Advertisement

A prospective multicentre study of the epidemiology and outcomes of bloodstream infection in cirrhotic patients

Open ArchivePublished:August 14, 2017DOI:https://doi.org/10.1016/j.cmi.2017.08.001

      Abstract

      Objectives

      To describe the current epidemiology of bloodstream infection (BSI) in patients with cirrhosis; and to analyse predictors of 30-day mortality and risk factors for antibiotic resistance.

      Methods

      Cirrhotic patients developing a BSI episode were prospectively included at 19 centres in five countries from September 2014 to December 2015. The discrimination of mortality risk scores for 30-day mortality were compared by area under the receiver operator risk and Cox regression models. Risk factors for multidrug-resistant organisms (MDRO) were assessed with a logistic regression model.

      Results

      We enrolled 312 patients. Gram-negative bacteria, Gram-positive bacteria and Candida spp. were the cause of BSI episodes in 53%, 47% and 7% of cases, respectively. The 30-day mortality rate was 25% and was best predicted by the Sequential Organ Failure Assessment (SOFA) and Chronic Liver Failure–SOFA (CLIF-SOFA) score. In a Cox regression model, delayed (>24 hours) antibiotic treatment (hazard ratio (HR) 7.58; 95% confidence interval (CI) 3.29–18.67; p < 0.001), inadequate empirical therapy (HR 3.14; 95% CI 1.93–5.12; p < 0.001) and CLIF-SOFA score (HR 1.35; 95% CI 1.28–1.43; p < 0.001) were independently associated with 30-day mortality. Independent risk factors for MDRO (31% of BSIs) were previous antimicrobial exposure (odds ratio (OR) 2.91; 95% CI 1.73–4.88; p < 0.001) and previous invasive procedures (OR 2.51; 95% CI 1.48–4.24; p 0.001), whereas spontaneous bacterial peritonitis as BSI source was associated with a lower odds of MDRO (OR 0.30; 95% CI 0.12–0.73; p 0.008).

      Conclusions

      MDRO account for nearly one-third of BSI in cirrhotic patients, often resulting in delayed or inadequate empirical antimicrobial therapy and increased mortality rates. Our data suggest that improved prevention and treatment strategies for MDRO are urgently needed in the liver cirrhosis patients.

      Keywords

      Introduction

      Infections are one of the most important complications of liver cirrhosis [
      • Jalan R.
      • Fernandez J.
      • Wiest R.
      • Schnabl B.
      • Moreau R.
      • Angeli P.
      • et al.
      Bacterial infections in cirrhosis: a position statement based on the EASL special conference, 2013.
      ,
      • Leber B.
      • Spindelboeck W.
      • Stadlbauer V.
      Infectious complications of acute and chronic liver disease.
      ,
      • Arvaniti V.
      • D'Amico G.
      • Fede G.
      • Manousou P.
      • Tsochatzis E.
      • Pleguezuelo M.
      • et al.
      Infections in patients with cirrhosis increase mortality four-fold and should be used in determining prognosis.
      ]. Bloodstream infections (BSI) occur in 4% to 21% of patients with cirrhosis, with a ten-fold higher frequency than in non-cirrhotic population and an estimated incidence of 11.7 cases per 10 000 patient-days [
      • Bartoletti M.
      • Giannella M.
      • Caraceni P.
      • Domenicali M.
      • Ambretti S.
      • Tedeschi S.
      • et al.
      Epidemiology and outcomes of bloodstream infection in patients with cirrhosis.
      ]. The mortality rates of BSI in patients with liver cirrhosis range between 23% and 58% [
      • Hsieh C.C.
      • Lee C.C.
      • Chan T.Y.
      • Hong M.Y.
      • Chi C.H.
      • Ko W.C.
      Clinical features and impact of empirical therapy in cirrhotic adults with community-onset bacteremia.
      ,
      • Campillo B.
      • Richardet J.P.
      • Kheo T.
      • Dupeyron C.
      Nosocomial spontaneous bacterial peritonitis and bacteremia in cirrhotic patients: impact of isolate type on prognosis and characteristics of infection.
      ].
      The epidemiology and outcomes of BSI in cirrhotic patients have been described mostly by retrospective single-centre studies [
      • Bartoletti M.
      • Giannella M.
      • Lewis R.E.
      • Viale P.
      Bloodstream infections in patients with liver cirrhosis.
      ,
      • Bang J.H.
      • Jung Y.
      • Cheon S.
      • Kim C.J.
      • Song K.H.
      • Choe P.G.
      • et al.
      Pseudomonas aeruginosa bacteremia in patients with liver cirrhosis: a comparison with bacteremia caused by enterobacteriaceae.
      ,
      • Park H.
      • Jang K.J.
      • Jang W.
      • Park S.H.
      • Park J.Y.
      • Jeon T.J.
      • et al.
      Appropriate empirical antibiotic use and 30-d mortality in cirrhotic patients with bacteremia.
      ]. Recent studies have suggested that the major challenges in the management of BSI in patients with liver cirrhosis are the lack of validated criteria in this population to predict poor outcome, as classical sepsis criteria have been shown to be inaccurate in cirrhotic patients; and limited data concerning the contemporary prevalence of multidrug-resistant organisms (MDRO) and their impact on treatment outcomes [
      • Jalan R.
      • Fernandez J.
      • Wiest R.
      • Schnabl B.
      • Moreau R.
      • Angeli P.
      • et al.
      Bacterial infections in cirrhosis: a position statement based on the EASL special conference, 2013.
      ,
      • Fernández J.
      • Bert F.
      • Nicolas-Chanoine M.H.
      The challenges of multi-drug-resistance in hepatology. The challenges of multi-drug-resistance in hepatology.
      ].
      To address these knowledge gaps, the aim of this study was to describe the contemporary epidemiology and outcomes of BSI in a large prospective multicentre cohort of patients with liver cirrhosis to identify the best clinical prediction score for 30-day mortality at the time of BSI onset in patients with liver cirrhosis; and to identify which pathogens are most frequently associated with inadequate empirical treatment. We also explored risk factors for BSI with MDRO in patients with liver cirrhosis.

      Methods

       Study design

      We performed a prospective multicentre observational cohort study with the endorsement of European Study Group of Bloodstream Infections and Sepsis (ESGBIS). The ESGBIS members who agreed to participate were asked to report all consecutive patients with liver cirrhosis who developed a BSI from September 2014 to December 2015.
      The study was approved by all local institutional review boards at participating hospitals. Written informed consent was obtained from patients or from legal surrogates before enrolment.

       Population

      All adult (>18 years) patients with liver cirrhosis who developed BSI at the participating centres were included in the study. The diagnosis of liver cirrhosis was based on previous liver biopsy results or a composite of clinical signs and findings provided by laboratory test results, endoscopy and radiologic imaging [
      • Moreau R.
      • Jalan R.
      • Gines P.
      • Pavesi M.
      • Angeli P.
      • Cordoba J.
      • et al.
      Acute-on-chronic liver failure is a distinct syndrome that develops in patients with acute decompensation of cirrhosis.
      ].
      BSI was defined the growth of a noncommon skin contaminant from one or more blood cultures (BC) and of a common skin contaminant such as diphtheroids, Bacillus species, Propionibacterium species, coagulase-negative staphylococci or micrococci from two or more BCs drawn from separate sites and reporting the same antimicrobial susceptibility test profile.
      Patients with previous liver transplants were excluded. Patients with subsequent episodes of BSI with an interval between BSIs of less than 3 months were excluded. Patients were followed up to 30 days after BSI onset, defined by the first positive BC.

       Patient recruitment and management

      In each study centre, the screening of enrollable patients was conducted hospitalwide. The centre coordinator was an infectious disease specialist (n = 13) or a hepatologist (n = 6). In both cases, the centre coordinator was appointed to screen for enrollable patients by both revising positive BCs and collaborating with the local liver units.
      To ensure the best enrolment rate, participating centres were contacted monthly and participated in investigator meetings. Patients were managed according with international or local guidelines; therapeutic management was not dictated by study protocol; and infectious disease consultation was available on demand in all centres.

       Data collection

      We collected data using an electronic case report form available at the study website. The integrity of data was systematically checked by an investigator before being entered into the database by a monthly assessment of data completeness and consistency. In case of inconsistent or missing data, queries were generated and distributed to the participating site's investigators for reconciliation. The following variables were collected at the moment of enrolment: demographic variables (sex, age); cause and severity of liver disease according to the Model for End Stage Liver Disease (MELD) collected at baseline, and BSI-onset; presence of hepatocellular carcinoma; presence of other comorbidities according with the Charlson score [
      • Charlson M.E.
      • Pompei P.
      • Ales K.L.
      • MacKenzie C.R.
      A new method of classifying prognostic comorbidity in longitudinal studies: development and validation.
      ]; invasive procedure performed within 30 days before BSI onset were collected (gastrointestinal endoscopy including esophagogastroduodenoscopy, colonoscopy, endoscopic retrograde cholangiopancreatography, endoscopic ultrasound); transjugular intrahepatic portosystemic shunt insertion; biliary procedures including biliary percutaneous drainage and/or stenting; hepatocellular carcinoma treatments; date, ward and cause of hospitalization; epidemiologic classification of BSI; and infection severity according to Bone's [
      • Bone R.C.
      Sepsis, the sepsis syndrome, multi-organ failure: a plea for comparable definitions.
      ] score, Sequential Organ Failure Assessment (SOFA) or Chronic Liver Failure–SOFA (CLIF-SOFA) (Supplementary Table S1), including the presence of acute-on-chronic liver failure (ACLF), as described by Moreau et al. [
      • Moreau R.
      • Jalan R.
      • Gines P.
      • Pavesi M.
      • Angeli P.
      • Cordoba J.
      • et al.
      Acute-on-chronic liver failure is a distinct syndrome that develops in patients with acute decompensation of cirrhosis.
      ]. We also collected data on the pathogens isolated and their susceptibility patterns; antibiotics administered as empirical therapy; and antibiotics administered as definitive therapy. Outcome variables were collected after 30 days from BSI onset during either bedside evaluation, outpatient visit or telephone call. These included the need of intensive care unit admission, length of hospital stay and 30-day transplant-free mortality.

       Microbiology

      Before study onset, the use of standard diagnostic methods was required and agreed with all the participating centres. This included the use of an automated BC detector system, the performance of Gram stain and/or rapid test (such as matrix-assisted desorption ionization–time of flight mass spectrometry or fluorescence in situ hybridization using peptide nucleic acid probes) with immediate communication of the preliminary information to the attending physicians and the use of an automated system (VITEK n = 17, MicroScan n = 2) for susceptibility testing. Breakpoints, screening and conformation of the main mechanisms of resistance were done according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) guidelines (http://www.amcli.it/wp-content/uploads/2015/10/EUCAST_detection_resistance_mechanisms_V1.pdf). Isolates considered as MDRO included the following: extended-spectrum β-lactamase-producing Enterobacteriaceae, carbapenem-resistant Enterobacteriaceae, multidrug resistant nonfermenting bacilli, methicillin-resistant Staphylococcus aureus, methicillin-resistant coagulase-negative staphylococci and Enterococcus faecium.

       Statistical analysis

      Categorical variables are presented as absolute numbers and their relative frequencies, and were compared by the chi-square test or Fisher exact test when appropriate. Quantitative variables are presented as mean and standard deviation if normally distributed, or as median and interquartile range if nonnormally distributed. Nonnormally distributed continuous variables were compared by the Mann-Whitney U test, and normally distributed continuous variables were compared by the t test.

       Predictors of mortality

      The discrimination of six established mortality risk scores for all-cause 30-day mortality were analysed by the area under the receiver operator curve (aROC). Risk scores with the highest aROC were then evaluated in a multivariate Cox regression model to identify the risk score with the best overall fit (Akaike information criterion, AIC) and discrimination (Harrell's C statistic) for 30-day all-cause mortality. Next, factors associated (p < 0.1) with 30-day mortality in the univariate analysis were entered stepwise into a Cox regression model with the risk score that was previously identified to have the best discrimination for 30-day mortality. All variables were explored for interaction and collinearity. The impact of infection-related or treatment variables was assessed at low, medium and high levels of baseline mortality risk, as defined by classification and regression tree (CART) analysis. Martingale residuals from the Cox model were used to calculate chi-square values; all score cut points were set at p < 0.05.

       Correlation between pathogens and adequacy of empirical antimicrobial treatment

      The frequency and strength of association between specific pathogen resistance profiles and the inadequate antimicrobial therapy were analysed by Spearman's rank correlation coefficient and were plotted as a weighted bubble plot by pathogen prevalence.

       Analysis of risk factors for MDRO

      To assess the independent risk factors for MDRO isolation, all variables with p ≤ 0.10 at univariate analysis were included in a forward conditional stepwise multivariate logistic regression model. The validity of the final model was assessed by estimating goodness of fit to the data with the Hosmer-Lemeshow test.
      All analysis was performed by Stata IC 13 software (StataCorp, College Station, TX, USA).

      Results

       Patient recruitment and baseline characteristics

      Nineteen of the 25 invited centres agreed to participate. Participating hospitals were all tertiary-care teaching facilities from Italy (n = 10 centres), Spain (n = 5 centres), Germany (n = 2 centres), Croatia (n = 1 centre) and Israel (n = 1 centre).
      A total of 312 patients were analysed (Fig. 1). Overall, 204 (65%) of 312 were male, and mean ± standard deviation age was 61 ± 12 years. Alcohol abuse and hepatitis C virus infection were the cause of cirrhosis in 40% and 36% of cases, respectively. The most common reasons for hospital admission were suspected bacterial infection (42%), acute decompensation unrelated to bacterial infection (40%) or scheduled procedures (6%) (Table 1).
      Table 1Difference of underlying conditions, BSI data and therapeutic management between survivors and nonsurvivors within 30 days from BSI onset
      CharacteristicTotal, n = 312 (100%)Nonsurvivors, n = 79 (25%)Survivors n = 233 (75%)p
      Demographic data
       Age years, mean (± standard deviation)61 ± 1260 ± 1062 ± 120.26
       Male sex204 (65)52 (65)152 (65)0.92
      Liver disease
      Thirty patients had multiple underlying liver disease, mostly hepatitis C virus and alcohol-related cirrhosis (n = 19), hepatitis B virus and alcohol-related cirrhosis (n = 5) and other causes (n = 6).
       Hepatitis C112 (36)27 (34)85 (36)0.71
       Alcoholic125 (40)32 (41)93 (40)0.96
       Cryptogenic47 (15)15 (15)32 (14)0.25
       Nonalcoholic fatty liver disease24 (8)6 (7)18 (8)0.96
       Hepatitis B26 (8)3 (4)23 (9)0.09
       Hepatocellular carcinoma50 (16)11 (14)39 (16)0.72
       Baseline MELD,
      Defined as most recent MELD within 2 weeks before BC drawn in patient free from symptoms of infection and/or acute decompensation; available in 231 patients.
      median (IQR)
      15 (11–20)18 (14–21)14 (10–19)0.01
      Admission diagnosis
       Ascitic decompensation44/304 (14)13 (18)31 (13)0.48
       Hepatorenal syndrome14/304 (4)7 (9)7 (3)0.05
       Hepatic encephalopathy29/304 (9)13 (18)16 (7)0.01
       Suspected bacterial infection131/304 (42)20 (27)111 (48)0.01
      Comorbidities
       Charlson index, mean (± standard deviation)7 ± 37 ± 37 ± 30.81
      BSI data
       Site of infection acquisition
      Community-acquired BSI60 (19)12 (15)48 (21)0.29
      Hospital-acquired BSI170 (54)50 (63)120 (51)0.06
      Healthcare associated82 (26)17 (21)65 (28)0.30
      Primary99 (32)30 (38)70 (30)0.19
      SBP50 (16)18 (23)32 (13)0.05
      Urinary tract35 (11)5 (6)30 (13)0.09
       Infection severity
      ACLF113 (36)59 (74)54 (23)<0.001
      CLIF-SOFA score, median (IQR)6 (4–9)10 (7–15)5 (3–7)<0.001
       MELD at BSI, median (IQR)18 (12–24)26 (21–31)16 (12–20)<0.001
       ΔMELD (at BSI—baseline), median (IQR)2 (0–5)5 (2–10)1 (0–4)<0.001
      Severe sepsis45 (14)16 (20)29 (12)0.10
      Septic shock41 (13)27 (35)14 (6)<0.001
       MDRO isolation98 (31)35 (44)63 (27)00.004
      Therapeutic management
       Adequate empirical antibiotic treatment190 (61)31 (39)159 (68)<0.001
       <6 hours from index BC153 (49)20 (25)133 (57)<0.001
       Between 6 and 24 hours from index BC24 (8)4 (5)20 (9)0.31
       >24 hours from index BC13 (4)7 (9)6 (3)0.01
       Inadequate122 (39)48 (61)74 (32)<0.001
       ICU admission84 (27)39 (50)45 (19)<0.001
      Data are presented as n (%) unless otherwise indicated.
      ACLF, acute-on-chronic liver failure; BC, blood culture; BSI, bloodstream infection; CLIF-SOFA, Chronic Liver Failure–Sequential Organ Failure Assessment; ICU, intensive care unit; IQR, interquartile range; MDRO, multidrug-resistant organisms; MELD, Model for End Stage Liver Disease; SBP, spontaneous bacterial peritonitis.
      a Thirty patients had multiple underlying liver disease, mostly hepatitis C virus and alcohol-related cirrhosis (n = 19), hepatitis B virus and alcohol-related cirrhosis (n = 5) and other causes (n = 6).
      b Defined as most recent MELD within 2 weeks before BC drawn in patient free from symptoms of infection and/or acute decompensation; available in 231 patients.
      BSIs were classified as primary in 99 (32%) of 312; catheter related in 32 (10%) of 312; and secondary in 181 (58%) of 312. Secondary BSIs included: intra-abdominal sources (99/181, 32%), mostly represented by spontaneous bacterial peritonitis (SBP; 50/99 50%) and cholangitis (26/99, 26%); urinary tract (35, 11%); lower respiratory tract (19, 6%); and other (34, 11%) (Supplementary Fig. S1).

       Thirty-day mortality

      At the end of 30-day follow-up, 79 (25%) of 312 patients died. According with BSI source, SBP (18/50, 36%) and pneumonia (6/19, 31%) were associated with the highest mortality rates, followed by primary BSI (30/99, 29%). Demographic and clinical characteristics of surviving vs. nonsurviving patients is shown in in Table 1. Among the different risk scores calculated at the day of BSI onset, CLIF-SOFA (aROC 0.82; 95% confidence interval (CI) 0.78–0.86), SOFA (aROC 0.82; 95% CI 0.77–0.86), MELD (aROC 0.79; 95% CI 0.74–0.83) and ΔMELD (aROC 0.76; 95% CI 0.70–0.84) showed roughly equivalent discriminative performance for 30-day all-cause mortality (Fig. 2(A)).
      Fig. 2
      Fig. 2Relationship of mortality risk models and 30-day all-cause mortality. (a) Comparison of risk score discrimination for 30-day mortality. (b) CART-defined cut points for CLIF-SOFA score analysis for failure time data. Martingale residuals of Cox model were used to calculate chi-square values for all possible cut points on all CART covariates at p < 0.05. (c) Cox regression survival estimates of patients receiving adequate empirical antimicrobial therapy within 24 hours, stratified by CLIF-SOFA score. (d) Cox regression survival estimates of patients who did not receive adequate empirical antimicrobial therapy within 24 hours, stratified by CLIF-SOFA score. ACLF, acute-on-chronic liver failure; CART, classification and regression tree; CLIF, Chronic Liver Failure; F, numbers of failures; MELD, Model for End Stage Liver Disease; N, numbers of patients; RHR, relative hazard ratio; SOFA Sequential Organ Failure Assessment.
      In the Cox regression model, CLIF-SOFA was associated with the lowest AIC score (768.07) and Harrell's C statistic (0.80), followed by BSI-MELD (AIC 798; Harrell's C 0.76). Therefore, risk factors for poor survival identified in univariate analysis (p < 0.1) were entered into multivariate Cox regression model that included CLIF-SOFA score as a continuous variable. Severity of sepsis displayed a collinearity with CLIF-SOFA scores and was not retained in the final model. Delayed (>24 hours) empirical antibiotic treatment (hazard ratio (HR) 7.58; 95% CI 3.29–18.67; p < 0.0001), inadequate empirical treatment (HR 3.14; 95% CI 1.93–5.12; p < 0.0001) and CLIF-SOFA (HR 1.35; 95% CI 1.28–1.43; p < 0.0001) were independently associated with increased rates of mortality within 30 days of a positive BC. After stratification for CLIF-SOFA using CART analysis (Fig. 2(B)), the impact of adequate empirical therapy administered within and after 24 hours was evident in patients with low-, medium- and high-baseline risk for mortality (Fig. 2(C), (D)).

       Aaetiology and risk factors for multidrug-resistant pathogens

      Overall, a total of 337 isolates were identified from the 312 BSI episodes. The aetiology of BSI according to epidemiologic classification and source of infection is summarized in Table 2 and Supplementary Table S2.
      Table 2Aetiology of 312 episodes of BSI in patients with liver cirrhosis
      CharacteristicTotal (n = 312 episodes, 100%), n (%)Community-acquired BSI (n = 60, 19%), n (%)Healthcare-associated BSI (n = 82, 26%), n (%)Hospital-acquired BSI (n = 170, 54%), n (%)p
      Gram positive146 (47)26 (43)36 (44)84 (49)0.59
       Coagulase negative staphylococci25 (8)2 (3)6 (7)17 (10)0.25
      Staphylococcus aureus (MSSA)41 (13)7 (12)9 (11)25 (15)0.25
      S. aureus (MRSA)12 (4)1 (2)4 (5)7 (4)0.59
      Streptococcus spp.24 (8)13 (22)4 (5)7 (4)<0.001*#
      Enterococcus faecalis15 (5)1 (1)5 (6)9 (5)0.45
      Enterococcus faecium22 (7)1 (1)6 (7)15 (9)0.33
       Other Gram positive
      Enterococcus raffinosus (n = 4), Listeria monocytogenes (n = 3), Corynebacterium striatum (n = 1).
      8 (3)1 (1)2 (2)5 (3)0.56
      Gram negative164 (53)35 (58)48 (58)81 (48)0.16
      Enterobacteriaceae136 (44)32 (53)39 (48)65 (38)0.09
      Escherichia coli82 (26)26 (43)27 (33)29 (17)<0.001*§
      Klebsiella pneumoniae29 (9)3 (5)8 (10)18 (12)0.43
      Enterobacter spp.13 (4)0 (0)2 (2)11 (6)0.06
       Other Enterobacteriaceae
      Morganella morganii (n = 4), Citrobacter spp. (n = 3), Klebsiella oxytoca (n = 3), Roultella planticola (n = 2), Proteus mirabilis (n = 1), Pantoea agglomerans (n = 1).
      15 (5)3 (5)3 (4)9 (5)0.48
       FQR Enterobacteriaceae52 (17)5 (8)21 (26)26 (15)0.01#
       ESBL-producing Enterobacteriaceae45 (14)2 (3)13 (16)30 (18)0.02*
       Carbapenem-resistant Enterobacteriaceae9 (3)0 (0)1 (1)8 (5)0.10
       Nonfermenters28 (9)3 (10)9 (11)16 (10)0.44
      Pseudomonas aeruginosa16 (5)2 (3)6 (7)8 (5)
      Acinetobacter baumannii3 (1)0 (0)0 (0)3 (2)
      Stenotrophomonas maltophilia4 (1)0 (0)1 (1)3 (2)
      Other nonfermenters
      Acinetobacter lwoffii (n = 1), Aeromonas veronii (n = 1), Pseudomonas alcaligenes (n = 1), Pseudomonas stutzeri (n = 1), Pseudomonas oryzihabitans (n = 1).
      5 (2)1 (1)2 (2)2 (3)
      Other Gram-negative
      Moraxella catarrhalis (n = 1), Haemophilus influenzae (n = 1).
      2 (1)1 (2)1 (1)0 (0)
      Anaerobes6 (2)3 (2)1 (1)2 (3)
      Fungi
      Candida spp.21 (7)1 (2)2 (2)19 (11)0.76
       Mixed infections30 (10)3 (5)5 (6)22 (13)0.09
      BSI, bloodstream infection; CR, carbapenem resistant; ESBL, extended-spectrum β-lactamase; FQR, fluoroquinolone resistant; MRSA methicillin-resistant Staphylococcus aureus; MSSA methicillin-susceptible S. aureus.
      *p < 0.01 between community acquired and hospital-acquired BSI.
      #p < 0.01 between community-acquired and healthcare-associated BSI.
      §p <0.01 between healthcare-associated and hospital acquired BSI.
      a Enterococcus raffinosus (n = 4), Listeria monocytogenes (n = 3), Corynebacterium striatum (n = 1).
      b Morganella morganii (n = 4), Citrobacter spp. (n = 3), Klebsiella oxytoca (n = 3), Roultella planticola (n = 2), Proteus mirabilis (n = 1), Pantoea agglomerans (n = 1).
      c Acinetobacter lwoffii (n = 1), Aeromonas veronii (n = 1), Pseudomonas alcaligenes (n = 1), Pseudomonas stutzeri (n = 1), Pseudomonas oryzihabitans (n = 1).
      d Moraxella catarrhalis (n = 1), Haemophilus influenzae (n = 1).
      Pathogens associated to the highest mortality rate were carbapenem-resistant Enterobacteriaceae (4/9, 44%), Candida spp. (9/21, 43%), E. faecium (9/22, 41%), extended-spectrum β-lactamase-producing Enterobacteriaceae (16/45, 36%), Streptococcus pneumoniae (3/9, 33%) and methicillin-susceptible Staphylococcus aureus (13/41, 32%).
      Significant differences were found in the aetiology distribution of BSI reported from different countries (Supplementary Table S3).
      Overall, 98 (31%) BSIs were caused by MDRO. As shown in Fig. 3, identification of MDRO or Candida spp. was strongly associated with receipt of inappropriate antimicrobial or antifungal treatment in the first 24 hours.
      Fig. 3
      Fig. 3Correlation of isolated bloodstream pathogen with probability of inappropriate antibiotic therapy. Spearman rank coefficient and associated p values were calculated for most common isolated pathogens including (a) Gram-negative bacilli and (b) Gram-positive cocci plus Candida spp. Size of bubble is relative to number of isolates. CNS, coagulase-negative staphylococci; CR, carbapenem resistant; ESBL, extended-spectrum β-lactamase; FQR, fluoroquinolone resistant; MRSA methicillin-resistant Staphylococcus aureus; MSSA, methicillin-susceptible S. aureus; VRE, vancomycin-resistant enterococci.
      Finally, we compared patients with and without MDRO bacteria to identify possible risk factors for MDRO. In multivariate analysis adjusted for clinical severity (CLIF-SOFA) and length of in-hospital stay before the onset of BSI, we found that prior (<30 days) antimicrobial exposure (OR (odds ratio) 2.91; 95% CI 1.73–4.88; p < 0.001), and prior (<30 days) invasive procedures (OR 2.51; 95% CI 1.48–4.24; p 0.001) were associated with a higher odds of BSI caused by MDRO. SBP source of BSI was associated with a lower odds of BSI caused by MDRO (OR 0.30; 95% CI 0.12–0.73; p 0.008) (Supplementary Table S4). These independent risk factors for MDRO were confirmed when analysis was performed by country taking into account the different prevalence of MDRO in Italy (37.5%), Germany (26.3%), Spain (22.4%) and Israel (30.6%) (Supplementary Table S5).

      Discussion

      The main finding of this study was that MDRO account for nearly one-third of BSI in patients with liver cirrhosis and are frequently associated with delays in effective treatment or inadequate empirical therapy, which are independent risk factors for death after a positive BC in both low- and high-risk patients. Among commonly used mortality risk scores, we found that CLIF-SOFA and SOFA best discriminated nonsurviving from surviving cirrhotic patients. Finally, we found that previous antimicrobial exposure, invasive procedures and source of infection play a role in determining the presence of MDRO.
      Similar to earlier studies, we found that alcohol abuse and hepatitis C virus infection are still the main causes of underlying disease [
      • Fernández J.
      • Acevedo J.
      • Castro M.
      • Garcia O.
      • de Lope C.R.
      • Roca D.
      • et al.
      Prevalence and risk factors of infections by multiresistant bacteria in cirrhosis: a prospective study.
      ]. Most BSI episodes are acquired while in hospital or after frequent exposures to the healthcare environment [
      • Merli M.
      • Lucidi C.
      • Di Gregorio V.
      • Falcone M.
      • Giannelli V.
      • Lattanzi B.
      The spread of multi drug resistant infections is leading to an increase in the empirical antibiotic treatment failure in cirrhosis: a prospective survey.
      ].
      The mortality rate of BSI was 25%, which appears to be significantly higher than that associated with BSI in the general population and with other bacterial infections in patients with liver cirrhosis [
      • Fernández J.
      • Acevedo J.
      • Castro M.
      • Garcia O.
      • de Lope C.R.
      • Roca D.
      • et al.
      Prevalence and risk factors of infections by multiresistant bacteria in cirrhosis: a prospective study.
      ,
      • Weinstein M.P.
      • Towns M.L.
      • Quartey S.M.
      • Mirrett S.
      • Reimer L.G.
      • Parmigiani G.
      • et al.
      The clinical significance of positive blood cultures in the 1990s: a prospective comprehensive evaluation of the microbiology, epidemiology, and outcome of bacteremia and fungemia in adults.
      ,
      • Rodríguez-Baño J.
      • López-Prieto M.D.
      • Portillo M.M.
      • Retamar P.
      • Natera C.
      • Nuño E.
      • et al.
      Epidemiology and clinical features of community-acquired, healthcare-associated and nosocomial bloodstream infections in tertiary-care and community hospitals.
      ]. However, we found that common sepsis criteria used to discriminate infection severity were less predictive of outcomes in patients with liver cirrhosis, as previously described [
      • Kaukonen K.M.
      • Bailey M.
      • Pilcher D.
      • Cooper D.J.
      • Bellomo R.
      Systemic inflammatory response syndrome criteria in defining severe sepsis.
      ]. Severity assessments based on parameters of organ failure may be more accurate in predicting outcome in cirrhotic patients with BSI [
      • Shankar-Hari M.
      • Phillips G.S.
      • Levy M.L.
      • Seymour C.W.
      • Liu V.X.
      • Deutschman C.S.
      Developing a new definition and assessing new clinical criteria for septic shock: for the Third International Consensus Definitions for Sepsis and Septic Shock (SEPSIS-3).
      ]. The CLIF-SOFA and SOFA scores better predicted mortality status after both 7 and 30 days after positive blood cultures than the MELD score. However, the superiority of CLIF-SOFA over SOFA was due to a modestly better fit in the Cox regression model. In our opinion, SOFA and CLIF-SOFA can both be used in the common clinical practice; CLIF-SOFA might be preferable in non-intensive care unit patients. Further studies are needed to address this latter point.
      The other major finding of this study was that besides underlying condition and infection severity, timely and appropriate antimicrobial therapy has a major impact on the outcome of BSI, confirming prior results [
      • Bartoletti M.
      • Giannella M.
      • Caraceni P.
      • Domenicali M.
      • Ambretti S.
      • Tedeschi S.
      • et al.
      Epidemiology and outcomes of bloodstream infection in patients with cirrhosis.
      ,
      • Merli M.
      • Lucidi C.
      • Di Gregorio V.
      • Lattanzi B.
      • Giannelli V.
      • Giusto M.
      An empirical broad spectrum antibiotic therapy in health-care-associated infections improves survival in patients with cirrhosis: a randomized trial.
      ]. The main predictor of inappropriate therapy in our patient cohort was isolation of a MDRO or Candida spp.
      We observed a substantial rate of MDRO in several countries, especially Italy, confirming previous single-centre studies [
      • Bartoletti M.
      • Giannella M.
      • Lewis R.E.
      • Viale P.
      Bloodstream infections in patients with liver cirrhosis.
      ,
      • Fernández J.
      • Bert F.
      • Nicolas-Chanoine M.H.
      The challenges of multi-drug-resistance in hepatology. The challenges of multi-drug-resistance in hepatology.
      ,
      • Merli M.
      • Lucidi C.
      • Di Gregorio V.
      • Falcone M.
      • Giannelli V.
      • Lattanzi B.
      The spread of multi drug resistant infections is leading to an increase in the empirical antibiotic treatment failure in cirrhosis: a prospective survey.
      ].
      The problem of MDRO and the related ineffectiveness of empirical treatment is an increasingly important topic of importance in the management of liver cirrhosis. In our study, risk factors independently associated to MDRO were antimicrobial exposure or undergoing invasive procedures in the previous 30 days of infection onset. By contrast, having a SBP as a source of infection was associated with lower risk of MDRO. This latter factor may seem unexpected. However, in the study of Fernandez et al. [
      • Fernández J.
      • Acevedo J.
      • Castro M.
      • Garcia O.
      • de Lope C.R.
      • Roca D.
      • et al.
      Prevalence and risk factors of infections by multiresistant bacteria in cirrhosis: a prospective study.
      ], among all infection caused by MDRO, only 9% were represented by SBP. This finding requires further studies for confirmation.
      Our study has some limitations including the heterogeneity of data due to the different epidemiology and different practice patterns over the different centres. This latter, however, is in line with the exploratory and observational design of the study. Also, the number of enrolled patients seems lower than expected. Possibly this was related to the heterogeneity of enrolling centres, which included both hepatologic units and infectious disease units.
      We think that BSIs in the cirrhotic population should be managed by a multidisciplinary team that includes a hepatologist and an infectious disease specialist performing systematic consultation. Given the global heterogeneity of distribution of MDRO, all centres should adopt local guidelines for empirical antimicrobial treatment of infection, especially healthcare-related and hospital-acquired infections in cirrhotic patients.
      In conclusion, we believe that this study could give substantial, generalizable information on the epidemiology of BSI in liver cirrhosis and provide the basis for further interventional studies on the management of BSI in this setting.

      Acknowledgements

      We thank the members of the ESGBIS/BICHROME study group.
      ESGBIS/BICHROME Study Group members: C. Campoli, G. Siccardi (Infectious Diseases Unit, Department of Medical and Surgical Sciences, Sant’Orsola-Malpighi Hospital, University of Bologna, Bologna, Italy); S. Ambretti (Operative Unite of Microbiology, Sant’Orsola-Malpighi Hospital, University of Bologna, Bologna, Italy); A. Stallmach (Department of Internal Medicine IV, Jena University Hospital, Friedrich Schiller University, Jena, Germany); M. Venditti (Department of Public Health and Infectious Diseases, ‘Sapienza’ University of Rome, Rome, Italy); C. Lucidi (Division of Gastroenterology, Department of Clinical Medicine, ‘Sapienza’ University of Rome, Rome, Italy); S. Ludovisi (Department of Infectious Diseases, Fondazione IRCCS Policlinico San Matteo, Pavia, Italy); M. De Cueto, M. D. Navarro, E. Lopez Cortes (Unidad Clínica Intercentros de Enfermedades Infecciosas, Microbiología y Medicina Preventiva, Hospitales Universitarios Virgen Macarena y Virgen del Rocío-IBIS, Seville, Spain); E. Bouza, M. Valerio, A. Eworo (Clinical Microbiology and Infectious Diseases Department, Hospital General Universitario Gregorio Marañón, Madrid, Spain); R. Losito (Institute of Infectious Diseases, Policlinico Universitario Agostino Gemelli, Rome, Italy); M. Senzolo, E. Nadal (Multivisceral Transplant Unit, Department of Surgery, Oncology, and Gastroenterology, Padova University Hospital, Italy); A. Ottobrelli (Gastrohepatology Unit, AOU Città della Salute e della Scienza di Torino, University of Turin, Turin, Italy); M. Varguvic (Infectious Diseases, University Hospital for Infectious Diseases ‘Dr. Fran Mihaljevic,’ Zagreb, Croatia); C. Badia (Infectious Disease Unit, Service of Internal Medicine, Hospital Universitari Mútua de Terrassa, Terrassa, Barcelona, Spain); G. Borgia, I. Gentile, A.R. Buonomo (Department of Clinical Medicine and Surgery, Section of Infectious Diseases, University of Naples Federico II, Naples, Italy); E. Boumis (2nd Infectious Diseases Division, National Institute for Infectious Diseases ‘L. Spallanzani,’ Rome, Italy); A. Beteta-Lopez (Clinical Microbiology, Nuestra Senora del Prado Hospital); A. Rianda (Hepatology and Infectious Diseases Unit, National Institute for Infectious Diseases ‘L. Spallanzani,’ Rome, Italy); G. Taliani, S. Grieco (Infectious and Tropical Diseases Unit, Department of Clinical Medicine, Sapienza University of Rome, Rome, Italy).

      Transparency declaration

      All authors report no conflicts of interest relevant to this article.

      Appendix A. Supplementary data

      Figure thumbnail figs1

      References

        • Jalan R.
        • Fernandez J.
        • Wiest R.
        • Schnabl B.
        • Moreau R.
        • Angeli P.
        • et al.
        Bacterial infections in cirrhosis: a position statement based on the EASL special conference, 2013.
        J Hepatol. 2014; 60: 1310-1324
        • Leber B.
        • Spindelboeck W.
        • Stadlbauer V.
        Infectious complications of acute and chronic liver disease.
        Semin Respir Crit Care Med. 2012; 33: 80-95
        • Arvaniti V.
        • D'Amico G.
        • Fede G.
        • Manousou P.
        • Tsochatzis E.
        • Pleguezuelo M.
        • et al.
        Infections in patients with cirrhosis increase mortality four-fold and should be used in determining prognosis.
        Gastroenterology. 2010; 139 (e1241–5): 1246-1256
        • Bartoletti M.
        • Giannella M.
        • Caraceni P.
        • Domenicali M.
        • Ambretti S.
        • Tedeschi S.
        • et al.
        Epidemiology and outcomes of bloodstream infection in patients with cirrhosis.
        J Hepatol. 2014; 61: 51-58
        • Hsieh C.C.
        • Lee C.C.
        • Chan T.Y.
        • Hong M.Y.
        • Chi C.H.
        • Ko W.C.
        Clinical features and impact of empirical therapy in cirrhotic adults with community-onset bacteremia.
        Am J Emerg Med. 2015; 33: 222-228
        • Campillo B.
        • Richardet J.P.
        • Kheo T.
        • Dupeyron C.
        Nosocomial spontaneous bacterial peritonitis and bacteremia in cirrhotic patients: impact of isolate type on prognosis and characteristics of infection.
        Clin Infect Dis. 2002; 35: 1-10
        • Bartoletti M.
        • Giannella M.
        • Lewis R.E.
        • Viale P.
        Bloodstream infections in patients with liver cirrhosis.
        Virulence. 2016; 7: 309-319
        • Bang J.H.
        • Jung Y.
        • Cheon S.
        • Kim C.J.
        • Song K.H.
        • Choe P.G.
        • et al.
        Pseudomonas aeruginosa bacteremia in patients with liver cirrhosis: a comparison with bacteremia caused by enterobacteriaceae.
        BMC Infect Dis. 2013; 13: 332
        • Park H.
        • Jang K.J.
        • Jang W.
        • Park S.H.
        • Park J.Y.
        • Jeon T.J.
        • et al.
        Appropriate empirical antibiotic use and 30-d mortality in cirrhotic patients with bacteremia.
        World J Gastroenterol. 2015; 21: 3587-3592
        • Fernández J.
        • Bert F.
        • Nicolas-Chanoine M.H.
        The challenges of multi-drug-resistance in hepatology. The challenges of multi-drug-resistance in hepatology.
        J Hepatol. 2016; 65: 1043-1054
        • Moreau R.
        • Jalan R.
        • Gines P.
        • Pavesi M.
        • Angeli P.
        • Cordoba J.
        • et al.
        Acute-on-chronic liver failure is a distinct syndrome that develops in patients with acute decompensation of cirrhosis.
        Gastroenterology. 2013; 144 (e1421–9): 1426-1437
        • Charlson M.E.
        • Pompei P.
        • Ales K.L.
        • MacKenzie C.R.
        A new method of classifying prognostic comorbidity in longitudinal studies: development and validation.
        J Chronic Dis. 1987; 40: 373-383
        • Bone R.C.
        Sepsis, the sepsis syndrome, multi-organ failure: a plea for comparable definitions.
        Ann Intern Med. 1991; 114: 332-333
        • Fernández J.
        • Acevedo J.
        • Castro M.
        • Garcia O.
        • de Lope C.R.
        • Roca D.
        • et al.
        Prevalence and risk factors of infections by multiresistant bacteria in cirrhosis: a prospective study.
        Hepatology. 2012; 55: 1551-1561
        • Merli M.
        • Lucidi C.
        • Di Gregorio V.
        • Falcone M.
        • Giannelli V.
        • Lattanzi B.
        The spread of multi drug resistant infections is leading to an increase in the empirical antibiotic treatment failure in cirrhosis: a prospective survey.
        PLoS One. 2015; 10e0127448
        • Weinstein M.P.
        • Towns M.L.
        • Quartey S.M.
        • Mirrett S.
        • Reimer L.G.
        • Parmigiani G.
        • et al.
        The clinical significance of positive blood cultures in the 1990s: a prospective comprehensive evaluation of the microbiology, epidemiology, and outcome of bacteremia and fungemia in adults.
        Clin Infect Dis. 1997; 24: 584-602
        • Rodríguez-Baño J.
        • López-Prieto M.D.
        • Portillo M.M.
        • Retamar P.
        • Natera C.
        • Nuño E.
        • et al.
        Epidemiology and clinical features of community-acquired, healthcare-associated and nosocomial bloodstream infections in tertiary-care and community hospitals.
        Clin Microbiol Infect. 2010; 16: 1408-1413
        • Kaukonen K.M.
        • Bailey M.
        • Pilcher D.
        • Cooper D.J.
        • Bellomo R.
        Systemic inflammatory response syndrome criteria in defining severe sepsis.
        N Engl J Med. 2015; 372: 1629-1638
        • Shankar-Hari M.
        • Phillips G.S.
        • Levy M.L.
        • Seymour C.W.
        • Liu V.X.
        • Deutschman C.S.
        Developing a new definition and assessing new clinical criteria for septic shock: for the Third International Consensus Definitions for Sepsis and Septic Shock (SEPSIS-3).
        JAMA. 2016; 315: 775-787
        • Merli M.
        • Lucidi C.
        • Di Gregorio V.
        • Lattanzi B.
        • Giannelli V.
        • Giusto M.
        An empirical broad spectrum antibiotic therapy in health-care-associated infections improves survival in patients with cirrhosis: a randomized trial.
        Hepatology. 2016; 63: 1632-1639