Bloodstream infection due to Escherichia coli in liver cirrhosis patients: clinical features and outcomes

Objectives The study aimed to investigate the clinical characteristics and antibiotic management, as well as independent indicators for survival within 30 days for Escherichia coli bloodstream infection (BSI) in liver cirrhosis. Results Hospital-acquired BSI accounted for 60.07%, with prolonged hospital stay (P = 0.000). The prevalence of Extended Spectrum Beta-Lactamases (ESBL) producing bacteria was 48.26%, which correlated with ICU admission (P = 0.015) and high model for end-stage liver disease (MELD) score at onset of BSI (P = 0.035). Moreover, ESBL producing pathogens showed a high resistant to the common antibiotic families and 27.5% pathogens were confirmed as multidrug-resistant (MDR). MDR infection was significantly correlated with ESBL production, ICU admission, inappropriate empiric therapy, resistance to firstly selected antibiotic, and infection duration (P < 0.05 for all). In addition, appropriate empiric therapy within 48 h (HR = 2.581, 95% CI = 1.166–5.715), ICU admission (HR = 4.434, 95% CI = 2.130–8.823), HE (HR = 2.379, 95% CI = 1.115–5.073) and final MELD (HR = 1.074, 95% CI = 1.044–1.106) were independent indicators for 30-day mortality. Materials and Methods The clinical data were collected from 288 eligible patients, and compared according to survival status and sites of infection acquisition. Drug resistance was recorded according to ESBL. In addition, cox regression analysis model was applied to evaluate the risk factors for 30-day mortality. Conclusions ESBL production can promote resistance to antibiotics in Escherichia coli. Antibiotic regimens, ICU admission, HE and MELD score can help identify the risk individuals who will benefit from the improved therapeutic regimens.


INTRODUCTION
Liver cirrhosis is a troublesome problem for public health worldwide, with high occurrence rate and mortality [1]. Cirrhosis is an advance stage of chronic liver disease, due to chronic infection with hepatitis virus, alcohol abuse, fat accumulation, autoimmune and metabolic alterations [2,3]. Various complications are frequently observed in cirrhosis patients, such as acute kidney injury, portal vein thrombosis, variable abdominal pain [4,5]. Among the complications, bacterial infection is a leading cause for death [6]. Unfortunately, patients with liver cirrhosis are more easily to be infected by pathogens due to impaired immunity, consequently poor outcomes [7][8][9].
Bloodstream infection (BSI) is a serious problem in many hospitalized patients, and is referred as being primary

Clinical Research Paper
Oncotarget 35781 www.oncotarget.com without obvious infection source, or secondary, arising as a complication of infection elsewhere (such as pneumonia, urinary tract, skin and soft tissues, intra-abdominal, devicerelated, etc) [10]. BSI is a common complication in liver cirrhosis [11]. It was reported that the frequencies of BSI in cirrhosis patients were 10 times higher than that in the non-cirrhosis group [12]. Two major reasons may lead to the high occurrence of BSI in cirrhosis patients: dysregulated intestinal bacterial translocation and cirrhosis associated immune dysfunction (CAID) [13]. The abnormal intestinal barrier permeability, overgrowth of small intestinal bacterial, and immune dysfunction may promote the bacteria into the bloodstream, leading to BSI [14,15]. BSI is significantly associated with prolonged hospital stay, faster development of liver disease, and low survival rate [7,16]. A populationbased cohort study indicated that BSI was a predictor for mortality in liver cirrhosis patients [17] The most common pathogens for BSI include gram-negative entericbacilli, anaerobes, and Enterococcus spp. [13]. In the previous studies, several studies reported the clinical characteristics of BSI patients. However, few studies described the specific characteristics of Escherichia coli BSI patients.
Timely and appropriately empirical antibiotic treatments are important for survival among BSI patients. Growing evidences have indicated that inappropriate antibiotic regimens are associated with increased mortality [18,19]. Until now, the empirical antibiotics are still effective for infection treatments. However, its failure rate is increasing due to the growing prevalence of multidrugresistant multidrug resistant (MDR) bacteria. MDR is defined as acquired non-susceptibility to at least one agent in three or more antimicrobial categories [20]. There are several common MDR, including extended-spectrum β-lactamase-producing bacteria (ESBL), methicillinresistant Staphylococcus aureus (MRSA), etc [21]. ESBL productions is one of the major reasons for antibiotic resistance in Escherichia coli. Therefore, to investigate the association between ESBL status and clinical characteristics in liver cirrhosis patients presenting Escherichia coli BSI will be helpful to guide empirical therapy.
In the present study, we aimed to investigate the clinical characteristics and antibiotic management, as well as independent risk factors for Escherichia coli BSI in liver cirrhosis patients. The liver disease of the patients were evaluated by Child-Pugh and model for end-stage liver disease (MELD) scores. We investigated the clinical characteristics of patients according to ESBL status of pathogens and survival status within 30 days after BSI diagnosis as well as sites of acquisition infection.

Baseline characteristics of the study population
There were 288 eligible patients in the present study, with average age of 51.95 ± 11.71 years. The study population included 204 males and 84 females and their mean times for hospital admission was 2.76 ± 3.29. Among the recruited patients, 213 (73.96%), 29 (10.07%), 32 (11.11%) and 14 (4.86%) patients were diagnosed with viral cirrhosis, alcoholic cirrhosis, autoimmune cirrhosis cases, and cryptogenic cirrhosis, respectively. The majority of the patients were classed as Child-Pugh class B (98, 34.03%) and C (171, 59.37%), while only few of them were diagnosed with class A (19, 6.60%). 45 of the patients were admitted to intensive care unit (ICU) ward (15.63%) ( Table 1).
In addition, 60.07% (173/288) of the study subjects were diagnosed with hospital-acquired BSI.

Clinical effects of hospital-acquired BSI in cirrhosis patients with BSI
Among the collected patients, the prevalence of hospital acquired BSI was 60.07%. We evaluated the effects of sites of acquisition infection on clinical characteristics of the study population. Analysis results indicated that cirrhosis patients with hospital-acquired BSI had a longer hospital stay than those without hospital acquired BSI (P = 0.000). Moreover, the MELD score at the discharged was also significantly associated with hospital acquired BSI (P = 0.000) ( Table 2).

Effects of ESBLs status on clinical characteristics based on study group
The patients were dived into two groups according to ESBL status of their isolates: ESBL + (n = 139) and ESBL − (n = 149). The prevalence of ESBL producing bacteria was 48.26%. We compared the clinical characteristics between the two groups. Analysis results indicated that ICU (P = 0.018) and MELD at onset of BSI (P = 0.035) were significantly associated with status of ESBL. In addition, we also found that liver cirrhosis patients infected by ESBL positive bacteria showed low sensitivity to firstly Oncotarget 35782 www.oncotarget.com selected antibiotics (P = 0.017), moreover, ESBL infection was significantly correlated with inappropriate empiric therapy within 48 h (P = 0.049). Other factors were similar between the two groups (P > 0.05 for all) ( Table 3).
Based on their resistance to antibiotics, 27.5% of the isolated bacteria were confirmed as MDR pathogens. We investigated the association between MDR and clinical characteristics. The results suggested that ESBL status (P = 0.000), ICU admission (P = 0.005), susceptibility to the firstly selected antibiotic (P = 0.025), appropriate empiric therapy within 48h (P = 0.011), and duration of infection (P = 0.021) were significantly correlated with MDR. However, there were no statistic correlation between MDR and other characteristics (P > 0.05 for all) ( Table 3).
Timely and appropriate antibiotic treatments were extremely important for survival in liver cirrhosis patients developing to BSI. Survival curve demonstrated that patients received appropriate empiric therapy within 48 h after BSI diagnosis had a significantly higher survival rate than those treated with inappropriate antibiotics (log rank test, P = 0.001) (Figure 1).

DISCUSSION
Liver cirrhosis is a major healthy problems worldwide, and the critical complications resulted from cirrhosis are the leading reasons for ICU admission among the patients [22]. Liver cirrhosis can cause multiple organs failure, including liver, heart, lung, kidney, as well as immune systems [23]. Due to the functional disorder of the organs and immune system, patients with cirrhosis are susceptible to infections. BSI is a common complication for cirrhosis patients. In the present study, we evaluated the clinical characteristics and risk factors for mortality in liver cirrhosis patients with Escherichia coli BSI.
288 cirrhosis patients diagnosed with Escherichia coli BSI were collected in the present study. Among them, the occurrence rate of hospital-acquired BSI was 60.07%. We compared the clinical characteristics between hospital acquired BSI and non-hospital acquired BSI. The results suggested that patients with hospital-acquired BSI had a prolonged hospital stay. Related studies had indicated that hospital-acquired BSI could prolong hospital stay [24]. In addition, several studies indicated that hospital-acquired BSI was significantly associated with high mortality [25,26]. However, no significant correlation between the sites of acquisition infection and mortality was found in the present study. The differences might be caused by the different medical conditions and pathogens. The effects of the sites of acquisition infection on clinical symptoms were needed to be well investigated in the next study.
ESBL production is one of the main cause for antimicrobial resistance in Escherichia coli, which is significantly correlated with therapeutic failures and high mortality [27,28]. Unfortunately, the prevalence rate of ESBL positive bacteria is increasing. In the present study, 48.26% pathogens were confirmed as ESBL positive Escherichia coli. Drug sensitivity analysis demonstrated that these ESBL positive pathogen had a higher resistant rate to most of the common antibiotics, such as third-generation cephalosporin, quinolone. The liver cirrhosis patients infected by ESBL positive bacteria showed low sensitivity to the firstly selected antibiotics, and were more likely to undergo inappropriate empiric therapy within 48 h. Moreover, based on the drug sensitivity analysis, 27.5% pathogens were confirmed as MDR bacteria. Analysis results suggested that ESBL status, ICU admission, susceptibility to firstly selected antibiotic and appropriate empiric therapy within 48 h www.oncotarget.com were significantly different between MDR infection and non-MDR infection. The result might reveal that patients infected by MDR pathogens were more likely to undergo antibiotic treatments failure and poor survival. It was worthy noting that two of the isolated bacteria were confirmed as CRE in the present study. Currently, CRE became common with high mortality. Moreover, unlike other antibiotic resistance, carapenems resistance was a complex process and mediated by several mechanisms, leading to limited therapeutic effects based on the present antibiotics [29,30]. Related studies reported that combination antibiotic treatments could significantly reduce the mortality in patients infected with CRE [31]. The optimal antibiotic treatment regimens for CRE were needed to be investigated in further studies.
Child-Pugh score and MELD score are based on the specific characteristics of cirrhosis. These scoring standards were simple and widely applied to predict longand short-time mortality in liver cirrhosis patients, even among those combined with infection [32]. However, accumulating evidences demonstrated that besides of the general scores, the specific parameters about infection severity should also be used to predict outcomes in liver cirrhosis patients combined with BSI [33,34]. In the current study, Cox regression model was used to evaluate the prognostic value of clinical parameters. The results suggested that empiric therapy within 48 h, ICU admission, HE and final MELD score were independent biomarkers for prognosis in liver cirrhosis patients presenting Escherichia coli BSI. In the previous studies, several factors were confirmed as prognostic markers for liver cirrhosis patients combined with BSI, including MELD score, SBP as BSI source, appropriate antibiotic treatments, HE, and so on [35,36]. All these indicators could help identify the high-risk individuals as soon as possible, thereby the therapeutic regimens were adjusted to improve clinical outcomes. It was noted that cirrhosis patients who were admitted to ICU, with high MELD or presented HE had a complex disease process and several factors might contribute to their mortality besides of infection. Therefore, novel prognostic systems combined with general and specific characteristics were needed to predict prognosis in liver cirrhosis patients combined with Escherichia coli BSI.
In conclusion, ESBL production is an important influencing factor for antibiotic resistance in Escherichia  coli and patients carrying ESBL producing pathogens are more likely admitted to ICU ward, with worse MELD score. In addition, ICU admission, HE, MELD score and empiric therapy within 48h are independent predictors for outcomes in cirrhosis patients with Escherichia coli BSI. The present study may be useful for therapy of Escherichia coli BSI in patients with liver cirrhosis.

Study subjects
The present study was a retrospective and observational study. The liver cirrhosis patients developing to Escherichia coli BSI during the period of hospitalization in Beijing 302 Hospital from May 2009 to September 2014 were collected in the study. The collected patients should meet the following criterion: (1) The study subjects were adult group; (2) All of the patients were diagnosed with decompensated liver cirrhosis; (3) All of the patients were infected with single Escherichia coli. And no other pathogens were isolated from the infection specimens; (4) The clinical characteristics of the patients were available. For cases who developed multiple BSIs during the study period, only the first BSI episode was included in the analysis.

Data collection
The collected patients were pathologically diagnosed with liver cirrhosis and presented BSI during the study period. The demographics characteristics including gender and age, hospitalization information, pathogenesis and severity degree of the disease according to Child-Pugh and MELD scores [37,38], complications during the hospital stay, baseline characteristics of BSI, and outcomes at discharged were recorded in the present study. In addition, the results of drug sensitivity test and empirical antibiotic regimen were also recorded. 30-day mortality was used to measure the primary outcomes of