- Email: [email protected]
Epidemiology and outcomes of bloodstream infection in patients with cirrhosis
Research Article
Epidemiology and outcomes of bloodstream infection in patients with cirrhosis Michele Bartoletti1, Maddalena Giannella1, Paolo Caraceni1, Marco Domenicali1, Simone Ambretti2, Sara Tedeschi1, Gabriella Verucchi1, Lorenzo Badia1, Russell E. Lewis1, Mauro Bernardi1, Pierluigi Viale1,⇑ 1
Department of Medical and Surgical Sciences, University of Bologna, Bologna, Italy; 2Microbiology Unit, S. Orsola-Malpighi Hospital Bologna, Italy
Background & Aims: Bloodstream infections (BSIs) in cirrhotic patients are 10-fold more common than in non-cirrhotic patients and increasingly caused by resistant pathogens. We examined 162 BSI episodes in cirrhotic patients to describe the etiology and risk factors for 30-day mortality. Methods: We retrospectively analyzed all consecutive BSIs in patients with liver cirrhosis at our 1350-bed teaching hospital (January 2008 to June 2012). Cox-proportional hazard regression was used to analyze the impact of disease and treatment-related variables on the crude 30-day mortality. Results: BSI episodes were identified in 162 patients, including 29 mixed infections. Most of episodes were classified as hospital acquired or healthcare associated (93%). Gram-negative bacteria (GNB), Gram-positive bacteria and Candida spp. caused 64%, 38%, and 10% of episodes, respectively. GNB were classified as multi-drug resistant (MDR) and extensively drug resistant (XDR) in 25% and 21% of cases, respectively. The overall crude 30-day mortality rate was 29%. Four risk factors were independently associated with 30-day crude mortality: worsening of MELD score from baseline (the last MELD score available in the 2 weeks prior BSI) to that at BSI onset (HR 1.11 per point increase, 95% CI 1.07–1.15, p <0.0001), spontaneous bacterial peritonitis as BSI source (HR 4.42, 2.04–9.54, p = 0.002), sepsis grading (HR 2.18, 1.39–3.43, p = 0.0007), and inappropriate antibiotic therapy within 24 h from blood cultures (HR 2.82, 1.50–5.41, p = 0.002). Conclusion: An increasing proportion of BSIs in cirrhotic patients are caused by resistant GNB and Candida spp. Accurate evaluation of risk factors for mortality may improve early appropriate therapeutic management. Ó 2014 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved.
Keywords: Bloodstream infection; Liver cirrhosis; Multi-drug resistance; Extremely-drug resistance; Candidemia; Inappropriate empirical therapy; Mortality. Received 26 September 2013; received in revised form 7 March 2014; accepted 17 March 2014 ⇑ Corresponding author. Address: Infectious Disease Unit, Policlinico S. OrsolaMalpighi, Via Massarenti, 11, 40138 Bologna, Italy. E-mail address: [email protected] (P. Viale).
Introduction Liver cirrhosis is the 10th most common cause of death in the Western world [1]. Among the complications of the End Stage Liver Disease (ESLD), infection represents the leading cause of acute decompensation [2,3] and is associated with a high mortality ranging from 12% to 52% [4,5]. Bloodstream infections (BSIs) are a common complication in patients with ESLD, affecting 4–21% of patients with cirrhosis [2], making them 10 times more common than in non-cirrhotic patients [6]. Alterations in intestinal permeability induced by portal hypertension, prolonged intestinal transit, and bacterial overgrowth may facilitate a high rate of bacterial translocation from the gut into the bloodstream [7,8]. Cirrhotic patients who develop a BSI are 2.4 to 6.3 more likely to die within 30-days compared to non-cirrhotic patients with similar infections [9]. In four epidemiological studies of BSI, liver cirrhosis was a consistent variable independently associated with mortality [10–13]. However, few studies have examined the specific risk factors for BSI-mortality in patients with ESLD. ESLD patients are also highly prone to infections with multidrug-resistant (MDR) microorganisms, given the frequency of hospital admissions and antibiotic prophylaxis administered for bacterial peritonitis [4]. The two largest studies on the epidemiology and outcome of BSI in cirrhotic patients, however, were published prior to widespread fluoroquinone and cephalosporin resistance, and currently emerging carbapenem resistance among Enterobacteriaceae. Cirrhosis may also be a unique risk factor for invasive candidiasis [14], although definitive studies establishing this link are lacking. The purpose this study was to: (i) describe the contemporary etiology of BSI in cirrhotic patients, and (ii) examine independent risk factors for all-cause patient mortality within 30-days of the positive blood cultures. Patients and methods Study design We performed a retrospective cohort study of unselected adult (P18 years) patients admitted to our hospital with diagnosis of cirrhosis who developed a
Journal of Hepatology 2014 vol. xxx j xxx–xxx
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021
Research Article bloodstream infection (BSI) from 1st January 2008 to 30th June 2012. Patients were selected by matching the laboratory data with the discharge diagnosis of liver cirrhosis. In those patients who presented multiple BSIs during the study period, only the first episode was considered for analysis. Setting S’ Orsola Malipighi Hospital in Bologna is a 1350-bed tertiary teaching institution in Northern Italy, with approximately 72,000 hospital admissions per year. The Institutional Ethics Committee approved the study. In our liver disease unit, antibiotic prophylaxis with norfloxacin is not routinely prescribed in patients with cirrhosis due to high rates of fluoroquinone resistance. Blood cultures (BCs) are drawn whenever an infection is clinically suspected. The policy for empirical antimicrobial therapy in cirrhotic patients with suspicion of infection has changed over the study period. Until the end of 2009, a 3rd generation cephalosporin was usually initiated with or without coverage for resistant Gram-positive bacteria depending on patient specific risk factors. After 2009, piperacillin/tazobactam was primarily used as first-line empiric therapy for cirrhotic patients without a history of colonization or infection by multidrug resistant bacteria, in whom a carbapenem is considered the drug of choice. Coverage for resistant Gram positive bacteria (Enterococcus faecium, coagulase negative staphylococci, methicillin resistant Staphylococcus aureus [MRSA]) with vancomycin is tailored on pre-defined risk factors (MRSA carrier, central venous catheter, prior exposure to prolonged antibiotic therapy, history of prior infection with MRSA). Antifungal coverage is usually not included in the empirical antimicrobial therapy. Data collection Data were extracted from medical charts of patients using a standardized data collection form. Variables included patient demographics (sex, age), date and ward of hospitalization, underlying diseases according with Charlson comorbidity index [15], the cause and severity of liver disease according with the baseline Child Pugh and Model for End-stage Liver Disease (MELD) scores [16,17]. ESLD complications were recorded if they occurred during the hospital stay before the onset of infection symptoms. Date of BCs, epidemiological classification of BSI, BSI source, Bone score for severity of sepsis [18], MELD scores at the day of BSI onset and at 24, 48, 72, and 96 h after BSI onset. Microbiology records were used to document final pathogen identification and susceptibility patterns. Data concerning patient outcome included all-cause mortality within 30-days of the positive BC.
Definitions Diagnosis of liver cirrhosis was established by histology or by clinical, analytical, and ultrasonographic findings [4]. Non-infectious complications of cirrhosis (ascites, hepatorenal syndrome, hepatic encephalopathy, and hepatocellular carcinoma) were defined in patients using criteria from the European Association for the Study of the Liver and International Ascites Club [19]. When possible, we also analyzed changes in the patient MELD score from baseline (defined as the most recent score available at least 2 weeks prior to BC) to BSI onset. The difference between the MELD score calculated on the day of positive BC and the baseline MELD was reported as ‘‘DMELD’’. Differences between MELD scores on 24, 48, and 72 h after BSI onset and the day of positive BCs were also calculated and were reported as DMELD-24 h, DMELD-48 h, and DMELD-72 h, respectively. Chronic renal failure was defined as kidney damage or glomerular filtration rate (GFR) <60 ml/min/1.73m2 for three months or more, irrespective of cause [20]. Chronic heart failure was defined according with the American College of Cardiology/American Heart Association staging system [21]. BSI was defined as the growth of a non-common skin contaminant from P1 BCs, and of a common skin contaminant (e.g., diphtheroids, Bacillus species, Propionibacterium species, coagulase negative staphylococci, or micrococci) from at P2 BCs drawn on separate sites. To distinguish between true BSIs and contamination, each positive BC was analyzed during review of the medical and microbiology records to confirm it represented true infection. The primary source of BSI was defined according to the Centers for Diseases Control and Prevention criteria [22]. Spontaneous bacterial peritonitis (SBP) was determined to be the source of BSI when the same microorganism was isolated from the BCs and peritoneal fluid, in presence of P250 polymorphonuclear cells/ml of fluid, and without other evident infection source. BSI was further classified as community acquired (CA), health-care associated (HCA), and hospital acquired (HA) according with the Friedman’s criteria [23].
2
Gram-negative microorganisms were classified as susceptible, multi drug resistant (MDR), extensively drug resistant (XDR) and pan drug resistant (PDR) according to the European Society of Clinical Microbiology and Infectious Diseases (ESCMID) and consensus definitions [24]. Briefly, MDR was defined as non-susceptibility to at least one agent in three or more antimicrobial categories, XDR was defined as non-susceptibility to at least one agent in all but two or fewer antimicrobial categories (i.e., bacterial isolates remain susceptible to only one or two categories), and PDR was defined as non-susceptibility to all agents in all antimicrobial categories. Empirical therapy was considered as adequate when at least one active antibiotic against the isolated pathogen, according with species identification and susceptibility test, was administered during first 24 h after blood cultures were drawn (before microbiological results were available). Definitive therapy was considered as the antibiotics administered beyond 72 h from the blood cultures, it was considered adequate according to microbiological results. Microbiology Identification and susceptibility test of the isolated pathogens were carried out using the Vitek 2 system. Antibiotic minimum inhibitory concentrations (MICs) were classified according with Clinical and Laboratory Standards Institute (CLSI) until December 2010, after this date according with EUCAST breakpoints. Results of antimicrobial susceptibility tests were generally available within 72–96 h after BCs becoming positive. Statistical analysis Categorical variables were analyzed as absolute numbers and their relative frequencies. Continuous variables were analyzed as mean and standard deviation (SD) if normally distributed, or as median and interquartile range (IQR) if nonnormally distributed. In the univariate analysis of risk factors for all-cause 30day mortality, categorical variables were compared using the v2 test, whereas continuous variables were compared using the Mann-Whitney U or two-tailed Student’s t test, when appropriate. The primary study objective was to describe the contemporary etiology of bloodstream infections in cirrhotic patients. A secondary exploratory objective was to determine if increasing MDR and XDR in Enterobacteriaceae would result in high rates of inadequate empirical antibiotic therapy in cirrhosis patients, which would be independently associated with increased 30-day crude mortality. To account for non-antibiotic therapy covariates impacting 30-day crude mortality, we first compared discrimination of MELD scores calculated on day of blood culture vs. DMELD score calculated prior to the positive blood culture, or at various intervals following the positive blood culture. The MELD score calculation with the highest area under the ROC curve (aROC) in relation to 30-day mortality was then entered into the regression model as a continuous variable, with a limited number (<5) clinically-plausible covariates identified in univariate analysis (p <0.1). The proportional hazard assumptions were tested for the final Cox models by including the interactions of all the predictors with log of survival time. Model goodness-of-fit was assessed by Naglekerke’s R-square measure and Akaike information criterion (AIC). Calibration of the model was assessed by Hosmer-Lemeshow test as described by Parzen and Lipsitz [25]. Discrimination of model was assessed by aROC analysis of final model survival probability estimates. All tests were two-sided with a significance level of 0.05. The analyses were performed using SYSTAT version 13.1 (Systat Software, Inc., Chicago, IL) and Medcalc version 12.7 (Medcalc, Ostend, Belgium).
Results During the study period, 8,874 patients with liver cirrhosis were admitted to our hospital with a total 137,749 days of hospital stay. BSI was diagnosed in 162 patients. Thus the incidence of BSI in patients with liver cirrhosis hospitalized at our hospital was estimated of 11.7 per 10,000 patient-days. All patients were hospitalized at the time of diagnosis, in medical (91%) surgical (5%) or ICU (4%) wards. Hepatitis C was the most common cause of liver disease in the study population (52%) followed by alcoholic liver disease (33%) and hepatitis B (15%). Few patients were classified as Child-Pugh class A (19%),
Journal of Hepatology 2014 vol. xxx j xxx–xxx
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021
JOURNAL OF HEPATOLOGY Table 1. Comparison of underlying conditions, BSI data and therapeutic management between survivors and non-survivors patients.
Total, n = 162 (100%) Demographic data Age (years) [mean (± SD)] 62 (±11) Male sex 104 (64) Hospitalization unit Medical 148 (91.3) ICU 8 (4.9) Liver diseasea Hepatitis C 84 (51) Hepatits B 24 (14) Alcoholic 53 (32) Hepatocellular carcinoma 57 (35) 16 (±6) Baseline MELD [mean (± SD)]b Complications during hospital stay before BSI onset Ascites Any episode 95 (58.6) Grade 3 or refractory ascites 43 (26) Hepato-renal syndrome 42 (25) Hepatic encephalopathy 49 (30) Charlson indexd 6 (±4) BSI data Days of hospital admission before BSI 17 (±21) [mean (± SD)] Site of infection acquisition Community-acquired BSI 12 (7.4) Healthcare-associated BSI 33 (20.4) Hospital-acquired BSI 117 (72.2) BSI source e Primary 116 (71.6) Lung 16 (9) Abdominal (SBP) 13 (8) Urinary tract 11 (6.8) Central venous catheter 6 (3.7) MELD at BSI [mean (± SD)] 20 (±8) ∆ MELD (at BSI-baseline) [mean (± SD)] 3 (±5) Bone score Severe sepsis 62 (38) Septic shock 10 (6) Therapeutic management Inadequate empiric therapy at 24 h 62 (38.2) 14 (8.7) Inadequate definitive therapy at 72 h Adequate source controlc 16 (9) ICU admission 20 (12)
Survivors, n = 115 (71%)
Non-survivors, n = 47 (29%)
p value
63 (±12) 75 (64)
63 (±10) 29 (61)
0.70 0.40
107 (93) 5 (4)
41 (87) 3 (6)
0.23 0.69
58 (50) 17 (15) 36 (31) 40 (36) 16 (±6)
26 (55) 7 (15) 17 (37) 17 (36) 18 (±6)
0.57 0.52 0.39 0.45 0.001
60 (52.2) 22 (19.6) 18 (16) 32 (28) 7 (±4)
35 (74.5) 21 (42) 24 (51) 17 (37) 6 (±3)
0.01 0.004 <0.0001 0.19 0.60
14 (±16)
27 (±27)
0.003
11 (10) 23 (20) 81 (70)
1 (2) 10 (21) 36 (77)
0.18 0.86 0.43
87 (75.7) 12 (10) 4 (3) 10 (8.7) 4 (3) 18 (±7) 2 (±3)
29 (61.7) 4 (9) 9 (19) 1 (2.1) 2 (0) 27 (±12) 9 (±6)
0.99 1.00 0.002 0.32 0.56 0.001 <0.001
33 (29) 2 (2)
29 (64) 8 (17)
0.0002 0.0009
39 (34) 10 (8.7) 13 (11) 8 (7)
23 (49) 4 (8.6) 3 (6) 12 (27)
0.07 0.97 0.34 0.003
a
Some patients had multiple underlying liver disease. Baseline MELD score was available in only 150 patients. c Includes central venous catheter (6) and urinary catheter removal (10). d Charlson Comorbidity Index or Comorbidity-Adjusted Life Expectancy is a score used to assess whether a patient will live long enough to benefit from a specific medical intervention. It consists of 16 comorbidity components (each with an assigned determined point) and a scoring age (1 point every 10 years >40 years) [15]. e Some patients had multiple sites. SD, standard deviation; MELD, model end stage liver disease; BSI, bloodstream infection; SBP, spontaneous bacteria peritonitis; ICU, intensive care unit. b
with the majority of patients classified as class B (38%) or class C (43%) chronic liver disease (Table 1). The majority of 162 cirrhotic patients were classified as having hospital-acquired BSI (67%), occurring on average 17 days (range 0–121 days) (median 11 days, IQR 3–24) after
admission (Table 1). Most patients (71.6%) developed BSI without a clearly identifiable source. In the remaining 21% of patients, lung (9%), SBP (8%), urinary tract (6.8%) or central venous catheter (CVC) (3.7%) was deemed as the source of BSI.
Journal of Hepatology 2014 vol. xxx j xxx–xxx
3
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021
Research Article Etiology A total of 202 isolates were identified in BCs in the 162 BSI episodes. Gram-negative bacteria (GNB) accounted for 104/162 BSIs (64%). Most GNB were Enterobacteriaceae (76%), with Escherichia coli (54%), Klebsiella pneumoniae (37%), and Enterobacter spp. (9%) most commonly isolated. Extended-spectrum b-lactamase (ESBL) and K. pneumoniae carbapenemase (KPC) production was identified in 30% and 9% of Enterobacteriaceae, respectively. Non-fermenting GNB (P. aeruginosa, A. baumannii, S. maltophilia) were isolated in 15% of BSI episodes. GNB BSI were classified as MDR and XDR in 25% and 21% of cases, respectively. No PDR GNB was identified during the study period. MDR or XDR GNB were isolated more frequently from patients hospitalized in the last two years of the study period (p = 0.006) and with hospitalacquired BSI (p = 0.008). The median days from the hospital admission to BSI were 16 (IQR 6–43) and 8 (2–17) among patients with MDR/XDR GNB BSI and those with non-MDR/XDR GNB BSI (p = 0.001), respectively. Gram-positive bacteria were cultured in 38% of BSI episodes. The most common isolated species were methicillin-susceptible Staphylococcus aureus (MSSA, 20%), Enterococcus faecalis (20%), methicillin-resistant S. aureus (MRSA, 13%), E. faecium (16%), and coagulase-negative staphylococci (CNS, 13%). Candida species accounted for 10% (16/162) of BSIs. Among the 16 episodes of candidemia (11 primary BSI, 3 BSI secondary to SBP, and 2 catheter related bloodstream infections [CR-BSI]), the majority were caused by C. albicans (n = 14), followed by C. parapsilosis (n = 1), and a fluconazole-resistant C. glabrata (n = 1). Of the 11 patients with primary BSI, 9 (81%) were exposed to prolonged (>5 days) broad spectrum antibiotic therapy within the 30 days before candidemia, 5 (45%) were receiving total parenteral nutrition, 1 was undergone hepatic resection for a HCC nodule in the prior 7 days, and 1 was admitted to ICU at the time of positive BCs. None of them had urine, CVC or other clinical samples positive for Candida spp. Candida BSI were more common in cirrhotic patients with more than 6 days hospital stay prior to developing BSI (p = 0.009), hospital-acquired BSI (p = 0.03), prior surgery (p = 0.013), central venous catheter (p = 0.004), neutropenia (p = 0.05), or prior piperacillin-tazobactam (p = 0.05) or fluoroquinolone (p = 0.013) antibiotic therapy. Mixed-infections (2 or more microorganisms) were found in 19% of all BSI episodes and were frequently associated with isolation of antibiotic-resistant pathogens, including CNS (p = 0.003), MRSA (p = 0.001), ESBL-producing E. coli (p = 0.001), KPC-producing Enterobacteriaceae (p = 0.005), A. baumanii (p = 0.003) and Candida spp. (p = 0.05). Considering isolate susceptibility, the frontline use of a third generation cephalosporin for empirical coverage of GNB was adequate in only 60% of BSI episodes (92% for CA, 82% for HCA and 51% for HA BSI, p = 0.002) (see Supplementary Table 1). Mortality risk factors Overall, 47 patients (29%) died within 30 days after BSI onset. The causes of mortality based on the clinician evaluation were: acute on chronic liver failure 78% (70% of which caused by BSI) and other 21%. Mortality rates according with the source of BSI were 69% for SBP, 38% for pneumonia, 33% for CR-BSI, 25% for primary BSI, and 9% for urinary infection. Among patients with candidemia 56% died within 30 days after blood cultures were drawn.
4
Multiple candidate risk factors for 30-day crude mortality were identified in univariate analysis (Table 1). Crude mortality rates varied considerably depending on the infectious etiology and antibiotic susceptibility (Fig. 1). Among the 6-most common isolated pathogens (n = 89), MSSA and antibiotic-susceptible E. coli BSI were associated with 30-day mortality rates of <20%, whereas ESBL-E. coli and E. faecalis were associated with crude mortality rates of 23–30%. The highest BSI crude mortality rates were found for Candida spp. (56%) and KPC-producing Enterobacteriaceae (60%). A detailed comparison of microbial infection etiology between surviving and non-surviving cirrhotic patients with BSI is presented in Table 2. Distribution of the MELD values calculated at baseline, BSI onset, and at 24, 48, 72, and 96 h after the BSI onset between survivors and non-survivors is shown in Fig. 2A. Receiver-operator curve analysis suggested that DMELD calculated from baseline to the time of positive BC had the highest discrimination for 30-day mortality status (aROC 0.89, 0.83–0.92) even though it was not statistically better than MELD score calculated at the time of drawing blood cultures (aROC 0.81, 0.74–0.87) (Fig. 2B). Five candidate mortality variables identified by univariate analysis (hepatorenal syndrome, SBP as BSI source, presentation with severe sepsis or septic shock, delayed appropriate antimicrobial therapy >24 h) plus DMELD were analyzed using Cox proportional hazards regression. Four covariates were retained in the final model derived from 150 patient cases; DMELD, SBP as BSI source, severe sepsis or septic shock, and receipt of inappropriate empirical within the first 24 h of BC (Table 3). The model displayed acceptable goodness of fit, correctly predicting the 30-day survival status of 85% of the cirrhotic patient cohort with a sensitivity of 90% and a specificity of 75% (Table 3). Use of MELD scores calculated at the time of BSI rather than DMELD resulted in less favorable model performance (AIC 383 vs. 355, respectively). Of the 105 patients with a primary bacterial BSI, 24 (22.9%) died within 30 days from BSI onset. In this group, the variables associated with 30-day mortality were: hepatorenal syndrome, presentation with severe sepsis or septic shock and higher DMELD (data not shown). Examination of whether isolation of a specific pathogen may be associated with significantly lower odds of receiving appropriate empirical therapy suggested that Candida BSI has the strongest association with inappropriate empirical therapy (data not shown). The specific impact of inappropriate vs. appropriate antibiotic therapy for BSI in cirrhotic patients is presented in Fig. 3. The percentages of patients receiving appropriate empirical therapy within 24 h from BCs were 54% and 63% between patients with DMELD >4 and those with DMELD 64, respectively (p = 0.17). Discussion To the best of our knowledge, this retrospective study represents the largest contemporary epidemiological survey of BSI in cirrhotic patients. Gram-negative bacteria were the leading pathogens with a high prevalence of MDR and XDR strains. In addition, 10% of BSIs were caused by Candida spp. Patients who received an inappropriate empirical antimicrobial therapy within 24 h after blood cultures were drawn had higher 30-day mortality rate. The majority of studies on infection in cirrhotic patients have analyzed patients with SBP. Few studies have examined the
Journal of Hepatology 2014 vol. xxx j xxx–xxx
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021
JOURNAL OF HEPATOLOGY 1.00
Survival probability
0.90 0.80 0.70
Susceptible, n = 98; AHR 0.58 (0.31-1.09); p = 0.1
0.60
ESBL and FQ resistant, n = 32; AHR 0.63 (0.30-1.31); p = 0.21
0.50
Carbapenem-resistant, n = 20; AHR 1.26 (0.60-2.64); p = 0.54
0.40 0.30
Candida spp., n = 12; AHR 2.09; (0.93-4.67); p = 0.07
0.20 0.10 0.00 0
5
10
15
20
25
30
Days after positive blood culture Fig. 1. Impact of isolation of antimicrobial resistant pathogens on survival from bloodstream infection in cirrhotic patients. Survival probability was estimated for antibiotic susceptible (solid line), ESBL and fluoroquinolone-resistant (long dash), carbapenem-resistant (short dash) and Candida spp. (dash-dot) using a multivariate Cox proportional hazard regression model accounting for underlying illness severity and timing of antibiotic administration. Symbols represent the unadjusted observed mortality rates for each pathogen subset estimated from Kaplan-Meier curves. Antibiotic susceptible (d), ESBL and fluoroquinolone resistant (), carbapenem-resistant (N) and Candida spp. (j). AHR, adjusted hazard ratio. Table 2. Microbiology etiology distribution between survivors and non-survivors groups per BSI episode.
Gram positive Coagulase negative staphylococci Staphylococcus aureus (MSSA) Staphylococcus aureus (MRSA) Streptococcus pneumoniae Enterococcus faecalis Enterococcus faecium Other Gram positivea Gram negative Enterobacteriaceae Escherichia coli Escherichia coli (FQR) Escherichia coli (ESBL) Enterobacter species Klebsiella pneumoniae Klebsiella pneumoniae (FQR) Klebsiella pneumoniae (ESBL) Klebsiella pneumoniae (KPC) Other enterobactericeaeb Non-fermenters Pseudomonas aeruginosa Acinetobacter baumanii Stenotrophomonas maltophilia Fungi Candida species Mixed infections
Total, n = 162 episodes (100%) 62 (38) 10 (6) 15 (9) 6 (3) 4 (2) 15 (9) 12 (7) 13 (8) 104 (64) 77 (47) 22 (13) 4 (2) 17 (10) 7 (4) 4 (2) 4 (2) 7 (4) 14 (8) 3 (1) 25 (15) 10 (6) 10 (6) 5 (3)
Survivors, n = 115 episodes (71%) 43 (37) 6 (5) 13 (11) 4 (3) 2 (2) 10 (9) 8 (7) 13 (11) 76 (66) 57 (50) 18 (16) 4 (3) 13 (11) 6 (5) 3 (3) 4 (3) 2 (2) 7 (6) 2 (2) 19 (17) 8 (7) 8 (7) 3 (3)
Non-survivors, n = 47 episodes (29%) 19 (40) 4 (9) 2 (4) 2 (4) 2 (4) 5 (11) 4 (9) 0 (0) 28 (60) 22 (47) 4 (9) 0 (0) 4 (9) 1 (2) 1 (2) 0 (0) 5 (11) 7 (15) 1 (2) 6 (13) 2 (4) 2 (4) 2 (4)
p value
0.75 0.42 0.16 0.06 0.35 0.70 0.73 0.01 0.47 0.86 0.23 0.65 0.60 0.18 0.86 0.20 0.01 0.07 0.99 0.64 0.73 0.73 0.63
16 (9) 30 (18)
7 (6) 22 (19)
9 (19) 8 (17)
0.01 0.83
MSSA, methicillin susceptible S. aureus; MRSA, methicillin-resistant S. aureus; FQR, fluoroquinolone-resistant; ESBL, extended-spectrum beta-lactamase; KPC, Klebsiella pneumonia carbapenemase. a Aeromonas spp. (n = 1), Enterococcus durans (n = 1), Listeria monocytogenes (n = 1), Micrococcus luetues (n = 1), beta-hemolytic streptococci (n = 9). b Serratia marcescens (n = 1), Citrobacter freundii (n = 2).
general epidemiology of BSI in cirrhotic patients. The two largest studies on incidence, etiological distribution and outcome of BSI in cirrhotic patients were published more than a decade ago and did not evaluate the specific impact of inadequate empirical therapy on outcome [6,26].
In the present study of 162 cirrhotic patients with BSI, GNB accounted for more than 60% of BSIs with a prevalence of MDR and XDR pathogens of 25% and 21% among Gram-negative isolates, respectively. Therefore, cephaloporins, which are recommended in current guidelines for infection in cirrhotic patients,
Journal of Hepatology 2014 vol. xxx j xxx–xxx
5
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021
Research Article A
35
Non-survivors day 30; n = 47 (29%)
B 100
80 Sensitivity
Median MELD ± 95% CI
Δ MELD BSI
30
25
20
15
MELD at BSI
Δ MELD 72 h Δ MELD 48 h
60
Δ MELD 24 h
40 Variable MELD BSI Δ MELD BSI Δ MELD 24 h Δ MELD 48 h Δ MELD 72 h
20 Survivors day 30; n = 115 (71%)
10
aROC (95% CI) 0.81 (0.74-0.87) 0.89 (0.83-0.94) 0.54 (0.45-0.62) 0.66 (0.56-0.76) 0.76 (0.66-0.86)
0 Baseline
BSI Diagnosis
24
48
72
96 h
0
20
40
60
80
100
100-Specificity
Time of assessment Fig. 2. MELD score as prognostic index for 30-day survival in cirrhotic patients with bloodstream infection. (A) Temporal patterns of median MELD score in 30-day non-survivors (open circles) and survivors (closed circles); (B) comparison of aROC for MELD score variables as a predictor of 30-day survival status.
Table 3. Final Cox-proportional hazards regression model for 30-day BSI mortality in cirrhotic patients.
Model covariate Spontaneous bacterial peritonitis Sepsis grading Inadequate empirical therapy within 24 h from BCs ΔMELD (per 1 point increase)
Hazard ratio 4.42 2.18 2.92 1.11
95% CI 2.04-9.54 1.39-3.43 1.50-5.41 1.07-1.15
p value 0.0002 0.0007 0.002 <0.0001
Model goodness of fit: Naglekerke’s R2 = 0.62; p <0.0001; Calibration: Hosmer-Lemeshow v2 = 8.34, degrees freedom = 8; p = 0.40). Discrimination: aROC 0.91 (95% CI 0.85– 0.95), p <0.0001.
1.00 Adequate antibiotic therapy within 24 hours (n = 100)
Survival probability
0.90 0.80 0.70 0.60 0.50 0.40
Inadequate antibiotic therapy within 24 hours (n = 62)
0.30 0.20
Adjusted HR 0.37 (0.20-0.70) p = 0.002
0.10 0.00 0
5
10 15 20 25 Days after positive blood culture
30
Fig. 3. Impact of inadequate antimicrobial therapy in the first 24 h on survival of cirrhotic patients with bloodstream infection. Probability of survival in patients receiving adequate therapy during first 24 h (solid line) or inadequate empirical therapy (dashed line) was estimated using multivariate Cox-hazard proportional regression accounting for severity of underlying illness. Actual unadjusted observed mortality rates were determined by Kaplan-Meier curves, and plotted for patients receiving adequate (j) or inadequate (h) antimicrobial therapy in the first 24 h.
would be inadequate as empirical therapy in over 40% of cases, mainly in hospital-acquired BSI. In addition, 10% of BSIs were caused by Candida spp., and, although data should be interpreted with caution due to the small number of events, the isolation of fungi was associated with the highest 30-day crude mortality rate (56%). Importantly, liver cirrhosis is a chronic underlying chronic condition that results in frequent contacts with the 6
healthcare system, especially in the advanced phase of the disease when the risk for infection complications is higher. The mean MELD at hospital admission of our patients was 20 ± 8 and most patients have had a recent contact with the hospital that surely influenced the etiology of infection. In contrast, among patients with true community acquired BSI, the most common pathogen was Streptococcus pneumoniae associated with pneumonia. DMELD, SBP as BSI source and severe sepsis or septic shock were independently associated with 30-day all-cause mortality in our study. Whereas, receipt appropriate empirical therapy was independently associated with better survival. Cirrhosis-specific scores, such as the Child Pugh and MELD scores, have been reported to be good predictors for short and long term mortality in patients with cirrhosis even in presence of infection [27,28]. Indeed, infection severity in patients with cirrhosis should be assessed using a liver disease calibrated scores rather than that used in general population [29]. Our analysis suggests that temporal trends in MELD score may be useful for characterizing infection-associated decompensation in patients with liver cirrhosis, and patients at highest risk for death. However, our definition of DMELD may be a source of bias because the time interval between the baseline MELD, defined as the last MELD available within two weeks before the BSI, and the MELD score calculated at the time of BC may have been variable between patients. Further studies will be required to examine the performance of this risk factor and whether patients with a higher DMELD would benefit from broader-spectrum empirical therapy that includes coverage of resistant GNB and Candida spp. To note
Journal of Hepatology 2014 vol. xxx j xxx–xxx
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021
JOURNAL OF HEPATOLOGY that all patients had at least criteria of systemic inflammatory response syndrome, this may partially be explained with the unique characteristics of the cirrhosis such as the hyper-dynamic circulatory status, low peripheral white blood cell count, and impaired synthesis of acute-phase proteins which can alter the accuracy of the classical symptoms and signs of sepsis. However, despite this bias, the sepsis grading seems to work also in this population, and therefore should be routinely included in the evaluation work-up of patients with suspicion of infection. Our study had some limitations. It is a monocentric retrospective study, thus epidemiological data may not be representative of other institutions. However, our hospital is a referral center for liver disease with many patients coming from different regions of the country. Additionally, the development of a more robust risk model for predicting mortality in cirrhotic patients with BSI was limited by the number of study subjects in our dataset. Nevertheless, our findings suggest a growing proportion of BSIs in cirrhotic patients are now caused by resistant GNB and Candida spp., raising concerns whether empirical antimicrobial regimens for BSI in cirrhotic patients should be reconsidered, especially in institutions with high (>10%) rates of ESBL and carbapenemase-producing enterobacteriaceae. According to our results, empirical antibiotic treatment should be guided first of all by infection severity (severe sepsis or septic shock), concomitant presence of SBP, site of infection acquisition (community, healthcare associated, hospital), and worsening of liver failure according to the MELD score at the time of drawing positive BCs. All these variables could be the first step toward the creation of a multifaceted clinical risk score in order to help clinicians in indentifying the highest risk subset of patients who would likely most benefit from broader-spectrum empirical therapy.
Conflict of interest The authors who have taken part in this study declared that they do not have anything to disclose regarding funding or conflict of interest with respect to this manuscript.
Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.jhep.2014.03. 021.
References [1] Leon DA, McCambridge J. Liver cirrhosis mortality rates in Britain from 1950 to 2002: an analysis of routine data. Lancet 2006;367:52–56. [2] Leber B, Spindelboeck W, Stadlbauer V. Infectious complications of acute and chronic liver disease. Semin Respir Crit Care Med 2012;33:80–95. [3] Moreau R, Jalan R, Gines P, Pavesi M, Angeli P, Cordoba J, et al. Acute-onchronic liver failure is a distinct syndrome that develops in patients with acute decompensation of cirrhosis. Gastroenterology 2013;144:e1429. [4] Fernandez J, Acevedo J, Castro M, Garcia O, de Lope CR, Roca D, et al. Prevalence and risk factors of infections by multiresistant bacteria in cirrhosis: a prospective study. Hepatology 2012;55:1551–1561. [5] Kang CI, Song JH, Chung DR, Peck KR, Yeom JS, Ki HK, et al. Liver cirrhosis as a risk factor for mortality in a national cohort of patients with bacteremia. J Infect 2011;63:336–343.
[6] Thulstrup AM, Sorensen HT, Schonheyder HC, Moller JK, Tage-Jensen U. Population-based study of the risk and short-term prognosis for bacteremia in patients with liver cirrhosis. Clin Infect Dis 2000;31:1357–1361. [7] Reiberger T, Ferlitsch A, Payer BA, Mandorfer M, Heinisch BB, Hayden H, et al. Non-selective betablocker therapy decreases intestinal permeability and serum levels of LBP and IL-6 in patients with cirrhosis. J Hepatol 2013;58:911–921. [8] Bellot P, Frances R, Such J. Pathological bacterial translocation in cirrhosis: pathophysiology, diagnosis and clinical implications. Liver Int 2013;33:31–39. [9] Linderoth G, Jepsen P, Schonheyder HC, Johnsen SP, Sorensen HT. Short-term prognosis of community-acquired bacteremia in patients with liver cirrhosis or alcoholism: a population-based cohort study. Alcohol Clin Exp Res 2006;30:636–641. [10] Ortega M, Almela M, Soriano A, Marco F, Martinez JA, Munoz A, et al. Bloodstream infections among human immunodeficiency virus-infected adult patients: epidemiology and risk factors for mortality. Eur J Clin Microbiol Infect Dis 2008;27:969–976. [11] Chen SY, Tsai CL, Lin CH, Lee CC, Chiang WC, Wang JL, et al. Impact of liver cirrhosis on mortality in patients with community-acquired bacteremia. Diagn Microbiol Infect Dis 2009;64:124–130. [12] Chang TY, Lee CH, Liu JW. Clinical characteristics and risk factors for fatality in patients with bloodstream infections caused by glucose non-fermenting gram-negative Bacilli. J Microbiol Immunol Infect 2010;43:233–239. [13] Al-Hasan MN, Lahr BD, Eckel-Passow JE, Baddour LM. Predictive scoring model of mortality in Gram-negative bloodstream infection. Clin Microbiol Infect 2013;19:948–954. [14] Grim SA, Berger K, Teng C, Gupta S, Layden JE, Janda WM, et al. Timing of susceptibility-based antifungal drug administration in patients with Candida bloodstream infection: correlation with outcomes. J Antimicrob Chemother 2012;67:707–714. [15] Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40:373–383. [16] Kamath PS, Wiesner RH, Malinchoc M, Kremers W, Therneau TM, Kosberg CL, et al. A model to predict survival in patients with end-stage liver disease. Hepatology 2001;33:464–470. [17] Pugh RN, Murray-Lyon IM, Dawson JL, Pietroni MC, Williams R. Transection of the oesophagus for bleeding oesophageal varices. Br J Surg 1973;60:646–649. [18] Bone RC, Balk RA, Cerra FB, Dellinger RP, Fein AM, Knaus WA, et al. Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine. Chest 1992;101:1644–1655. [19] EASL clinical practice guidelines on the management of ascites, spontaneous bacterial peritonitis, and hepatorenal syndrome in cirrhosis. J Hepatol 2010;53:397–417. [20] Levey AS, Coresh J, Balk E, Kausz AT, Levin A, Steffes MW, et al. National Kidney Foundation practice guidelines for chronic kidney disease: evaluation, classification, and stratification. Ann Intern Med 2003;139:137–147. [21] Hunt SA, Abraham WT, Chin MH, Feldman AM, Francis GS, Ganiats TG, et al. ACC/AHA 2005 Guideline Update for the Diagnosis and Management of Chronic Heart Failure in the Adult: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Writing Committee to Update the 2001 Guidelines for the Evaluation and Management of Heart Failure): developed in collaboration with the American College of Chest Physicians and the International Society for Heart and Lung Transplantation: endorsed by the Heart Rhythm Society. Circulation 2005;112:e154–e235. [22] Horan TC, Andrus M, Dudeck MA. CDC/NHSN surveillance definition of health care-associated infection and criteria for specific types of infections in the acute care setting. Am J Infect Control 2008;36:309–332. [23] Friedman ND, Kaye KS, Stout JE, McGarry SA, Trivette SL, Briggs JP, et al. Health care–associated bloodstream infections in adults: a reason to change the accepted definition of community-acquired infections. Ann Intern Med 2002;137:791–797. [24] Magiorakos AP, Srinivasan A, Carey RB, Carmeli Y, Falagas ME, Giske CG, et al. Multidrug-resistant, extensively drug-resistant and pandrug-resistant bacteria: an international expert proposal for interim standard definitions for acquired resistance. Clin Microbiol Infect 2012;18:268–281. [25] Parzen M, Lipsitz SR. A global goodness-of-fit statistic for Cox regression models. Biometrics 1999;55:580–584. [26] Campillo B, Dupeyron C, Richardet JP. Epidemiology of hospital-acquired infections in cirrhotic patients: effect of carriage of methicillin-resistant
Journal of Hepatology 2014 vol. xxx j xxx–xxx
7
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021
Research Article Staphylococcus aureus and influence of previous antibiotic therapy and norfloxacin prophylaxis. Epidemiol Infect 2001;127:443–450. [27] Thabut D, Massard J, Gangloff A, Carbonell N, Francoz C, Nguyen-Khac E, et al. Model for end-stage liver disease score and systemic inflammatory response are major prognostic factors in patients with cirrhosis and acute functional renal failure. Hepatology 2007;46:1872–1882.
8
[28] Durand F, Valla D. Assessment of prognosis of cirrhosis. Semin Liver Dis 2008;28:110–122. [29] Viasus D, Garcia-Vidal C, Castellote J, Adamuz J, Verdaguer R, Dorca J, et al. Community-acquired pneumonia in patients with liver cirrhosis: clinical features, outcomes, and usefulness of severity scores. Medicine (Baltimore) 2011;90:110–118.
Journal of Hepatology 2014 vol. xxx j xxx–xxx
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021
Epidemiology and outcomes of bloodstream infection in patients with cirrhosis Michele Bartoletti1, Maddalena Giannella1, Paolo Caraceni1, Marco Domenicali1, Simone Ambretti2, Sara Tedeschi1, Gabriella Verucchi1, Lorenzo Badia1, Russell E. Lewis1, Mauro Bernardi1, Pierluigi Viale1,⇑ 1
Department of Medical and Surgical Sciences, University of Bologna, Bologna, Italy; 2Microbiology Unit, S. Orsola-Malpighi Hospital Bologna, Italy
Background & Aims: Bloodstream infections (BSIs) in cirrhotic patients are 10-fold more common than in non-cirrhotic patients and increasingly caused by resistant pathogens. We examined 162 BSI episodes in cirrhotic patients to describe the etiology and risk factors for 30-day mortality. Methods: We retrospectively analyzed all consecutive BSIs in patients with liver cirrhosis at our 1350-bed teaching hospital (January 2008 to June 2012). Cox-proportional hazard regression was used to analyze the impact of disease and treatment-related variables on the crude 30-day mortality. Results: BSI episodes were identified in 162 patients, including 29 mixed infections. Most of episodes were classified as hospital acquired or healthcare associated (93%). Gram-negative bacteria (GNB), Gram-positive bacteria and Candida spp. caused 64%, 38%, and 10% of episodes, respectively. GNB were classified as multi-drug resistant (MDR) and extensively drug resistant (XDR) in 25% and 21% of cases, respectively. The overall crude 30-day mortality rate was 29%. Four risk factors were independently associated with 30-day crude mortality: worsening of MELD score from baseline (the last MELD score available in the 2 weeks prior BSI) to that at BSI onset (HR 1.11 per point increase, 95% CI 1.07–1.15, p <0.0001), spontaneous bacterial peritonitis as BSI source (HR 4.42, 2.04–9.54, p = 0.002), sepsis grading (HR 2.18, 1.39–3.43, p = 0.0007), and inappropriate antibiotic therapy within 24 h from blood cultures (HR 2.82, 1.50–5.41, p = 0.002). Conclusion: An increasing proportion of BSIs in cirrhotic patients are caused by resistant GNB and Candida spp. Accurate evaluation of risk factors for mortality may improve early appropriate therapeutic management. Ó 2014 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved.
Keywords: Bloodstream infection; Liver cirrhosis; Multi-drug resistance; Extremely-drug resistance; Candidemia; Inappropriate empirical therapy; Mortality. Received 26 September 2013; received in revised form 7 March 2014; accepted 17 March 2014 ⇑ Corresponding author. Address: Infectious Disease Unit, Policlinico S. OrsolaMalpighi, Via Massarenti, 11, 40138 Bologna, Italy. E-mail address: [email protected] (P. Viale).
Introduction Liver cirrhosis is the 10th most common cause of death in the Western world [1]. Among the complications of the End Stage Liver Disease (ESLD), infection represents the leading cause of acute decompensation [2,3] and is associated with a high mortality ranging from 12% to 52% [4,5]. Bloodstream infections (BSIs) are a common complication in patients with ESLD, affecting 4–21% of patients with cirrhosis [2], making them 10 times more common than in non-cirrhotic patients [6]. Alterations in intestinal permeability induced by portal hypertension, prolonged intestinal transit, and bacterial overgrowth may facilitate a high rate of bacterial translocation from the gut into the bloodstream [7,8]. Cirrhotic patients who develop a BSI are 2.4 to 6.3 more likely to die within 30-days compared to non-cirrhotic patients with similar infections [9]. In four epidemiological studies of BSI, liver cirrhosis was a consistent variable independently associated with mortality [10–13]. However, few studies have examined the specific risk factors for BSI-mortality in patients with ESLD. ESLD patients are also highly prone to infections with multidrug-resistant (MDR) microorganisms, given the frequency of hospital admissions and antibiotic prophylaxis administered for bacterial peritonitis [4]. The two largest studies on the epidemiology and outcome of BSI in cirrhotic patients, however, were published prior to widespread fluoroquinone and cephalosporin resistance, and currently emerging carbapenem resistance among Enterobacteriaceae. Cirrhosis may also be a unique risk factor for invasive candidiasis [14], although definitive studies establishing this link are lacking. The purpose this study was to: (i) describe the contemporary etiology of BSI in cirrhotic patients, and (ii) examine independent risk factors for all-cause patient mortality within 30-days of the positive blood cultures. Patients and methods Study design We performed a retrospective cohort study of unselected adult (P18 years) patients admitted to our hospital with diagnosis of cirrhosis who developed a
Journal of Hepatology 2014 vol. xxx j xxx–xxx
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021
Research Article bloodstream infection (BSI) from 1st January 2008 to 30th June 2012. Patients were selected by matching the laboratory data with the discharge diagnosis of liver cirrhosis. In those patients who presented multiple BSIs during the study period, only the first episode was considered for analysis. Setting S’ Orsola Malipighi Hospital in Bologna is a 1350-bed tertiary teaching institution in Northern Italy, with approximately 72,000 hospital admissions per year. The Institutional Ethics Committee approved the study. In our liver disease unit, antibiotic prophylaxis with norfloxacin is not routinely prescribed in patients with cirrhosis due to high rates of fluoroquinone resistance. Blood cultures (BCs) are drawn whenever an infection is clinically suspected. The policy for empirical antimicrobial therapy in cirrhotic patients with suspicion of infection has changed over the study period. Until the end of 2009, a 3rd generation cephalosporin was usually initiated with or without coverage for resistant Gram-positive bacteria depending on patient specific risk factors. After 2009, piperacillin/tazobactam was primarily used as first-line empiric therapy for cirrhotic patients without a history of colonization or infection by multidrug resistant bacteria, in whom a carbapenem is considered the drug of choice. Coverage for resistant Gram positive bacteria (Enterococcus faecium, coagulase negative staphylococci, methicillin resistant Staphylococcus aureus [MRSA]) with vancomycin is tailored on pre-defined risk factors (MRSA carrier, central venous catheter, prior exposure to prolonged antibiotic therapy, history of prior infection with MRSA). Antifungal coverage is usually not included in the empirical antimicrobial therapy. Data collection Data were extracted from medical charts of patients using a standardized data collection form. Variables included patient demographics (sex, age), date and ward of hospitalization, underlying diseases according with Charlson comorbidity index [15], the cause and severity of liver disease according with the baseline Child Pugh and Model for End-stage Liver Disease (MELD) scores [16,17]. ESLD complications were recorded if they occurred during the hospital stay before the onset of infection symptoms. Date of BCs, epidemiological classification of BSI, BSI source, Bone score for severity of sepsis [18], MELD scores at the day of BSI onset and at 24, 48, 72, and 96 h after BSI onset. Microbiology records were used to document final pathogen identification and susceptibility patterns. Data concerning patient outcome included all-cause mortality within 30-days of the positive BC.
Definitions Diagnosis of liver cirrhosis was established by histology or by clinical, analytical, and ultrasonographic findings [4]. Non-infectious complications of cirrhosis (ascites, hepatorenal syndrome, hepatic encephalopathy, and hepatocellular carcinoma) were defined in patients using criteria from the European Association for the Study of the Liver and International Ascites Club [19]. When possible, we also analyzed changes in the patient MELD score from baseline (defined as the most recent score available at least 2 weeks prior to BC) to BSI onset. The difference between the MELD score calculated on the day of positive BC and the baseline MELD was reported as ‘‘DMELD’’. Differences between MELD scores on 24, 48, and 72 h after BSI onset and the day of positive BCs were also calculated and were reported as DMELD-24 h, DMELD-48 h, and DMELD-72 h, respectively. Chronic renal failure was defined as kidney damage or glomerular filtration rate (GFR) <60 ml/min/1.73m2 for three months or more, irrespective of cause [20]. Chronic heart failure was defined according with the American College of Cardiology/American Heart Association staging system [21]. BSI was defined as the growth of a non-common skin contaminant from P1 BCs, and of a common skin contaminant (e.g., diphtheroids, Bacillus species, Propionibacterium species, coagulase negative staphylococci, or micrococci) from at P2 BCs drawn on separate sites. To distinguish between true BSIs and contamination, each positive BC was analyzed during review of the medical and microbiology records to confirm it represented true infection. The primary source of BSI was defined according to the Centers for Diseases Control and Prevention criteria [22]. Spontaneous bacterial peritonitis (SBP) was determined to be the source of BSI when the same microorganism was isolated from the BCs and peritoneal fluid, in presence of P250 polymorphonuclear cells/ml of fluid, and without other evident infection source. BSI was further classified as community acquired (CA), health-care associated (HCA), and hospital acquired (HA) according with the Friedman’s criteria [23].
2
Gram-negative microorganisms were classified as susceptible, multi drug resistant (MDR), extensively drug resistant (XDR) and pan drug resistant (PDR) according to the European Society of Clinical Microbiology and Infectious Diseases (ESCMID) and consensus definitions [24]. Briefly, MDR was defined as non-susceptibility to at least one agent in three or more antimicrobial categories, XDR was defined as non-susceptibility to at least one agent in all but two or fewer antimicrobial categories (i.e., bacterial isolates remain susceptible to only one or two categories), and PDR was defined as non-susceptibility to all agents in all antimicrobial categories. Empirical therapy was considered as adequate when at least one active antibiotic against the isolated pathogen, according with species identification and susceptibility test, was administered during first 24 h after blood cultures were drawn (before microbiological results were available). Definitive therapy was considered as the antibiotics administered beyond 72 h from the blood cultures, it was considered adequate according to microbiological results. Microbiology Identification and susceptibility test of the isolated pathogens were carried out using the Vitek 2 system. Antibiotic minimum inhibitory concentrations (MICs) were classified according with Clinical and Laboratory Standards Institute (CLSI) until December 2010, after this date according with EUCAST breakpoints. Results of antimicrobial susceptibility tests were generally available within 72–96 h after BCs becoming positive. Statistical analysis Categorical variables were analyzed as absolute numbers and their relative frequencies. Continuous variables were analyzed as mean and standard deviation (SD) if normally distributed, or as median and interquartile range (IQR) if nonnormally distributed. In the univariate analysis of risk factors for all-cause 30day mortality, categorical variables were compared using the v2 test, whereas continuous variables were compared using the Mann-Whitney U or two-tailed Student’s t test, when appropriate. The primary study objective was to describe the contemporary etiology of bloodstream infections in cirrhotic patients. A secondary exploratory objective was to determine if increasing MDR and XDR in Enterobacteriaceae would result in high rates of inadequate empirical antibiotic therapy in cirrhosis patients, which would be independently associated with increased 30-day crude mortality. To account for non-antibiotic therapy covariates impacting 30-day crude mortality, we first compared discrimination of MELD scores calculated on day of blood culture vs. DMELD score calculated prior to the positive blood culture, or at various intervals following the positive blood culture. The MELD score calculation with the highest area under the ROC curve (aROC) in relation to 30-day mortality was then entered into the regression model as a continuous variable, with a limited number (<5) clinically-plausible covariates identified in univariate analysis (p <0.1). The proportional hazard assumptions were tested for the final Cox models by including the interactions of all the predictors with log of survival time. Model goodness-of-fit was assessed by Naglekerke’s R-square measure and Akaike information criterion (AIC). Calibration of the model was assessed by Hosmer-Lemeshow test as described by Parzen and Lipsitz [25]. Discrimination of model was assessed by aROC analysis of final model survival probability estimates. All tests were two-sided with a significance level of 0.05. The analyses were performed using SYSTAT version 13.1 (Systat Software, Inc., Chicago, IL) and Medcalc version 12.7 (Medcalc, Ostend, Belgium).
Results During the study period, 8,874 patients with liver cirrhosis were admitted to our hospital with a total 137,749 days of hospital stay. BSI was diagnosed in 162 patients. Thus the incidence of BSI in patients with liver cirrhosis hospitalized at our hospital was estimated of 11.7 per 10,000 patient-days. All patients were hospitalized at the time of diagnosis, in medical (91%) surgical (5%) or ICU (4%) wards. Hepatitis C was the most common cause of liver disease in the study population (52%) followed by alcoholic liver disease (33%) and hepatitis B (15%). Few patients were classified as Child-Pugh class A (19%),
Journal of Hepatology 2014 vol. xxx j xxx–xxx
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021
JOURNAL OF HEPATOLOGY Table 1. Comparison of underlying conditions, BSI data and therapeutic management between survivors and non-survivors patients.
Total, n = 162 (100%) Demographic data Age (years) [mean (± SD)] 62 (±11) Male sex 104 (64) Hospitalization unit Medical 148 (91.3) ICU 8 (4.9) Liver diseasea Hepatitis C 84 (51) Hepatits B 24 (14) Alcoholic 53 (32) Hepatocellular carcinoma 57 (35) 16 (±6) Baseline MELD [mean (± SD)]b Complications during hospital stay before BSI onset Ascites Any episode 95 (58.6) Grade 3 or refractory ascites 43 (26) Hepato-renal syndrome 42 (25) Hepatic encephalopathy 49 (30) Charlson indexd 6 (±4) BSI data Days of hospital admission before BSI 17 (±21) [mean (± SD)] Site of infection acquisition Community-acquired BSI 12 (7.4) Healthcare-associated BSI 33 (20.4) Hospital-acquired BSI 117 (72.2) BSI source e Primary 116 (71.6) Lung 16 (9) Abdominal (SBP) 13 (8) Urinary tract 11 (6.8) Central venous catheter 6 (3.7) MELD at BSI [mean (± SD)] 20 (±8) ∆ MELD (at BSI-baseline) [mean (± SD)] 3 (±5) Bone score Severe sepsis 62 (38) Septic shock 10 (6) Therapeutic management Inadequate empiric therapy at 24 h 62 (38.2) 14 (8.7) Inadequate definitive therapy at 72 h Adequate source controlc 16 (9) ICU admission 20 (12)
Survivors, n = 115 (71%)
Non-survivors, n = 47 (29%)
p value
63 (±12) 75 (64)
63 (±10) 29 (61)
0.70 0.40
107 (93) 5 (4)
41 (87) 3 (6)
0.23 0.69
58 (50) 17 (15) 36 (31) 40 (36) 16 (±6)
26 (55) 7 (15) 17 (37) 17 (36) 18 (±6)
0.57 0.52 0.39 0.45 0.001
60 (52.2) 22 (19.6) 18 (16) 32 (28) 7 (±4)
35 (74.5) 21 (42) 24 (51) 17 (37) 6 (±3)
0.01 0.004 <0.0001 0.19 0.60
14 (±16)
27 (±27)
0.003
11 (10) 23 (20) 81 (70)
1 (2) 10 (21) 36 (77)
0.18 0.86 0.43
87 (75.7) 12 (10) 4 (3) 10 (8.7) 4 (3) 18 (±7) 2 (±3)
29 (61.7) 4 (9) 9 (19) 1 (2.1) 2 (0) 27 (±12) 9 (±6)
0.99 1.00 0.002 0.32 0.56 0.001 <0.001
33 (29) 2 (2)
29 (64) 8 (17)
0.0002 0.0009
39 (34) 10 (8.7) 13 (11) 8 (7)
23 (49) 4 (8.6) 3 (6) 12 (27)
0.07 0.97 0.34 0.003
a
Some patients had multiple underlying liver disease. Baseline MELD score was available in only 150 patients. c Includes central venous catheter (6) and urinary catheter removal (10). d Charlson Comorbidity Index or Comorbidity-Adjusted Life Expectancy is a score used to assess whether a patient will live long enough to benefit from a specific medical intervention. It consists of 16 comorbidity components (each with an assigned determined point) and a scoring age (1 point every 10 years >40 years) [15]. e Some patients had multiple sites. SD, standard deviation; MELD, model end stage liver disease; BSI, bloodstream infection; SBP, spontaneous bacteria peritonitis; ICU, intensive care unit. b
with the majority of patients classified as class B (38%) or class C (43%) chronic liver disease (Table 1). The majority of 162 cirrhotic patients were classified as having hospital-acquired BSI (67%), occurring on average 17 days (range 0–121 days) (median 11 days, IQR 3–24) after
admission (Table 1). Most patients (71.6%) developed BSI without a clearly identifiable source. In the remaining 21% of patients, lung (9%), SBP (8%), urinary tract (6.8%) or central venous catheter (CVC) (3.7%) was deemed as the source of BSI.
Journal of Hepatology 2014 vol. xxx j xxx–xxx
3
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021
Research Article Etiology A total of 202 isolates were identified in BCs in the 162 BSI episodes. Gram-negative bacteria (GNB) accounted for 104/162 BSIs (64%). Most GNB were Enterobacteriaceae (76%), with Escherichia coli (54%), Klebsiella pneumoniae (37%), and Enterobacter spp. (9%) most commonly isolated. Extended-spectrum b-lactamase (ESBL) and K. pneumoniae carbapenemase (KPC) production was identified in 30% and 9% of Enterobacteriaceae, respectively. Non-fermenting GNB (P. aeruginosa, A. baumannii, S. maltophilia) were isolated in 15% of BSI episodes. GNB BSI were classified as MDR and XDR in 25% and 21% of cases, respectively. No PDR GNB was identified during the study period. MDR or XDR GNB were isolated more frequently from patients hospitalized in the last two years of the study period (p = 0.006) and with hospitalacquired BSI (p = 0.008). The median days from the hospital admission to BSI were 16 (IQR 6–43) and 8 (2–17) among patients with MDR/XDR GNB BSI and those with non-MDR/XDR GNB BSI (p = 0.001), respectively. Gram-positive bacteria were cultured in 38% of BSI episodes. The most common isolated species were methicillin-susceptible Staphylococcus aureus (MSSA, 20%), Enterococcus faecalis (20%), methicillin-resistant S. aureus (MRSA, 13%), E. faecium (16%), and coagulase-negative staphylococci (CNS, 13%). Candida species accounted for 10% (16/162) of BSIs. Among the 16 episodes of candidemia (11 primary BSI, 3 BSI secondary to SBP, and 2 catheter related bloodstream infections [CR-BSI]), the majority were caused by C. albicans (n = 14), followed by C. parapsilosis (n = 1), and a fluconazole-resistant C. glabrata (n = 1). Of the 11 patients with primary BSI, 9 (81%) were exposed to prolonged (>5 days) broad spectrum antibiotic therapy within the 30 days before candidemia, 5 (45%) were receiving total parenteral nutrition, 1 was undergone hepatic resection for a HCC nodule in the prior 7 days, and 1 was admitted to ICU at the time of positive BCs. None of them had urine, CVC or other clinical samples positive for Candida spp. Candida BSI were more common in cirrhotic patients with more than 6 days hospital stay prior to developing BSI (p = 0.009), hospital-acquired BSI (p = 0.03), prior surgery (p = 0.013), central venous catheter (p = 0.004), neutropenia (p = 0.05), or prior piperacillin-tazobactam (p = 0.05) or fluoroquinolone (p = 0.013) antibiotic therapy. Mixed-infections (2 or more microorganisms) were found in 19% of all BSI episodes and were frequently associated with isolation of antibiotic-resistant pathogens, including CNS (p = 0.003), MRSA (p = 0.001), ESBL-producing E. coli (p = 0.001), KPC-producing Enterobacteriaceae (p = 0.005), A. baumanii (p = 0.003) and Candida spp. (p = 0.05). Considering isolate susceptibility, the frontline use of a third generation cephalosporin for empirical coverage of GNB was adequate in only 60% of BSI episodes (92% for CA, 82% for HCA and 51% for HA BSI, p = 0.002) (see Supplementary Table 1). Mortality risk factors Overall, 47 patients (29%) died within 30 days after BSI onset. The causes of mortality based on the clinician evaluation were: acute on chronic liver failure 78% (70% of which caused by BSI) and other 21%. Mortality rates according with the source of BSI were 69% for SBP, 38% for pneumonia, 33% for CR-BSI, 25% for primary BSI, and 9% for urinary infection. Among patients with candidemia 56% died within 30 days after blood cultures were drawn.
4
Multiple candidate risk factors for 30-day crude mortality were identified in univariate analysis (Table 1). Crude mortality rates varied considerably depending on the infectious etiology and antibiotic susceptibility (Fig. 1). Among the 6-most common isolated pathogens (n = 89), MSSA and antibiotic-susceptible E. coli BSI were associated with 30-day mortality rates of <20%, whereas ESBL-E. coli and E. faecalis were associated with crude mortality rates of 23–30%. The highest BSI crude mortality rates were found for Candida spp. (56%) and KPC-producing Enterobacteriaceae (60%). A detailed comparison of microbial infection etiology between surviving and non-surviving cirrhotic patients with BSI is presented in Table 2. Distribution of the MELD values calculated at baseline, BSI onset, and at 24, 48, 72, and 96 h after the BSI onset between survivors and non-survivors is shown in Fig. 2A. Receiver-operator curve analysis suggested that DMELD calculated from baseline to the time of positive BC had the highest discrimination for 30-day mortality status (aROC 0.89, 0.83–0.92) even though it was not statistically better than MELD score calculated at the time of drawing blood cultures (aROC 0.81, 0.74–0.87) (Fig. 2B). Five candidate mortality variables identified by univariate analysis (hepatorenal syndrome, SBP as BSI source, presentation with severe sepsis or septic shock, delayed appropriate antimicrobial therapy >24 h) plus DMELD were analyzed using Cox proportional hazards regression. Four covariates were retained in the final model derived from 150 patient cases; DMELD, SBP as BSI source, severe sepsis or septic shock, and receipt of inappropriate empirical within the first 24 h of BC (Table 3). The model displayed acceptable goodness of fit, correctly predicting the 30-day survival status of 85% of the cirrhotic patient cohort with a sensitivity of 90% and a specificity of 75% (Table 3). Use of MELD scores calculated at the time of BSI rather than DMELD resulted in less favorable model performance (AIC 383 vs. 355, respectively). Of the 105 patients with a primary bacterial BSI, 24 (22.9%) died within 30 days from BSI onset. In this group, the variables associated with 30-day mortality were: hepatorenal syndrome, presentation with severe sepsis or septic shock and higher DMELD (data not shown). Examination of whether isolation of a specific pathogen may be associated with significantly lower odds of receiving appropriate empirical therapy suggested that Candida BSI has the strongest association with inappropriate empirical therapy (data not shown). The specific impact of inappropriate vs. appropriate antibiotic therapy for BSI in cirrhotic patients is presented in Fig. 3. The percentages of patients receiving appropriate empirical therapy within 24 h from BCs were 54% and 63% between patients with DMELD >4 and those with DMELD 64, respectively (p = 0.17). Discussion To the best of our knowledge, this retrospective study represents the largest contemporary epidemiological survey of BSI in cirrhotic patients. Gram-negative bacteria were the leading pathogens with a high prevalence of MDR and XDR strains. In addition, 10% of BSIs were caused by Candida spp. Patients who received an inappropriate empirical antimicrobial therapy within 24 h after blood cultures were drawn had higher 30-day mortality rate. The majority of studies on infection in cirrhotic patients have analyzed patients with SBP. Few studies have examined the
Journal of Hepatology 2014 vol. xxx j xxx–xxx
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021
JOURNAL OF HEPATOLOGY 1.00
Survival probability
0.90 0.80 0.70
Susceptible, n = 98; AHR 0.58 (0.31-1.09); p = 0.1
0.60
ESBL and FQ resistant, n = 32; AHR 0.63 (0.30-1.31); p = 0.21
0.50
Carbapenem-resistant, n = 20; AHR 1.26 (0.60-2.64); p = 0.54
0.40 0.30
Candida spp., n = 12; AHR 2.09; (0.93-4.67); p = 0.07
0.20 0.10 0.00 0
5
10
15
20
25
30
Days after positive blood culture Fig. 1. Impact of isolation of antimicrobial resistant pathogens on survival from bloodstream infection in cirrhotic patients. Survival probability was estimated for antibiotic susceptible (solid line), ESBL and fluoroquinolone-resistant (long dash), carbapenem-resistant (short dash) and Candida spp. (dash-dot) using a multivariate Cox proportional hazard regression model accounting for underlying illness severity and timing of antibiotic administration. Symbols represent the unadjusted observed mortality rates for each pathogen subset estimated from Kaplan-Meier curves. Antibiotic susceptible (d), ESBL and fluoroquinolone resistant (), carbapenem-resistant (N) and Candida spp. (j). AHR, adjusted hazard ratio. Table 2. Microbiology etiology distribution between survivors and non-survivors groups per BSI episode.
Gram positive Coagulase negative staphylococci Staphylococcus aureus (MSSA) Staphylococcus aureus (MRSA) Streptococcus pneumoniae Enterococcus faecalis Enterococcus faecium Other Gram positivea Gram negative Enterobacteriaceae Escherichia coli Escherichia coli (FQR) Escherichia coli (ESBL) Enterobacter species Klebsiella pneumoniae Klebsiella pneumoniae (FQR) Klebsiella pneumoniae (ESBL) Klebsiella pneumoniae (KPC) Other enterobactericeaeb Non-fermenters Pseudomonas aeruginosa Acinetobacter baumanii Stenotrophomonas maltophilia Fungi Candida species Mixed infections
Total, n = 162 episodes (100%) 62 (38) 10 (6) 15 (9) 6 (3) 4 (2) 15 (9) 12 (7) 13 (8) 104 (64) 77 (47) 22 (13) 4 (2) 17 (10) 7 (4) 4 (2) 4 (2) 7 (4) 14 (8) 3 (1) 25 (15) 10 (6) 10 (6) 5 (3)
Survivors, n = 115 episodes (71%) 43 (37) 6 (5) 13 (11) 4 (3) 2 (2) 10 (9) 8 (7) 13 (11) 76 (66) 57 (50) 18 (16) 4 (3) 13 (11) 6 (5) 3 (3) 4 (3) 2 (2) 7 (6) 2 (2) 19 (17) 8 (7) 8 (7) 3 (3)
Non-survivors, n = 47 episodes (29%) 19 (40) 4 (9) 2 (4) 2 (4) 2 (4) 5 (11) 4 (9) 0 (0) 28 (60) 22 (47) 4 (9) 0 (0) 4 (9) 1 (2) 1 (2) 0 (0) 5 (11) 7 (15) 1 (2) 6 (13) 2 (4) 2 (4) 2 (4)
p value
0.75 0.42 0.16 0.06 0.35 0.70 0.73 0.01 0.47 0.86 0.23 0.65 0.60 0.18 0.86 0.20 0.01 0.07 0.99 0.64 0.73 0.73 0.63
16 (9) 30 (18)
7 (6) 22 (19)
9 (19) 8 (17)
0.01 0.83
MSSA, methicillin susceptible S. aureus; MRSA, methicillin-resistant S. aureus; FQR, fluoroquinolone-resistant; ESBL, extended-spectrum beta-lactamase; KPC, Klebsiella pneumonia carbapenemase. a Aeromonas spp. (n = 1), Enterococcus durans (n = 1), Listeria monocytogenes (n = 1), Micrococcus luetues (n = 1), beta-hemolytic streptococci (n = 9). b Serratia marcescens (n = 1), Citrobacter freundii (n = 2).
general epidemiology of BSI in cirrhotic patients. The two largest studies on incidence, etiological distribution and outcome of BSI in cirrhotic patients were published more than a decade ago and did not evaluate the specific impact of inadequate empirical therapy on outcome [6,26].
In the present study of 162 cirrhotic patients with BSI, GNB accounted for more than 60% of BSIs with a prevalence of MDR and XDR pathogens of 25% and 21% among Gram-negative isolates, respectively. Therefore, cephaloporins, which are recommended in current guidelines for infection in cirrhotic patients,
Journal of Hepatology 2014 vol. xxx j xxx–xxx
5
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021
Research Article A
35
Non-survivors day 30; n = 47 (29%)
B 100
80 Sensitivity
Median MELD ± 95% CI
Δ MELD BSI
30
25
20
15
MELD at BSI
Δ MELD 72 h Δ MELD 48 h
60
Δ MELD 24 h
40 Variable MELD BSI Δ MELD BSI Δ MELD 24 h Δ MELD 48 h Δ MELD 72 h
20 Survivors day 30; n = 115 (71%)
10
aROC (95% CI) 0.81 (0.74-0.87) 0.89 (0.83-0.94) 0.54 (0.45-0.62) 0.66 (0.56-0.76) 0.76 (0.66-0.86)
0 Baseline
BSI Diagnosis
24
48
72
96 h
0
20
40
60
80
100
100-Specificity
Time of assessment Fig. 2. MELD score as prognostic index for 30-day survival in cirrhotic patients with bloodstream infection. (A) Temporal patterns of median MELD score in 30-day non-survivors (open circles) and survivors (closed circles); (B) comparison of aROC for MELD score variables as a predictor of 30-day survival status.
Table 3. Final Cox-proportional hazards regression model for 30-day BSI mortality in cirrhotic patients.
Model covariate Spontaneous bacterial peritonitis Sepsis grading Inadequate empirical therapy within 24 h from BCs ΔMELD (per 1 point increase)
Hazard ratio 4.42 2.18 2.92 1.11
95% CI 2.04-9.54 1.39-3.43 1.50-5.41 1.07-1.15
p value 0.0002 0.0007 0.002 <0.0001
Model goodness of fit: Naglekerke’s R2 = 0.62; p <0.0001; Calibration: Hosmer-Lemeshow v2 = 8.34, degrees freedom = 8; p = 0.40). Discrimination: aROC 0.91 (95% CI 0.85– 0.95), p <0.0001.
1.00 Adequate antibiotic therapy within 24 hours (n = 100)
Survival probability
0.90 0.80 0.70 0.60 0.50 0.40
Inadequate antibiotic therapy within 24 hours (n = 62)
0.30 0.20
Adjusted HR 0.37 (0.20-0.70) p = 0.002
0.10 0.00 0
5
10 15 20 25 Days after positive blood culture
30
Fig. 3. Impact of inadequate antimicrobial therapy in the first 24 h on survival of cirrhotic patients with bloodstream infection. Probability of survival in patients receiving adequate therapy during first 24 h (solid line) or inadequate empirical therapy (dashed line) was estimated using multivariate Cox-hazard proportional regression accounting for severity of underlying illness. Actual unadjusted observed mortality rates were determined by Kaplan-Meier curves, and plotted for patients receiving adequate (j) or inadequate (h) antimicrobial therapy in the first 24 h.
would be inadequate as empirical therapy in over 40% of cases, mainly in hospital-acquired BSI. In addition, 10% of BSIs were caused by Candida spp., and, although data should be interpreted with caution due to the small number of events, the isolation of fungi was associated with the highest 30-day crude mortality rate (56%). Importantly, liver cirrhosis is a chronic underlying chronic condition that results in frequent contacts with the 6
healthcare system, especially in the advanced phase of the disease when the risk for infection complications is higher. The mean MELD at hospital admission of our patients was 20 ± 8 and most patients have had a recent contact with the hospital that surely influenced the etiology of infection. In contrast, among patients with true community acquired BSI, the most common pathogen was Streptococcus pneumoniae associated with pneumonia. DMELD, SBP as BSI source and severe sepsis or septic shock were independently associated with 30-day all-cause mortality in our study. Whereas, receipt appropriate empirical therapy was independently associated with better survival. Cirrhosis-specific scores, such as the Child Pugh and MELD scores, have been reported to be good predictors for short and long term mortality in patients with cirrhosis even in presence of infection [27,28]. Indeed, infection severity in patients with cirrhosis should be assessed using a liver disease calibrated scores rather than that used in general population [29]. Our analysis suggests that temporal trends in MELD score may be useful for characterizing infection-associated decompensation in patients with liver cirrhosis, and patients at highest risk for death. However, our definition of DMELD may be a source of bias because the time interval between the baseline MELD, defined as the last MELD available within two weeks before the BSI, and the MELD score calculated at the time of BC may have been variable between patients. Further studies will be required to examine the performance of this risk factor and whether patients with a higher DMELD would benefit from broader-spectrum empirical therapy that includes coverage of resistant GNB and Candida spp. To note
Journal of Hepatology 2014 vol. xxx j xxx–xxx
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021
JOURNAL OF HEPATOLOGY that all patients had at least criteria of systemic inflammatory response syndrome, this may partially be explained with the unique characteristics of the cirrhosis such as the hyper-dynamic circulatory status, low peripheral white blood cell count, and impaired synthesis of acute-phase proteins which can alter the accuracy of the classical symptoms and signs of sepsis. However, despite this bias, the sepsis grading seems to work also in this population, and therefore should be routinely included in the evaluation work-up of patients with suspicion of infection. Our study had some limitations. It is a monocentric retrospective study, thus epidemiological data may not be representative of other institutions. However, our hospital is a referral center for liver disease with many patients coming from different regions of the country. Additionally, the development of a more robust risk model for predicting mortality in cirrhotic patients with BSI was limited by the number of study subjects in our dataset. Nevertheless, our findings suggest a growing proportion of BSIs in cirrhotic patients are now caused by resistant GNB and Candida spp., raising concerns whether empirical antimicrobial regimens for BSI in cirrhotic patients should be reconsidered, especially in institutions with high (>10%) rates of ESBL and carbapenemase-producing enterobacteriaceae. According to our results, empirical antibiotic treatment should be guided first of all by infection severity (severe sepsis or septic shock), concomitant presence of SBP, site of infection acquisition (community, healthcare associated, hospital), and worsening of liver failure according to the MELD score at the time of drawing positive BCs. All these variables could be the first step toward the creation of a multifaceted clinical risk score in order to help clinicians in indentifying the highest risk subset of patients who would likely most benefit from broader-spectrum empirical therapy.
Conflict of interest The authors who have taken part in this study declared that they do not have anything to disclose regarding funding or conflict of interest with respect to this manuscript.
Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.jhep.2014.03. 021.
References [1] Leon DA, McCambridge J. Liver cirrhosis mortality rates in Britain from 1950 to 2002: an analysis of routine data. Lancet 2006;367:52–56. [2] Leber B, Spindelboeck W, Stadlbauer V. Infectious complications of acute and chronic liver disease. Semin Respir Crit Care Med 2012;33:80–95. [3] Moreau R, Jalan R, Gines P, Pavesi M, Angeli P, Cordoba J, et al. Acute-onchronic liver failure is a distinct syndrome that develops in patients with acute decompensation of cirrhosis. Gastroenterology 2013;144:e1429. [4] Fernandez J, Acevedo J, Castro M, Garcia O, de Lope CR, Roca D, et al. Prevalence and risk factors of infections by multiresistant bacteria in cirrhosis: a prospective study. Hepatology 2012;55:1551–1561. [5] Kang CI, Song JH, Chung DR, Peck KR, Yeom JS, Ki HK, et al. Liver cirrhosis as a risk factor for mortality in a national cohort of patients with bacteremia. J Infect 2011;63:336–343.
[6] Thulstrup AM, Sorensen HT, Schonheyder HC, Moller JK, Tage-Jensen U. Population-based study of the risk and short-term prognosis for bacteremia in patients with liver cirrhosis. Clin Infect Dis 2000;31:1357–1361. [7] Reiberger T, Ferlitsch A, Payer BA, Mandorfer M, Heinisch BB, Hayden H, et al. Non-selective betablocker therapy decreases intestinal permeability and serum levels of LBP and IL-6 in patients with cirrhosis. J Hepatol 2013;58:911–921. [8] Bellot P, Frances R, Such J. Pathological bacterial translocation in cirrhosis: pathophysiology, diagnosis and clinical implications. Liver Int 2013;33:31–39. [9] Linderoth G, Jepsen P, Schonheyder HC, Johnsen SP, Sorensen HT. Short-term prognosis of community-acquired bacteremia in patients with liver cirrhosis or alcoholism: a population-based cohort study. Alcohol Clin Exp Res 2006;30:636–641. [10] Ortega M, Almela M, Soriano A, Marco F, Martinez JA, Munoz A, et al. Bloodstream infections among human immunodeficiency virus-infected adult patients: epidemiology and risk factors for mortality. Eur J Clin Microbiol Infect Dis 2008;27:969–976. [11] Chen SY, Tsai CL, Lin CH, Lee CC, Chiang WC, Wang JL, et al. Impact of liver cirrhosis on mortality in patients with community-acquired bacteremia. Diagn Microbiol Infect Dis 2009;64:124–130. [12] Chang TY, Lee CH, Liu JW. Clinical characteristics and risk factors for fatality in patients with bloodstream infections caused by glucose non-fermenting gram-negative Bacilli. J Microbiol Immunol Infect 2010;43:233–239. [13] Al-Hasan MN, Lahr BD, Eckel-Passow JE, Baddour LM. Predictive scoring model of mortality in Gram-negative bloodstream infection. Clin Microbiol Infect 2013;19:948–954. [14] Grim SA, Berger K, Teng C, Gupta S, Layden JE, Janda WM, et al. Timing of susceptibility-based antifungal drug administration in patients with Candida bloodstream infection: correlation with outcomes. J Antimicrob Chemother 2012;67:707–714. [15] Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40:373–383. [16] Kamath PS, Wiesner RH, Malinchoc M, Kremers W, Therneau TM, Kosberg CL, et al. A model to predict survival in patients with end-stage liver disease. Hepatology 2001;33:464–470. [17] Pugh RN, Murray-Lyon IM, Dawson JL, Pietroni MC, Williams R. Transection of the oesophagus for bleeding oesophageal varices. Br J Surg 1973;60:646–649. [18] Bone RC, Balk RA, Cerra FB, Dellinger RP, Fein AM, Knaus WA, et al. Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine. Chest 1992;101:1644–1655. [19] EASL clinical practice guidelines on the management of ascites, spontaneous bacterial peritonitis, and hepatorenal syndrome in cirrhosis. J Hepatol 2010;53:397–417. [20] Levey AS, Coresh J, Balk E, Kausz AT, Levin A, Steffes MW, et al. National Kidney Foundation practice guidelines for chronic kidney disease: evaluation, classification, and stratification. Ann Intern Med 2003;139:137–147. [21] Hunt SA, Abraham WT, Chin MH, Feldman AM, Francis GS, Ganiats TG, et al. ACC/AHA 2005 Guideline Update for the Diagnosis and Management of Chronic Heart Failure in the Adult: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Writing Committee to Update the 2001 Guidelines for the Evaluation and Management of Heart Failure): developed in collaboration with the American College of Chest Physicians and the International Society for Heart and Lung Transplantation: endorsed by the Heart Rhythm Society. Circulation 2005;112:e154–e235. [22] Horan TC, Andrus M, Dudeck MA. CDC/NHSN surveillance definition of health care-associated infection and criteria for specific types of infections in the acute care setting. Am J Infect Control 2008;36:309–332. [23] Friedman ND, Kaye KS, Stout JE, McGarry SA, Trivette SL, Briggs JP, et al. Health care–associated bloodstream infections in adults: a reason to change the accepted definition of community-acquired infections. Ann Intern Med 2002;137:791–797. [24] Magiorakos AP, Srinivasan A, Carey RB, Carmeli Y, Falagas ME, Giske CG, et al. Multidrug-resistant, extensively drug-resistant and pandrug-resistant bacteria: an international expert proposal for interim standard definitions for acquired resistance. Clin Microbiol Infect 2012;18:268–281. [25] Parzen M, Lipsitz SR. A global goodness-of-fit statistic for Cox regression models. Biometrics 1999;55:580–584. [26] Campillo B, Dupeyron C, Richardet JP. Epidemiology of hospital-acquired infections in cirrhotic patients: effect of carriage of methicillin-resistant
Journal of Hepatology 2014 vol. xxx j xxx–xxx
7
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021
Research Article Staphylococcus aureus and influence of previous antibiotic therapy and norfloxacin prophylaxis. Epidemiol Infect 2001;127:443–450. [27] Thabut D, Massard J, Gangloff A, Carbonell N, Francoz C, Nguyen-Khac E, et al. Model for end-stage liver disease score and systemic inflammatory response are major prognostic factors in patients with cirrhosis and acute functional renal failure. Hepatology 2007;46:1872–1882.
8
[28] Durand F, Valla D. Assessment of prognosis of cirrhosis. Semin Liver Dis 2008;28:110–122. [29] Viasus D, Garcia-Vidal C, Castellote J, Adamuz J, Verdaguer R, Dorca J, et al. Community-acquired pneumonia in patients with liver cirrhosis: clinical features, outcomes, and usefulness of severity scores. Medicine (Baltimore) 2011;90:110–118.
Journal of Hepatology 2014 vol. xxx j xxx–xxx
Please cite this article in press as: Bartoletti M et al. Epidemiology and outcomes of bloodstream infection in patients with cirrhosis. J Hepatol (2014), http://dx.doi.org/10.1016/j.jhep.2014.03.021