Antimicrobial resistance in patients with decompensated liver cirrhosis and bacterial infections in a tertiary center in Northern Germany

Background and aims Bacterial infections are common in patients with decompensated liver cirrhosis and a leading cause of death. Reliable data on antibiotic resistance are required to initiate effective empiric therapy. We here aim to assess the antimicrobial resistance profile of bacteria among patients with liver cirrhosis and infection. Methods Overall, 666 cirrhotic patients admitted to Hannover Medical School between January 2012 and April 2018 with ascites were assessed for bacterial infection. In case of infection, bacteria cultured from microbiological specimens of ascites, blood or urine were identified and analyzed for resistances against common antibiotic agents. Furthermore, analyses compared two periods of time and community-acquired vs. nosocomial infections. Results In 281 patients with infection, microbiological sampling was performed and culture-positive results were obtained in 56.9%. Multidrug-resistant (MDR)-bacteria were found in 54 patients (19.2%). Gram-positive organisms were more common (n = 141/261, 54.0%) and detected in 116/192 culture-positive infections (60.4%). Comparing infections before and after 2015, a numerical decline for MDR-bacteria (23.8% vs. 15.6%, p = 0.08) was observed with a significant decline in meropenem resistance (34.9% vs. 19.5%, p = 0.03). MDR-bacteria were more frequent in the case of nosocomial infections. Of note, in ascites the majority of the tested bacteria were resistant against ceftriaxone (73.8%) whereas significantly less were resistant against meropenem (27.0%) and vancomycin (25.9%). Conclusions In our tertiary center, distinct ratios of gram-positive infection with overall low ratios of MDR-bacteria were found. Adequate gram-positive coverage in the empiric therapy should be considered. Carbapenem treatment may be omitted even in nosocomial infection. In contrast, 3rd generation cephalosporins cannot be recommended even in community-acquired infection in our cirrhotic population. Supplementary Information The online version contains supplementary material available at 10.1186/s12876-021-01871-w.


Introduction
Bacterial infections increase the mortality of patients with liver cirrhosis up to four times [1] and are one of the leading causes of death in this population [2]. Particular hazards are numerous consecutive nosocomial infections decreasing liver transplant-free survival significantly [3].
The most frequent sites of infection are spontaneous bacterial peritonitis (SBP) and urinary tract infections (UTI) [2][3][4]. Patients with liver cirrhosis have a higher susceptibility for infections resulting in an increased risk for sepsis and sepsis-related mortality [5]. Early empiric antibiotic treatment can decrease the mortality of severe sepsis and is strongly recommended [6][7][8][9][10].
In 2018 the European Association for the Study of the Liver (EASL) published a guideline for the management of patients with decompensated cirrhosis [11]. Empirical antibiotic treatment recommendations for communityacquired SBP and bloodstream infection (BSI) are 3rd generation cephalosporins and meropenem for nosocomial SBP and BSI. Treatment recommendations for UTI include ciprofloxacin or nitrofurantoin. In addition, it is recommended to consider the local resistance profiles [11,12]. The EASL guideline distinguishes between countries with high and low ratios of bacterial resistance [11].
International studies with individual data for single countries and regions show a prevalence of multidrugresistant (MDR)-bacteria in patients with cirrhosis in Germany between 26% [13] and 34.1% [14] of culturepositive patients. Importantly, the prevalence varies significantly between local regions, e.g., 4.2% and 41.2% [14].
Antibiotic therapy is challenged by several trends. An increasing number of gram-positive pathogens can be found in patients with SBP [4,15,16] which makes 3rd generation cephalosporins less effective [17]. Additionally, due to frequent hospitalization and regular use of antibiotics in patients with liver cirrhosis, cirrhotic patients have a higher risk for developing infections caused by MDR-bacteria [18]. Those infections are associated with a higher incidence of treatment failure, septic shock, and hospital mortality ] [13,14,18] and are increasing dramatically according to several studies [4,18,19].
The EASL recommendations for the antibiotic treatment of patients with liver cirrhosis are based on pan-European studies on pathogens and their antibiotic resistance and aim to provide valid recommendations across Europe. However, given the changing epidemiology of pathogens and resistances as well as local regional differences, it may be better to rely on local data than global recommendations.
With the present study, we aim to present a big database of frequently found pathogens and their resistances in microbiology samples taken during SBP, BSI, and UTI from patients with decompensated cirrhosis and question whether the EASL recommendations are adequate in our region. In addition, further analyses distinguish between two periods of time to assess trends in MDR-infection and between community-acquired vs. nosocomial infections. Based on these findings we aim to provide empiric treatment recommendations for our tertiary center in Northern Germany.

Patient cohort
The patients were recruited from the Hannover Ascites Cohort, which considered all consecutive patients with liver cirrhosis who underwent paracentesis during their hospital stay at the Hannover Medical School between January 2012 and April 2018. The patients were automatically identified retrospectively by the Enterprise Clinical Research Data Warehouse (ECRDW) to minimize a potential selection bias. Overall, 1314 patients were screened and 666 patients were included after manual assessment of the patient's charts (Fig. 1). Routine laboratory data were collected using the ECRDW. Routine clinical data including the date of infection and antibiotic susceptibility testing (AST) results were collected manually from the patients' files. Diagnosis of liver cirrhosis was based on ultrasound, FibroScan (≥ 14.5 kPa), biochemical results, and/or liver biopsy.
Exclusion criteria were lack of evidence of cirrhosis, extrahepatic malignancy, hepatocellular carcinoma beyond the Milan criteria, secondary intraabdominal inflammation, infection with HIV, congenital immune dysfunction, non-liver organ transplantation, lack of data, and absence of written informed consent.

Ethics
The local ethics committee of Hannover Medical School approved the study (ethic no. 7935_BO_K_2018 from June 22nd 2018) and patients provided written informed consent for the scientific use of their clinical data. The

Data assessment
The observation period included the complete hospital stay until discharge, transfer to another hospital or nursing home, liver transplantation, or death. Infections were diagnosed by the attending physician or in case of uncertainty by infectiologists (BM, MC, JR, CHZS) based on clinical symptoms and the following laboratory results: • SBP: ≥ 500 nuclei containing cells/µl ascites fluid [20] • BSI: positive blood cultures • UTI: leukocyturia and/or positive urine cultures with significant germination number (≥ 10 5 CFU/ml urine).
In this study, ascites and blood were considered sterile. Therefore, all the pathogens found in ascites or blood during infection were included in the analyses. Pathogens found in urine during infection were only included if at least 10 5 CFU/ml urine were detected except for Escherichia coli according to local guidelines [21].
Infection was classified based on the time from admission as community-acquired (< 48 h) or nosocomial (> 48 h) [22]. If the patient was hospitalized within the previous three months, infections in the first 48 h after admission were classified as nosocomial.
Pathogens were counted once per site and patient only. In case, a pathogen was redetected, only the first Antimicrobial Susceptibility Testing was included in our study. For easier reading, a sample with a detected pathogen is called positive.
The AST was evaluated according to European Committee on Antimicrobial Susceptibility Testing (EUCAST) international guidelines and expert microbiologists (RV, SZ). Resistant results include intrinsic resistances as well as acquired resistances. Intermediate results were stated as resistant.
Definitions for multiresistant bacteria are not consistent. For better distinction the existence of one of the following resistances classified the pathogen as severe resistant: • Vancomycin-or Linezolid-resistant Enterococcus  Table 1).
Enterobacteriaceae, Staphylococcus aureus, Enterococcus spp., and Acinetobacter spp. were classified as multidrug-resistant (MDR) if the isolate is non-susceptible to at least 1 agent in ≥ 3 antimicrobial categories [24]. In our study, all pathogens with a severe resistance met MDRcriteria as well. Additional file 2: Table 2 shows the antimicrobial categories and agents used for the definition of MDR.

Study design
Three main analyses are part of this study. In the first analysis, the development of antibiotic resistance over time was observed. For this part, the cohort was divided into two groups. The first one ("period 1") contains the samples that were taken during an infection between January 1st 2012 and December 31st 2014. The second period ("period 2") includes the samples that were taken between January 1st 2015 and April 30th 2018. The analysis compares the number of severe resistances, MDR, and the ratios of resistances against common antibiotics.
The second analysis presents the ratios of bacteria with resistance against common antibiotics in SBP, BSI, and UTI. In a further analysis, these samples are subdivided into community-acquired and nosocomial infections.
The third analysis evaluates the treatment options for SBP, BSI and UTI recommended by EASL. For this part, we analyzed the ratios of patients with resistance against the recommended antibiotics in mono-and combination-therapy. Suboptimal antibiotic selection was assumed in patients if at least one pathogen was resistant to all applied antibiotic agents.

Statistics
Statistical analyses were performed with SPSS (IBM SPSS Statistics 26). Continuous data were analyzed with unpaired t-test and categorical data with Fisher´s exact test.

Baseline characteristics
Overall, 666 patients were included in this study. The majority (64.3%) was male and the median age was 56.7 (17-87) years. The major causes of liver cirrhosis were alcohol (50.5%) and hepatitis C virus infection (13.8%) with a median MELD-Score of 17.6. In half of the patients, at least one infection was diagnosed. The most common infection was SBP followed by UTI and BSI (Table 1).
Microbiological sampling was performed in 281 patients, 56.9% of them had at least one culture-positive sample. Samples, in which more than one pathogen was detected, were found in 46 patients (16.4%). Pathogens with severe resistance as defined above were found in twelve patients (4.3%). MDR-bacteria were found in 54 patients (19.2%). Table 1 Patients' characteristics * 1 Only patients with microbiological sampling during infection included. One patient was excluded because the time of his infection overlapped the cutoff date for both periods * 2 Comparison of the two groups "period 1" and "period 2" * 3 The 90 days were counted starting from the end of hospitalization

Changes in antibiotic resistance over time
To assess the development of antibiotic resistance over time, all 280 patients were compared depending on the date of infection. One hundred and twenty-six patients were allotted for period 1 (2012-2015) and 154 patients for period 2 (2015-2018). The baseline parameters of the two study groups are depicted in Table 1.

Antibiotic resistance in common locations for infection
Pathogens were detected in more than half of the urine samples taken during UTI (66.9%, n = 83/124) and blood cultures taken during BSI (56.1%, n = 55/98) but in less than a quarter of ascites samples taken during SBP (23.2%, n = 72/310). The highest ratio of MDR-bacteria was detected in ascites samples (ascites: 24.4%, n = 22/90; urine: 22.2%, n = 24/108; blood: 22.2%, n = 14/63). Pathogens with severe resistance were more likely found in urine samples (7.4%, n = 8/108; blood: 4.8%, n = 3/63; ascites: 3.3%, n = 3/90) (Additional file 4: Table 4). Table 4 shows the AST results for commonly used antibiotics for the three different types of infection. In ascites the majority of the tested bacteria were resistant against ceftriaxone (73.8%), whereas only around a quarter was resistant against meropenem (27.0%) and vancomycin (25.9%). Piperacillin/tazobactam showed similar effectiveness as meropenem with only 34.5% (n = 29/55) of the tested bacteria being resistant. Bacteria found in blood were slightly more resistant to commonly used antibiotics. In urine, 23 of the detected pathogens had resistances against at least four antibiotics commonly used for UTI (susceptibility and resistances for distinct antibiotics and combinations are depicted in Table 4).
For further analyses, the SBP and UTI cohorts were divided into community-acquired and nosocomial infections. This separation was not possible for BSI, because only three patients had a community-acquired infection. In ascites the ratio of patients with positive cultures was significantly higher in nosocomial SBP versus community-acquired infection (36.9% vs. 20.0%, p = 0.04). MDR-bacteria were slightly more frequent in nosocomial episodes (11.5% vs. 7.5%, p = 0.46). In addition, bacteria with severe resistances were only found in patients with nosocomial episodes of SBP. Similarly, for UTI the patients with nosocomial infections were more likely to have positive cultures compared with communityacquired UTI (79.5% vs. 72.7%, p = 0.61). Moreover, MDR-bacteria were also more frequent in patients with nosocomial UTI (24.1% vs. 18.2%, p = 0.66). However, patients with severe resistant pathogens were found   equally in both groups (nosocomial vs. communityacquired, 9.1% vs. 8.4%, p = 0.94). No major differences in resistances between community-acquired and nosocomial infections were found (Additional file 5: Table 5).

Application of treatment guidelines
To assess the effectivity of the EASL guidelines for the treatment of UTI, BSI or SBP at our center, we compared our local resistances with the EASL recommendations. For community-acquired SBP or septic UTI the recommended antibiotic regimen are 3rd generation cephalosporines or piperacillin/tazobactam [11]. For nosocomial infection, combination therapy is recommended if risk factors are present, usually a carbapenem plus vancomycin, daptomycin, or linezolid. In our cohort, the combination therapy with meropenem plus vancomycin showed susceptibility in all patients with culture-positive SBP and UTI, except for one patient with BSI (2.0%, n = 1/49). Combination therapy with ampicillin/sulbactam plus vancomycin showed corresponding resistances in 11.6% of patients with culture-positive SBP (n = 8/69) and 12.2% in patients with BSI (n = 6/49). Therapy with piperacillin/tazobactam plus vancomycin had corresponding resistances in 16.3% of patients with BSI (n = 8/46), 7.2% of patients with culture-positive SBP (n = 5/69) and 0.0% of patients with culture-positive UTI (n = 0/74). The combination of ceftriaxone plus vancomycin was affected by corresponding resistances in 7.5% (n = 12/160) of patients with culture-positive infections. The highest ratio was shown in patients with culturepositive SBP (10.1%, n = 7/69) followed by BSI (8.2%, n = 4/49) and UTI (1.4%, n = 1/74).

Discussion
Cirrhosis is associated with a high frequency of infections, most notably SBP and UTI. To decrease resulting complications and mortality, early diagnosis and adequate empiric treatment based on the knowledge of local antibiotic resistances are necessary.
We here show that (1) a distinct ratio of infections was caused by gram-positive bacteria, (2) 3rd generation cephalosporins, commonly recommended as firstline therapy, were subject to a high ratio of antimicrobial resistance in our center, even in community-acquired infections, (3) the overall frequency of infections caused by MDR-bacteria was declining and resistance to meropenem was declining in particular and (4) other antibiotics seemed to be almost equally effective as meropenem even for the treatment of nosocomial infection, in particular, if combination therapy with vancomycin was applied.
In our study, the 90-day mortality was 30.5% for patients with infection (n = 108/354). This ratio is a little bit lower compared to other studies, whereby there is a big range (1). The 90-day mortality differed according to the time of infection. Whereas patients with nosocomial infection showed a 90-day mortality of 34.4% (n = 99/288), the ratio was significantly lower for patients with community-acquired infection (13.2%, n = 7/53, p = 0.002).
Whereas in the past the majority of infections in cirrhotic patients were caused by gram-negative bacteria [25,26], we currently observe a shift to the gram-positive bacteria. Several reasons have been linked to this shift: the more frequent use of invasive procedures [4], the frequent use of proton pump inhibitors (PPI) with a consecutive shift in intestinal microbiota [27], and alcohol intake with subsequent failure of the intestinal barrier and bacterial translocation [28]. In our cohort 542 patients (81.4%) had PPI in their daily medication and 336 patients (50.5%) had alcohol-related liver damage. However, no difference in ratios of gram-positive bacteria were found (ASH vs. non-ASH: 53.5%, n = 69/129 vs. 54.5%, n = 72/132; p = 0.86. PPI vs. non-PPI: 54.3%, n = 127/234 vs. 50.0%, n = 12/24; p = 0.69). Gram-positive bacteria were especially frequent (72.2%) in ascites which is in line with recent studies [15,17,29]. This is important to acknowledge as single antibiotic therapy will usually fail in our cohort, and the addition of vancomycin drastically increases the chances of effective antibiotic therapy. However, when interpreting these data, it has to be considered, that in some cases the detected pathogens belonged to the group of facultative pathogenic bacteria. Thus, we have to acknowledge, that the findings of e.g., certain gram-positive pathogens may be due to contamination in some cases while the true origin of infection is to be found elsewhere.
Third-generation cephalosporins are a first-line recommendation for community-acquired SBP [30,31]. Our results showed high ratios of resistances against ceftriaxone in patients with SBP (73.8%). Resistance was more frequent in nosocomial episodes (75.0%) compared to community-acquired episodes (62.5%) and increased over time by 28.0% (from 63.9% in period 1-81.8% in period 2, p = 0.07). This finding is in line with other recent reports [32][33][34][35]. Thus, special consideration should be given to the use of 3 rd generation cephalosporins according to local resistance profiles.
Resistance is usually higher in nosocomial infections than in community-acquired infections [18,33]. We did not find a significant difference in resistances between these two groups. However, similar to the results of Umgelter et al. [32], the majority of patients (81.4%, n = 288/354) in our study were classified as having a nosocomial infection. Other studies report lower rates [36,37]. The high rate of nosocomial infection is well explained by the fact, that we provide tertiary care for patients with liver diseases. Thus, patients are more likely to be referred from other hospitals and tend to have more advanced liver disease (median MELD 17.6). The majority of classifications as nosocomial infection were caused by recent hospitalizations for various reasons-a common event in patients with decompensated cirrhosis [38]. Only 20 patients were directly admitted for infection, whereas the vast majority came to our hospital for ascitic decompensation or evaluation of transjugular intrahepatic portosystemic shunt placement or liver transplantation and infection was diagnosed on further work-up. It should be noted, that repeated hospitalization is a common event in cirrhotic patients. Therefore, per definition community-acquired infection may be caused by a different group of bacteria, which may influence resistance weights.
Based on our results first-line empiric therapy with piperacillin/tazobactam plus vancomycin for nosocomial SBP and UTI and meropenem plus vancomycin for nosocomial BSI might be the best choice for our center. However, this assumption can't simply be transferred to other hospitals due to regional differences in resistance prevalence.
Due to repeated hospitalizations and exposure to antibiotics, cirrhotic patients are increasingly challenged by MDR-infections [39]. Comparison between different studies is difficult due to inconsistent definitions of multiresistance. We used the definition of Magiorakos et al. [24] as well as specific resistances like oxacillin-resistance in Staphylococcus aureus to describe our cohort. Overall, 60 (31.3%) of 192 culture-positive infections were caused by MDR-bacteria. This ratio was similar in all culture sites (SBP: 31.8%, n = 22/69; BSI: 28.6%, n = 14/49; UTI: 32.4%, n = 24/74) and is similar to results from other recent studies [18,40]. Compared to other German centers, we found no VRE in ascites during SBP [35,41], despite high ratios of Enterococcus spp., and could not detect an increase of MDR-bacteria [42]. However, parts of the gram-positive pathogens must be considered as possible contamination. Analyzing only patients with possible MDR-bacteria (Enterobacteriaceae, Staphylococcus aureus, Enterococcus spp. or Acinetobacter spp.) the ratio of MDR-classified pathogens decreased over time (data not shown). In patients with gram-negative pathogens ratio of MDR-bacteria did not change over time (data not shown). Therefore, the missing increase of MDR-bacteria in this study can't be fully explained by contamination. For single antibiotics, the resistance against ceftriaxone was distinctly higher in our cohort [17,29], whereas vancomycin resistance distinctly was lower [17]. In addition to local differences, some of the different findings may be explained by the time point of sampling, as antibiotic resistance may shift over time. Overall, the present data underline the recommendation to assess and track local rather than national resistance profiles.
In addition to the analyzed infections, pneumonia is common in patients with liver cirrhosis as well [2]. However, in this study cohort, respiratory material was infrequently collected or consisted mainly of sputum and the culture results were commonly positive for several bacteria of the oral flora. Unfortunately, these conditions were impeding analysis for patients with pneumonia.
This study has some limitations. First, it is based on retrospective data. As retrospective estimation of pathogenicity is difficult, we did not exclude pathogens that may be cultured due to contamination (i.e., coagulasenegative staphylococci, Corynebacterium, Propionibacterium, or Bacillus spp.). As mentioned before, this can partly explain the high ratio of gram-positive bacteria compared to other studies [33]. In addition, ratios of MDR-bacteria and other resistances could be stated too low due to including non-pathogenic germs. Second, there is a lack of microbiology sampling in many patients. Third, most patients had nosocomial infections by our definition, which makes the comparison between community-acquired and nosocomial infections difficult. Nevertheless, due to the large number of included patients and observed episodes of infections the results can help to adapt guidelines for empirical antibiotic treatment.
In conclusion, our study shows that (1) a distinct number of infections was caused by gram-positive bacteria, (2) 3rd generation cephalosporins had high ratios of antibiotic resistance at our center, (3) no increase in infections with MDR-bacteria occurred between 2012 and 2018, (4) for our center the first-line empiric therapy for patients with nosocomial SBP was sufficient with piperacillin/tazobactam, even despite declining resistances to meropenem. Thus, we can confirm the higher frequency