higher mortality (p = 0.02). When patients on chronic antibiotic treatment (with norfloxacin or rifaximin) were considered as a single group, bowel decontamination had a protective effect at one year (OR 0.32, 95% CI 0.13-0.78, p = 0.012) and at two years (OR 0.25, 95% CI 0.083-0.76, p = 0.015).

After two years of follow-up, no significant increase in mortality was found in patients with SBP relapse OR 0.5 (95% CI 0.23-1.10; p = 0.89). No significant differences were found with regard to long-term mortality between patients with and without volume expansion with SA (at one year p = 0.7, and at two years p = 0.8). One-year survival was 209.8 days (SD 14.3) in patients who received SA for volume expansion without having an indication, as compared to 291.1 days (SD 18.8) in patients administered SA based on a clinical indication or not given SA (p = 0.009; OR 0.37, 95% CI 0.17-0.82 p = 0.01). There were no significant differences in mortality between patients who did not receive SA or who received SA based on an indication (p = 0.5).

There were no significant differences in long-term mortality in patients in whom antibiotic treatment failed (p = 0.79 at one year, p = 0.94 at two years). Age and hepatic encephalopathy at diagnosis were independent predictors of mortality at one and two years (Table 3).

Assessment of prognostic indices

At 30 days, neither CPT (p = 0.25) nor CCI (p = 0.93) were significantly related to mortality. However, there was a trend for a longer survival in Child A patients. MELD was found to be a good predictor of short-term mortality (OR 1.067, 95% CI 1.023-1.4, p = 0.002). Survival with a MELD ≥ 15 was 24.4 days (SD 0.95) and that of patients with MELD < 15 was 28.5 days (SD 0.8, p < 0.005). In ROC curve analysis, the AUC for short-term mortality was 0.71 (95% CI 0.63-0.79, p = 0.001). The optimum cut-off point was a MELD value of 16.5.

At two years, all three indices were predictors of mortality according to Kaplan-Meier curve analysis: CPT (p = 0.03), CCI (p = 0.001) and MELD (p = 0.021). Survival analysis (Fig. 2) showed that MELD continued to be a predictor of higher mortality in patients with scores ≥ 15, AUC 0.63 (95% CI 0.53-0.72, p = 0.008).

Fig. 2 Cumulative survival after two years of follow-up categorized by MELD, Charlson Comorbidity Index and Child Pugh Score. 

With regard to CCI, survival was poorer in patients with CCI ≥ 8 (p = 0.001) and there were no differences between patients with a lower CCI (p = 0.79). Mean survival after two years of follow-up was 447 days (SD 30.3) in patients with CCI < 8, and 240 days (SD 32.3) in those with CCI ≥ 8 (p = 0.001). The ROC curve had an AUC of 0.68 (95% CI 0.6-0.77, p = 0.001) with a CCI of 7.5, which was an optimum cut-off point according to the Youden index.

Finally, the analysis of CPT as a prognostic factor for long-term survival showed that Child C patients had a greater mortality than those with Child A (p = 0.011). Child B patients have a dual behavior. During the first few months, the mortality rate was very similar to that of Child C patients but the survival curve then became less steep. Therefore, despite a clear trend, survival was not statistically significantly higher in Child B patients as compared to Child A patients (p = 0.2). ROC analysis of CPT, considered as a continuous variable, was not statistically significant (AUC 0.58; p = 0.08).