Introduction

Spontaneous bacterial peritonitis (SBP) is one of the most devastating complications of cirrhosis and is associated with an extremely high mortality rate ranging from 20 to 40% [1]. SBP is an infection of ascitic fluid and occurs frequently in patients with advanced cirrhosis [1]. The pathophysiology underlying SBP is considered multifactorial, primarily mediated through the impaired intestinal immunity in decompensated cirrhosis along with overgrowth of gut microbiota, which predisposes to bacterial translocation into mesenteric lymphatics and subsequently to the peritoneum [2]. An insufficiency of peritoneal antibacterial immunity is considered a secondary prerequisite as it permits colonization of translocated bacteria towards the development of SBP following systemic dissemination of the pathogen.

The delay in SBP diagnosis and treatment initiation is coupled with increasing risk of death [3]. Thus, prompt diagnosis and administration of appropriate antibiotics is the critical determinant for successful management, which requires the identification of the causative pathogen. Obligate anaerobic bacteria outnumber aerobic bacteria by more than 100-fold in gut microbiota, however they rarely translocate from the gastrointestinal (GI) tract to the ascitic fluid [4]. Conversely, specific bacteria types (such as gram-negative organisms) are more adept at translocating to mesenteric lymph nodes as compared to other bacteria types (such as gram-positive organisms) conceivably as a result of a greater ability to adhere to the intestinal mucosal surface [5]. Consequently, gram-negative organisms are more commonly isolated from the ascitic fluid of SBP cases; thereby forming the basis of guidelines recommending the empiric use of a third-generation cephalosporin as the first choice and this has traditionally been the standard of care for the last few decades [6]. This recommendation also applies for the treatment of culture-negative neutrocytic ascites (CNNA), as this entity is considered a SBP subtype except that the ascitic culture fails to yield the responsible pathogen.

On the other hand, the epidemiology of gut microbiota has been changing rapidly over the past few decades. The increasing rates of SBP cases with gram-positive and multi-drug resistant (MDR) organisms have been documented widely in studies from European countries as well as the U.S. [1, 7]. While these epidemiological changes presumably involve multifaceted mechanisms, it has been speculated that the widespread abuse of antibiotics worldwide is one of the predominant factors. Consequently, there has been an exponential increase in the number of SBP cases with bacterial pathogens that are resistant to the first-line therapy; thereby associated with worse outcomes including mortality even with prompt transition to 2nd-line therapies [1]. The evolution of MDR organisms also compromises the preventative management of SBP. Prophylactic antibiotic therapy has been shown to effectively reduce the incidence of SBP; however, SBP in individuals who are on prophylaxis have a higher risk of MDR microorganisms, which are associated with worse prognosis [8].

These observations collectively indicate that the characteristics of bacterial pathogens have a direct relevance to the clinical presentations including outcomes and therapeutic decision-making in patients presenting with SBP. While the clinical significance of MDR bacteria has been well established, there has been sparse information on the impact of bacterial taxonomy on clinical presentations and outcomes. To date, there has not been a study investigating the pathogenesis of individual organisms in patients with SBP. Consequently, the purpose of this study is to elucidate the epidemiology and microbiology of bacterial pathogens isolated from the ascitic fluid of SBP cases and their correlation with clinical presentation and outcomes, which can be exploitable to improve the outcomes of patients with SBP.

Materials and Methods

Study Design and Patient Populations

This is a retrospective study conducted at two large academic tertiary care centers in Los Angeles, California. The study was approved by the Institutional Review Board at both LAC + USC Medical Center and Keck Hospital of USC.

The study included adults (≥ 18 years) hospitalized at either hospital from 01/2015 to 01/2021 for SBP in patients with cirrhosis. Cirrhosis was determined based on clinical, laboratory, histopathology, or radiographic data. SBP was defined as ascitic absolute neutrophil [often referred to as polymorphonuclear (PMN)] count ≥ 250 cells/mm3 without evidence of intra-abdominal, surgically treatable source of infection on either diagnostic or therapeutic paracentesis [9]. Culture-positive SBP is defined as the growth of one species of bacteria or fungus in ascitic fluid within five days of culture set-up [10]. All patients received standard of care, empiric antibiotics upon presentation.

Cases with suspected peritoneal carcinomatosis, hepatocellular carcinoma outside of the Milan criteria, polymicrobial growths (≥ 2 organisms), or peritoneal dialysis were excluded given their higher likelihood of secondary peritonitis. In addition, patients with cirrhosis due to autoimmune hepatitis who were on immunosuppressive therapies, or previous liver transplant (LT) cases were excluded. When repeat episodes of SBP were noted, we captured the last episode since patients may have had SBP at outside hospitals and been referred our tertiary center for subsequent episodes.

Data Collection

Data were collected retrospectively from 01/2015 to 01/2021, ascertaining patient demographics, etiology of cirrhosis, MELD-Na score, ascitic PMN counts, albumin at diagnosis of SBP, types of causative pathogens, MDR status, SBP prophylaxis use, mortality, release to hospice or liver transplant (LT). MDR is defined as resistance to at least three antimicrobial drug classes, extensively drug resistant if only sensitive to one or two classes and pandrug-resistant if resistant to all classes [11].

Outcomes

The primary outcome was SBP progression, the composite event of mortality or LT. Mortality was defined as death from any cause or release to hospice because these patients presumably die in hospice and their deaths are not readily reported to hospitals. LT is included as a primary endpoint as it is the only curative treatment for decompensated liver failure in patients with high MELD-Na scores who would have died if they did not receive a LT.

Statistical Analysis

Patient characteristics are presented as frequencies (percentages) and medians (interquartile ranges (IQR)) by bacterial pathogen. Differences by outcome were assessed using χ2 and Wilcoxon rank sum tests, respectively, applying exact tests when needed.

The Kaplan–Meier method was used to estimate the cumulative incidence of SBP progression at 1- and 6 months after the diagnostic paracentesis by the type of causative bacteria. Event rates were compared using the log-rank test with Sidak adjusted p-values to account for multiple comparisons. Patients were followed from the time of the diagnostic paracentesis to SBP progression or date of last follow-up with the hospital system (censored). Patients were censored at last follow-up to avoid assuming that those without 1- or 6 months follow-up did not experience the event. Cox regression estimated hazard ratios (HR) and 95% confidence intervals (CI) for risk of SBP progression within 1-month by bacterial type. We chose the 1-month outcome as most events could be attributed to SBP and for comparison to prior studies [10]. Bacterial pathogens were modeled as Klebsiella, Streptococcus, or other based on the Kaplan–Meier curves demonstrating similar event rates for the remaining bacteria. MDR and MELD-Na were selected a priori for inclusion in the multivariable model because they are well established predictors for mortality in SBP patients. Multicollinearity was assessed with variance inflation factors less than 10 confirmed. Analysis was completed using SAS version 9.4 (SAS Institute, Inc., Cary, NC).

Results

267 cirrhosis patients whose PMN count in ascitic fluid exceeded 250 cells/mm3 were identified, of which ascitic fluid cultures yielded a causative pathogen in 88 patients as follows: E. coli (33%), Streptococcus (15%), Klebsiella (13%), Enterococcus (13%), Staphylococcus(9%) and others (Pseudomonas, Enterobacter, Listeria, Pasteurella, Citrobacter, Lactobacillus, Serratia and fungal; 18%). Two patients had SBP due to fungal infections. The median age at paracentesis was 57 years (IQR 52–64); 68% male; 59% Hispanic (Table 1). Cirrhosis was predominantly caused by alcohol (61%) with other etiologies being hepatitis C (18%), non-alcoholic fatty liver disease (12%), hepatitis B (5%), primary biliary cholangitis (1%), and cryptogenic or unknown (3%). Patients with end-stage renal disease on hemodialysis comprised 23% of the study group. Only 30 (34%) patients had a prior episode of documented SBP with 27 (31%) patients on SBP prophylactic antibiotics at time of admission.

Table 1 Demographics by bacteria organism
Full size table

Upon presentation, baseline characteristics and laboratory findings were obtained (Table 1). Overall, median MELD-Na was 29 (IQR 23–35), lowest for Streptococcus at 28 (IQR 16–30) and highest for Enterococcus and Staphylococcus at 32 (IQR 25–36 and 23–35, respectively). White blood cell count, systemic inflammatory response syndrome (SIRS), concurrent bacteremia, and serum albumin were similar by bacteria type. The overall median lactate was 3.6 (IQR 2.2–7.2) with Klebsiella having the highest value at 13.9 (IQR 2.7–11.0). Median absolute PMN leukocyte count was 2723 cells/mm3 (IQR 866–6381) with Streptococcus having the highest at 5500 cells/mm3 (IQR 2275–8424). Overall, 64 patients (73%) met SIRS diagnosis criteria on admission, in which all cases with Klebsiella had SIRS (100%). At time of paracentesis, 14 (16%) of total patients were on vasopressors with only one case associated with Streptococcus (7.7%). MDR status was seen in 36 (41%) patients with E. coli having the highest frequency (76%) and Streptococcus the lowest (0%). Of the organisms with MDR, 18 (50%) were extensively drug resistant with Klebsiella having the highest XDR at 3 (75%) (Table 1). When assessing for ESBL in patients with MDR, E. coli had 20 (91%) and Klebsiella had 2 (50%). Furthermore, 5 (83%) of patients with MDR Enterococcus had Vancomycin-resistant Enterococcus. No patients had documented Carbapenem-resistant Enterobacteriaceae.

SBP progression occurred in 57 patients, 45 (79%) within 1 month including 10 having LT, and 35 died of which 3 were released to hospice (Table 2). The cumulative incidence of SBP progression was highest for Klebsiella and lowest for Streptococcus (p < 0.001) with no statistically significant differences detected for the remaining bacterial types (Fig. 1). At 1-month after paracentesis, the cumulative incidence of SBP progression for Klebsiella was 91% (95% CI 67–100), Staphylococcus 66% (95% CI 32–95), Enterococcus 64% (95% CI 35–91), other 64% (95% CI 4–87), and E. coli 59% (95% CI 42–76) while Streptococcus was 16% (95% CI 4–51). For comparison, 1-month cumulative incidence of SBP was 40% (95% CI 32–48) for culture-negative neutrocytic ascites (Supplemental Fig. 1).

Fig. 1
figure 1

Kaplan–Meier cumulative curve of SBP progression stratified by type of bacteria

Full size image
Table 2 Outcomes by bacterial organism
Full size table

Clinical characteristics are compared by SBP progression within 1-month (Table 3). The distribution of bacteria type differed significantly by outcome (p < 0.001); a higher percentage with Klebsiella had SBP progression (22%) versus no progression (2%) and lower percentage with Streptococcus (4% vs 26%, respectively). Median MELD-Na (33 vs 24, p < 0.001) and MDR (21% vs 15%, p = 0.19) were higher for those with SBP progression, although the latter did not achieve statistical significance. Patients with SBP progression also had higher median WBC (12.8 vs 8.4, p = 0.003), lactate (6.4 vs 2.4, p ≤ 0.001), frequency of SIRS (38% vs 26%, p = 0.016), and vasopressor requirements (13% vs 1%, p ≤ 0.001) at presentation.

Table 3 Characteristics of those with SBP Progression within one month of paracentesis versus those without
Full size table

When examining 1-month risk of SBP progression in univariable analysis, risk increased with the presence of Klebsiella (HR 2.68, 95% CI 1.31–5.49) and decreased with the presence of Streptococcus (HR 0.18, 95% CI 0.04–0.77) compared to all other bacteria (Table 4). Additionally, for every 1-unit increase in MELD-NA, the risk of SBP progression increased by 11% (95% CI 1.06–1.16) while MDR had a 61% increased risk of SBP progression (95% CI 0.89–2.89) but was not statistically significant. After adjusting for MELD-Na and MDR in a multivariable analysis, SBP progression remained elevated for Klebsiella with a twofold increased risk (HR 2.07, 95% CI 0.988–4.24) and decreased for Streptococcus with HR of 0.28 (95% CI 0.06–1.21) compared to all other bacteria.

Table 4 Cox proportional hazard model for risk of SOP proyraaslon within one month of paracentesis
Full size table

Discussion

Overall, the development of SBP is associated with higher mortality and morbidity rates. In particular, SBP cases with renal dysfunction, high MELD scores, and hospital-acquired are associated with poor clinical outcomes. Apart from these host factors, our understanding of the role of bacterial virulence factors has been limited, with the exception of the link between MDR status of the causative pathogen and poor clinical outcomes [8]. Accordingly, little is known whether the bacterial pathogen correlates with clinical outcomes as compared to other identifiable bacteria. Our study revealed that, for the first time, SBP cases with Klebsiella pneumoniae are associated with worse outcomes such as high mortality, transfer to hospice or liver transplantation, as demonstrated by a hazard ratio twice that compared to other identifiable bacteria. This is further supported when comparing SBP progression for Klebsiella compared to culture-negative neutrocytic ascites (Supplemental Fig. 1). Moreover, our study showed that SBP with Klebsiella pneumoniae, regardless of MDR status, is associated with the highest WBC levels, lactate levels and highest percentage of meeting SIRS criteria. Conversely, cases with Streptococcus have the most favorable outcomes, even after adjusting for MDR and MELD-Na. These observations collectively suggest that the type of bacteria, along with sensitivity to empiric therapy with 3rd generation cephalosporin, might serve as a critical and independent determinant for clinical outcomes of SBP.

Classically, Klebsiella pneumoniae has been considered a commensal and opportunistic microorganism; therefore, the progression to overt infection requires additional risk factors, which include but are not limited to diabetes, alcohol use, fluid, and electrolyte disorders [12], all of which are highly relevant to patients with advanced cirrhosis, the patients represented in our study. Historically, Klebsiella pneumoniae has accounted for a relatively small proportion SBP cases compared with other pathogens such as E. coli, consisting of approximately 10% of all SBP cases depending on the region and ethnicity [13]. However, the number of SBP cases with Klebsiella pneumoniae is expected to grow as epidemiological studies demonstrated an increasing prevalence of colonization over the last decades both at the community level as well as in hospitalized patients. Currently, the rate of community acquired Klebsiella pneumoniae colonization in nasopharyngeal and GI tract are as high as 15% and 35%, respectively. Of note, the rate of GI tract colonization in hospitalized patients, especially those who had been exposed to antibiotics, has been reported to be nearly 80% [14,15,16]. Since the GI tract is the predominant reservoir site for the SBP pathogens, the endemic colonization strongly suggests a spike in the number of cases with Klebsiella pneumoniae [17].

The increasing Klebsiella prevalence is concerning; however, this alone does not fully explain the previously unrecognized high pathogenicity of SBP with Klebsiella pneumoniae, for which the evolution of virulence factors likely serves as an explanation according to cumulative evidence. Recent epidemiological studies with molecular and genetic approaches demonstrated the emergence of strains harboring genes responsible for enhanced virulence [18, 19]. The acquisition of virulence factors as well as antibiotics resistant genes are largely mediated through genetic alterations of the accessory genome [20,21,22,23,24,25,26]. For example, the well-accepted hyper-virulence factors, RmpA, K2A, and MagA genes are all encoded in the accessory genome [27, 28]. Due to the high flexibility of the accessory genome for the horizontal gene transfer among species, the emergence of dual-risk strains that carry both antibiotic-resistant and hypervirulent genes is becoming evident, which makes it progressively pathogenic and difficult to treat. This concept is not only applicable for SBP pathogenesis but also for infections in other organ systems. Indeed, over the past 30 years, the increasing incidence of liver abscesses and septic metastasis caused by hypervirulent strains of Klebsiella pneumoniae has been reported, and has emerged as a leading cause of pyogenic liver abscesses, initially documented in several Asian countries and now in the United States [29].

Detection of virulence factors requires the genotyping of the accessory genome which is beyond the capacity of a typical clinical microbiology laboratories [30]. Accordingly, this is one of the limitations of the present study, as we are unable to report on virulence factors and thus, cannot definitively correlate poor outcomes in cases with Klebsiella pneumoniae results with the presence of bacterial virulence. An assessment of hypervirulent Klebsiella pneumoniae carriage rates at the local community, regional, and national levels is required to define the generalizability of our study results. Another limitation is the relatively small number of patients for each identified organism group. However this study contained one of the largest number of culture-positive SBP subjects in the United States. Taken together, further prospective molecular epidemiological studies with a larger number of subjects along with molecular genetic approaches are warranted.

Conclusion

Our study is the first to demonstrate the significance of bacteria type on clinical outcomes after adjusting for known risk factors (MDR and MELD-Na) in cirrhotic patients presenting with SBP. Currently, ascitic fluid cultures are performed mainly to determine MDR status by performing sensitivities; however, our study highlights the importance of determining bacterial types for management strategy and prognostication. Lastly, this study is the first to report worse clinical outcomes with Klebsiella and favorable clinical outcomes with Streptococcus after adjusting for MDR and MELD-Na.