Bloodstream infection after elective liver transplantation is associated with increased mortality in patients with cirrhosis

Journal of Critical Care (2011) 26, 468–474

Outcomes/Predictors

Bloodstream infection after elective liver transplantationis associated with increased mortality in patientswith cirrhosis☆

Constantine J. Karvellas MD FRCPCa,b,⁎, Mark McPhail MBBS, PhDb, Fred Pink MBBSb,Sonal Asthana MBBS MRSC c, Paolo Muiesand, Nigel Heaton MBBS FRSCb,Georg Auzinger MBBSb, William Bernal MBBS, MDb, Ian Eltringham MBBSb,Julia A. Wendon MBChB, FRCPb

aDivisions of Gastroenterology (Liver Unit) and Critical Care Medicine, University of Alberta, Edmonton, CanadabInstitute of Liver Studies, King's College Hospital, Denmark Hill, London, United KingdomcDepartment of Surgery, University of Alberta, Edmonton, CanadadThe Liver Unit, Birmingham University Hospital, Birmingham, United Kingdom

lidL

C

s(G

0d

Keywords:Chronic liver disease;Sepsis;Bloodstream infection;Bacteremia;Liver transplantation

AbstractPurpose: This study aims to investigate what factors predict the development of postoperativebloodstream infection (BSI) in patients transplanted electively for chronic liver disease and compareoutcomes in infected transplant recipients (BCLD) with noninfected patients (CLD).Methods: A retrospective cohort study of 218 patients who had elective liver transplantation (LT)between January 2003 and July 2005 and admitted to a specialist intensive care unit (ICU) was done.Results: Fifteen patients had BSI post-LT (BCLD, 29 isolates) while in the ICU, and 203 patients didnot (CLD). Thirty-eight percent of isolates were gram negatives; 55%, gram positives; and 7%,fungemia. Median time to first BSI post-LT was 11 days (range, 3-16 days). On admission post-LT tothe ICU, patients with BCLD had higher Acute Physiology and Chronic Health Evaluation II scores(23 vs 10, P b .001). While in the ICU, patients with BCLD had greater requirements for renalreplacement therapy (73% vs 8%) and days on mechanical ventilation (17 vs 2 days) and longermedian ICU stay (21 vs 3 days, P b .001 for all). One-year survival was worse in the BCLD group(40% vs 94%, P b .001). On multivariate analysis, Acute Physiology and Chronic Health Evaluation II

Abbreviations: APACHE II, Acute Physiology and Chronic Health Evaluation II score; BCLD, transplanted patient with cirrhosis (chronicver disease) with bloodstream infection; BSI, bloodstream infection (bacterial, fungal); CLD, transplanted patient with cirrhosis (chronic liverisease) without bloodstream infection (control); ICU, intensive care unit; LT, liver transplantation (elective); MELD, Model for End-Stageiver Disease score; RRT, renal replacement therapy.

☆ This work was performed initially at King's College Hospital.⁎ Corresponding author at: Division of Gastroenterology (Liver Unit), Division of Critical Care Medicine, University of Alberta, Edmonton, Alberta,

anada T6G 2X8. Tel.: +1 780 248 1555; fax: +1 780 492 5643.E-mail addresses: [email protected] (C.J. Karvellas), [email protected] (M. McPhail), [email protected] (F. Pink),

[email protected] (S. Asthana), [email protected] (P. Muiesan), [email protected] (N. Heaton), [email protected]. Auzinger), [email protected] (W. Bernal), [email protected] (I. Eltringham), [email protected] (J.A. Wendon).

883-9441/$ – see front matter © 2011 Elsevier Inc. All rights reserved.oi:10.1016/j.jcrc.2010.12.018

469Bloodstream infection post liver transplant

score (odds ratio, 1.36) post-LT was independently associated with subsequent BSI. Bloodstreaminfection (hazards ratio, 8.7) was independently associated with mortality.Conclusion: Bloodstream infection post-LT was associated with increased severity of illness onadmission, greater requirements for organ support, and increased mortality.© 2011 Elsevier Inc. All rights reserved.

1. Introduction

Patients with cirrhosis are abnormally susceptible toinfection as a result of immunologic deficits in thepretransplant period. Mechanisms previously describedinclude decreased hepatic production of complement (re-duced C3 and C5 levels), impaired Kupffer cell function(phagocytosis), altered neutrophil chemotaxis, and impairedclearance of inflammatory cytokines [1-4]. Patients may alsodemonstrate decreased bactericidal activities of immunoglo-bulins against gram-negative pathogens [5]. The reticuloen-dothelial cells represent a potent filtering mechanism forportal bloodborne pathogens. In the presence of cirrhosis,filtering is impaired by impaired macrophage mobilizationand diminished killing of bacteria [6].

Perhaps as a result of these deficits, bloodstream infection(BSI) remains a significant concern in the early posttrans-plant period and by far the most frequently occurringinfection complication after liver transplantation [7-9].Bacteremia accounts for up to 30% of all major infectionsafter liver transplantation with an attributable mortality of24% to 36% [8,10]. Most bacteremias occur within the first 3months after transplantation with vascular catheters, the lungand biliary tract being the most common sources [8].Recently, there has been interest in the shift of responsiblepathogens in elective liver transplantation (LT) patients. Inthe 1980s, BSIs were predominantly gram negative (Enter-obacter sp), most often associated with an intra-abdominalsource [11,12]. Studies from the 1990s showed anemergence of methicillin-resistant Staphylococcus aureus(MRSA) as the leading causative agent likely related tocatheter-related bloodstream infections [13]. However, 2more recent studies from Pittsburgh (USA) and Clichy(France) have shown a shift toward gram-negative patho-gens, possibly related to improvement in screening andprevention of MRSA [10,14].

In this study, we reviewed electively (from home ormedical ward) transplanted patients with cirrhosis admittedto a specialized intensive care unit (ICU) posttransplantbetween January 2003 and July 2005. We comparedphysiologic and biochemical profiles along with clinicaloutcomes of patients who developed a significant BSI duringtheir posttransplant ICU admission (BCLD) with 203patients admitted during the same period who did not goon to develop a BSI during admission (CLD). We wereparticularly interested in answering the following questions:what physiologic, biochemical, and donor factors predict the

development of BSI in patients transplanted electively forchronic liver disease in the initial perioperative ICU period;and upon diagnosis of BSI, what is their natural historycompared with noninfected controls?

2. Materials and methods

Physiologic and microbiologic data were prospectivelycollected on admission for all elective (nonurgent) transplantpatients admitted to a specialized ICU at King's CollegeHospital between January 2003 and July 2005 and enteredinto a database. Patients were included if they weretransplanted for cirrhosis from home or the medical wardand were admitted to intensive care after transplant. Patientsrequiring organ support in ICU pretransplant were excluded.Retransplanted patients and combined liver/kidney trans-plant recipients were also excluded from this analysis. Theinstitutional review board waived the need for informedconsent for use of these data.

2.1. Study definitions

Illness severity was defined according to the AcutePhysiology and Chronic Health Evaluation (APACHE) IIscore. The APACHE II uses a point score based on initialvalues of 12 routine physiologic measurements, age, andprevious health status (cirrhosis, severe cardiopulmonarydisease, chronic renal failure, immunocompromized) toprovide a general measure of severity of disease. Anincreasing score (range, 0-71) correlates with the subsequentrisk of hospital death in the general ICU population [15].Severity of liver dysfunction was defined by the ModifiedEnd-Stage Liver Disease (MELD) score [16].

2.2. Blood cultures/BSI

Standard aerobic and anaerobic (2 × 10 mL of blood)paired samples were taken on admission to the ICU fromnew central lines inserted as well as 1 peripheral site. Bloodcultures were also taken upon clinical suspicion of sepsis(temperature, hypotension, and abnormalities of central linesite). A significant BSI was defined as 1 culture yielding arecognized pathogen (eg, S aureus, enterococcus, member ofEnterobacteriaceae, Pseudomonas sp, Candida sp) from asingle sample or an organism that forms part of the normal

Table 1 Etiologies of 218 patients nonurgently transplantedpatients with cirrhosis

Etiology CLD (n = 203) BCLD(n = 15)

Alcohol 48 (24) 2 (13)Viral (HBV, HCV) 56 (28) 4 (27)Biliary cirrhosis 23 (11) 2 (13)Primary sclerosing cholangitis 20 (10) 0Autoimmune 11 (5) 1 (7)NASH/cryptogenic 17 (8) 2 (13)Metabolic 18 (9) 2 (13)Other 10 (5) 2 (13)

Values are presented as n (%).CLD indicates transplanted patient with cirrhosis; BCLD, transplantedpatient with cirrhosis with a bloodstream infection; HBV/HCV, hepatitisB/C; NASH, nonalchoholic steatohepatitis. Metabolic indicates hemo-chromatosis, amyloid, Wilson, and α1-antitrypsin deficiency. Otherindicates Budd-Chiari syndrome, venoocclusive disease, and liver cysts.

Fig. 1 Etiologies and frequencies of 29 blood culture isolates in15 patients nonurgently transplanted for chronic liver disease.

470 C.J. Karvellas et al.

skin flora isolated from more than 1 culture occurring within a48-hour period. This is consistent with the NosocomialInfection National Surveillance Scheme criterion 1 for thedefinition of BSI [17]. The isolation of different species fromdifferent samples or isolates of the same species from samplestaken more than 14 days apart were defined as separateepisodes. Blood culture samples were only included in thisanalysis if they were taken post-LT. Patients were excludedfrom this analysis if positive blood culture samples were takenbefore admission to the ICU, either from the ward or from areferring hospital. Positive blood culture dates and organismswere collected from microbiology databases. Nonguidedqualitative tracheal aspirate data were collected for all patientsincluded in this study posttransplant. Correlation betweenchest and bloodstream pathogens was determined. Chestradiograph data were not captured in this study.

For all patients analyzed (infected and noninfected),lengths of stay (in hospital and ICU); requirement of renalreplacement therapy (RRT); and mechanical ventilationalong with other physiologic, biochemical, and outcomedata were obtained from the MedTRACK database, patientcharts, and discharge summaries.

2.3. Posttransplant ICU care

Nonurgent transplant recipients were admitted to the ICUsedated and ventilated in the immediate postoperative periodto allow confirmation of hemodynamic stability, normother-mia, and acceptable graft and renal function. In uncompli-cated cases, the patients were extubated once the abovecriteria have been satisfied. Postoperative immunosuppres-sion regimens (assuming normal renal function) consisted ofintravenous methylprednisolone/oral prednisone and tacro-limus depending on graft and renal function. All posttrans-plant patients received prophylactic antibiotics for 48 hourspostoperative. Lymphocyte-depleting therapies (OKT3,

antithymocyte globulin) were not used routinely in thispatient population.

Elective transplant recipients received flucloxacillin andgentamicin for 48 hours posttransplant. Patients who hadrecently received intravenous antibiotics or recently beenhospitalized were given a 5-day course of piperacillin/tazobactam ormeropenem (if they were known to be penicillinallergic) in the first instance. If the patient was known to be acarrier of multidrug-resistant organisms (eg, Pseudomonas/MRSA), then amikacin or vancomycin was added. Subsequentchanges to antibiotics were based on cultures (blood, urine,bronchial wash, and ascites). In situations where empiricaltherapy was required, antibiotic choice was driven by clinicalcontext and surveillance swab–positive status. Vancomycin-resistant enterococcus was treated with linezolid, whereasfungal sepsis was treated with a mixture of agents dependingon isolate and likely sensitivity pattern.

2.4. Statistics

Statistical analysis was performed using the StatisticalPackage for the Social Sciences version 17 (SPSS, Inc,Chicago, Ill) and STATA 10 (College Station, Tex). In theunivariate analysis, categorical variables were analyzedusing the χ2 and Fisher exact tests as appropriate. Measuresof central tendency for continuous variables were comparedusing the t test (parametric variables) following normalitytesting and the Mann-Whitney U test (nonparametricvariables). Parametric variables were reported as mean(SD); and nonparametric variables, as medians (interquartileranges). Statistical significance (95% confidence interval[CI]) was reported for P b .05 with a trend towardsignificance for P b .10. Variables that came out statisticallydifferent between the groups in univariate analysis wereentered into multivariate analysis if the P value achieved onunivariate analysis was .10 or less. Logistic regression wasused to determine predictors of BSI, and Cox proportionalhazards regression model was used to determine predictorsof mortality. Where results were found to be significant

471Bloodstream infection post liver transplant

(ie, P b .05; CIs do not cross unity), values are reportedas odds/hazards ratios with 95% CIs.

3. Results

3.1. Etiology and microbiology information forpatients with BSIs

Fifteen patients met the Nosocomial Infection NationalSurveillance Scheme criteria for significant BSI (BCLD);203 patient controls did not (CLD). A complete list ofetiologies of pretransplant diagnoses is listed in Table 1. Themost common etiologies were viral (BCLD, 27%; CLD,28%) and alcohol (BCLD, 13%; CLD, 24%).

From the 15 patients who had significant BSI while in theICU, 29 isolates were identified. The distribution of

Table 2 Comparison of admission and outcome data (univariate) for

Variable CLD

203/218 (93%)Age (y) 53 (17-54)Sex, male/female 141/62Home/in hospital 179/24PretransplantMELD 13 (10-18)Creatinine (mmol/L) 93 (76-108)INR 1.2 (1.1-1.4)On admissionRespiratory rate 15 (10-35)Temperature (°C) 37 (34-39)Mean arterial pressure (mm Hg) 71 (50-132)Heart rate 95 (20)On vasopressors 19 (9%)Glasgow Coma Score 15 (3-15)White blood count 5.8 (0.9-35)PCO2 (mm Hg) 39.8 (24.0-62.3)PO2/Fio2 307 (105)INR 1.3 (0.9-2.5)Creatinine (mmol/L) 93 (10-418)Bilirubin (mmol/L) 55 (4-783)Lactate (mmol/L) 2.1 (0-11)pH 7.36 (7.1-7.54)APACHE II 10 (4-23)During ICU stayRRT during admission 16/203 (8%)Days on RRT 0 (0-16)Days on ventilator 2 (0-18)Length of ICU stay 3 (1-59)Long-term outcomesRetransplant 11/203 (5%)Survival (mo) 58.7 (0.1-75.8)1-y survival 190/203 (94%)Cumulative survival 178/203 (88%)

Parametric continuous data are represented as mean (SD); and nonparametric, asINR indicates international normalized ratio; FIO2, fraction of inspired oxygen.

frequency of isolates is shown in Fig. 1. Gram positiveswere seen in 16 isolates (55%); and gram negatives, in 11(38%). Candidemia was observed in 2 (7%) isolates. Themost frequent organisms causing bacteremia were thefollowing: enterococcus (Enterococcus faecalis [n = 6isolates, 21%], Enterococcus facium [n = 4, 14%],vancomycin-resistant enterococcus [n = 4, 14%]), Klebsiellasp (n = 4, 14%), and Pseudomonas (n = 2, 7%). The mediantime to first bacteremia was 11 days (range, 3-16 days).

3.2. Univariate analysis

Results of univariate analysis of all 218 patients arepresented in Table 2. Pretransplant, there were no significantdifferences in age, sex, and biochemical profile includingwhite blood count, creatinine, international normalized ratio,and pretransplant MELD score (CLD, 13, vs BCLD, 17;

patients with/without BSIs transplanted electively for cirrhosis

BCLD P

15/218 (7%)49 (18-66) .5610/5 .9515/0

17 (12-23) .09100 (82-168) .301.2 (1.1-1.3) .68

18 (12-44) .00436 (35-39) .1262 (501-22) .079101 (30) .413 (20%) .3212 (7-15) b.0017.0 (0.8-33) .1939.8 (28.5-60.8) .24236 (137) .0211.2 (1.0-2.2) .29147 (69-318) b.00136 (10-372) .172.7 (1.3-8.7) .127.26 (7.06-7.44) .00723 (10-42) b.001

11/15 (73%) b.00116 (0-62) b.00117 (1-66) b.00121 (1-121) b.001

3/15 (20%) .093.2 (0.8-65.0) b.0016/15 (40%) b.0015/15 (33%) b.001

median (interquartile range). P = not significant if not otherwise specified.Boldface indicates a statistically significant P-value (b.05).

Table 3 Donor and operative factors for 218 nonurgentlytransplanted patients with cirrhosis

CLD(n = 203)

BCLD(n = 15)

P-value

Donor factorsCold ischemia time, N12 h 67 (33%) 4 (27%) 0.47Split graft 28 (14%) 0 0.12Donor in ICU, N96 h 26 (13%) 0 0.22Severe steatosis (N40%) 9 (4%) 0 1.0Donor age, N60 y 31 (15%) 2 (13%) 1.0Donor score (out of 5) 1 (0-1) 0 (0-1) 0.08Operative factorsOperating room time (h) 6.5 (1.8) 8.0 (2.7) 0.065Roux-en-Yhepaticojejunostomy

27 (13%) 6 (40%) 0.014

472 C.J. Karvellas et al.

P = .09) between the 2 groups. Post-LT, several factorsshowed statistically significant differences. On admission tothe ICU posttransplant (first 24 hours posttransplant),patients with BCLD had significantly higher APACHE IIscores (23 vs 10, P b .001) compared with controls. Patientswith BCLD also presented with worse encephalopathy(Glasgow Coma Score, 12 vs 15; P b .001) and respiratoryparameters (PaO2/fraction of inspired oxygen ratio andrespiratory rate, P b .05 for both). There were no significantdifferences in other biochemical or physiologic markers onpost-LT admission to the ICU.

During the course of overall ICU stay, more patients withBCLD required RRT (BCLD, 73%, vs CLD, 8%) and for alonger period (16 vs 0 days, P b .001 for both). Mean lengthof mechanical ventilation was significantly longer in theBCLD group (17 vs 2 days). Length of ICU stay wassignificantly higher in the BCLD group (21 vs 3 days). Graftfailure requiring retransplantation was not significantlyhigher (BCLD, 20%; CLD, 5%). Three-year survival wassignificantly worse in the BCLD group (BCLD, 33%; CLD,90%; P b .001). This was also demonstrated by Kaplan-Meier survival analysis (log-rank, b0.001) (see Fig. 2).

3.3. Graft/operative data

The following donor criteria were assessed: cold ischemiatime more than 12 hours, split graft, donor in ICU more than96 hours preorgan recovery, severe donor hepatic steatosis(N40%), and donor age older than 60 years. In comparing the2 groups, there were no statistically significant differencesin any donor category/score (see Table 3). However, patientswith BCLD had longer operating room times (8.0 vs6.5 hours, P = .065) and more hepaticojejunostomies(Roux-en-Y) (40% vs 13%, P = .014).

Fig. 2 Kaplan-Meier survival curve for electively transplantedpatients with cirrhosis (log-rank, b0.001).

3.4. Multivariate analysis (BSI and mortality)

Logistical regression (multivariate) modeling was used todetermine independent associations with BSI. Physiologic,donor and operative factors with P b 0.1 on univariateanalysis were included in our model. Factors that werecolinear with the APACHE II score were not included. In thefinal model, the only independent predictor of BSI wasAPACHE II score post-LT on admission to ICU (odds ratio1.36/U; 95% CI, 1.19-1.54; P b .001). We then performedmultivariate analysis looking for independent clinicalassociations with cumulative survival (Cox proportionalhazards regression model). Factors that were colinear withthe APACHE II score were not included. In the final model,the only independent predictor of mortality was BSI(hazards ratio, 8.7; 95% CI, 4.1-18; P b .001). Surgicalfactors (Roux-en-Y and operating room time) were notsignificant in either model.

3.5. Chest infections vs BSI: nonguided trachealaspirate data

Tracheal aspirate specimens collected between January2003 and July 2005 were retrospectively reviewed for allpatients with a diagnosis of BCLD during their ICUadmission post-LT. Of 15 patients with BCLD, 6 hadpositive tracheal aspirates (14 isolates). Two patients grew thesame bacterial species in both blood and chest samples (bothbefore diagnosis of BSI). Of all isolates (n = 14), 13 (93%)were gram-negative species (Acinetobacter, 6; Escherichiacoli, 2; Klebsiella, 2; Stenotrophomonas, 2; and Pseudomo-nas, 1); and 1 isolate was gram-positive species (MRSA, 1).

4. Discussion

This study has shown that patients transplanted nonurgentlyfor cirrhosis develop BSIs late (day 11) during ICU admission.

473Bloodstream infection post liver transplant

Recent studies examining BSI posttransplant have shownthat gram negatives, in particular, Klebsiella and Pseudomo-nas, tend to be the predominant species [8,13,14,18]. In ourcohort, the predominant species was enterococcus. Giventhat significantly more patients in the infected group hadRoux-en-Y anastomoses, these may have reflected intra-abdominal sepsis. Previous reports have demonstratedincreasing rates of bacteremia and, in particular, enterococcalbacteremia in patients undergoing Roux-en-Y hepaticojeju-nostomy in living-related donor transplantation [19,20].Other well-known risk factors for invasive enterococcaldisease include the presence of immunosuppression, struc-tural anatomic abnormalities/intestinal disruption, and expo-sure to broad-spectrum antimicrobial therapy [21].

Differences in etiology of infections may also beattributable to the use of different antimicrobial regimensafter transplantation, in particular, the use of empiricalflucloxacillin in our cohort. For contrast, Bert et al [14] statethat in Clichy (France), intraoperative cefoxitin (andvancomycin in MRSA positive) was standard prophylaxisin 2004.

In this study, patients who went on to develop BSI duringtheir posttransplant ICU stay were admitted after surgerywith a higher severity of illness as per APACHE II scores.This is consistent with previous studies that have shown thatcritically ill patients are at an increased risk for BSIs withincreasing APACHE II score, reflecting worse multiorganfailure [22,23]. Although donor factors were similar betweenthe BSI group and controls, patients who went on to developBSIs in the ICU after transplant underwent longer procedures(operating room time) and had more complex surgery (notsignificant on multivariate analysis). Although there was atrend toward a higher preoperative MELD score, this was notstatistically significant. The APACHE II was not associatedwith mortality on multivariate analysis, but presence of a BSIin ICU was.

In our cohort, BSIs were associated with significantlymore intensive support: RRT, mechanical ventilation, andprolonged ICU stay. Furthermore, Kaplan-Meier analysisshowed that patients with BSI were more likely to die aftertransplant than those who did not develop BSIs. Kim et al[24] showed a 1-year survival of 93% in patients without BSIcompared with 73% in patients with BSI. Bert et al [14] alsorecently demonstrated that BSIs are significant predictors of1-year outcome and, in particular, patients who developedBSI within 1 month. Bloodstream infections may also be amarker for other factors that predispose to higher mortality,including poor graft function, surgical complications, andprolonged ICU stay. In our cohort, although our numberswere small, there was a trend for increasing rates of graftfailure requiring retransplant in patients who developedinfection in ICU post-LT. Early bacterial infection haspreviously been shown to be a risk factor of a prolongedhospital stay, as seen in our cohort, which is associated withincreased mortality during the first year [25]. Hence, wespeculate that a de novo BSI may initiate a cascade of

systemic inflammatory events that could ultimately lead toorgan failure and impaired survival.

Our data must be interpreted within the study design. Thisis a single-center cohort study; data were analyzedretrospectively; and as such, we can only comment onassociations rather than causation. The APACHE II, asurrogate of severity of illness, was calculated on admissionposttransplant and not serially. We chose APACHE II overother organ failure scoring systems because it is readilyavailable to most physicians without proprietary softwareand acknowledge that this study may be too underpowered toshow a relation between APACHE II and mortality. Wechose to look only at nonurgently transplanted patients to tryto decipher the influence of BSIs from severity of illnesspretransplant and cannot comment on preoperative patientswho required organ support before urgent transplant.Differences in etiologies from other studies may reflectcatheter-related bloodstream, and MRSA infections may bemore preventable than intra-abdominal and biliary infectionsthat are more inherent to the complexity of surgery and lesseasily preventable. Despite these limitations, our studyappears to suggest that patients who undergo nonurgentliver transplant requiring longer operating room times andRoux-en-Y biliary anastomosis and have a higher APACHEII score posttransplant may be at higher risk for developingposttransplant BSI and may benefit from longer antimicro-bial prophylaxis with greater coverage of gram-positivepathogens, including enterococcus

5. Conclusion

In nonurgently transplanted patients with cirrhosis, BSIswere associated with increased severity of illness onadmission; greater requirements for organ support; andindependently, adversely impacted on survival. AlthoughAPACHE II was associated with BSIs posttransplant, it wasnot independently associated with mortality.

References

[1] Adams HG, Jordan C. Infections in the alcoholic. Med Clin North Am1984;68:179-200.

[2] Wyke RJ. Problems of bacterial infection in patients with liver disease.Gut 1987;28:623-41.

[3] Johnson DH, Cunha BA. Infections in cirrhosis. Infect Dis Clin NorthAm 2001;15:363-71 vii.

[4] Campbell AC, Dronfield MW, Toghill PJ, et al. Neutrophil function inchronic liver disease. Clin Exp Immunol 1981;45:81-9.

[5] Fierer J, Finley F. Deficient bactericidal activity against Escherichiacoli in patients with cirrhosis of the liver. J Clin Invest 1979;63:912.

[6] Liu YK. Effects of alcohol on granulocytes and lymphocytes. SeminHematol 1980;17:130-6.

[7] Fortun J, Martin-Davila P, Moreno S, et al. Risk factors for invasiveaspergillosis in liver transplant recipients. Liver Transpl 2002;8:1065-70.

474 C.J. Karvellas et al.

[8] Singh N,WagenerMM, Obman A, et al. Bacteremias in liver transplantrecipients: shift toward gram-negative bacteria as predominantpathogens. Liver Transpl 2004;10:844-9.

[9] Gavalda J, Len O, San Juan R, et al. Risk factors for invasiveaspergillosis in solid-organ transplant recipients: a case-control study.Clin Infect Dis 2005;41:52-9.

[10] Wade JJ, Rolando N, Hayllar K, et al. Bacterial and fungal infectionsafter liver transplantation: an analysis of 284 patients. Hepatology1995;21:1328-36.

[11] Kusne S, Dummer JS, Singh N, et al. Infections after liver trans-plantation. An analysis of 101 consecutive cases. Medicine (Baltimore)1988;67:132-43.

[12] Wagener MM, Yu VL. Bacteremia in transplant recipients: aprospective study of demographics, etiologic agents, risk factors,and outcomes. Am J Infect Control 1992;20:239-47.

[13] Singh N, Arnow PM, Bonham A, et al. Invasive aspergillosis inliver transplant recipients in the 1990s. Transplantation 1997;64:716-20.

[14] Bert F, Larroque B, Paugam-Burtz C, et al.: Microbial epidemiologyand outcome of bloodstream infections in liver transplant recipients: ananalysis of 259 episodes. Liver Transpl 16:393-401.

[15] Knaus WA, Draper EA, Wagner DP, et al. APACHE II: a severity ofdisease classification system. Crit Care Med 1985;13:818-29.

[16] Wiesner R, Edwards E, Freeman R, et al. Model for End-Stage LiverDisease (MELD) and allocation of donor livers. Gastroenterology2003;124:91-6.

[17] Document N. Nosocomial Infection National Surveillance Scheme(UK): protocol for the surveillance of hospital acquired bacteremia.London: PHLS Central Public Health Laboratory; 1997.

[18] Fortun J, Martin-Davila P, Moreno S, et al. Prevention of invasivefungal infections in liver transplant recipients: the role of prophylaxiswith lipid formulations of amphotericin B in high-risk patients.J Antimicrob Chemother 2003;52:813-9.

[19] Yamamoto S, Sato Y, Oya H, et al. Risk factors and prevention ofbiliary anastomotic complications in adult living donor livertransplantation. World J Gastroenterol 2007;13:4236-41.

[20] Patel R, Badley AD, Larson-Keller J, et al. Relevance and risk factorsof enterococcal bacteremia following liver transplantation. Transplan-tation 1996;61:1192-7.

[21] Maki DG, Agger WA. Enterococcal bacteremia: clinical features, the riskof endocarditis, and management. Medicine (Baltimore) 1988;67:248-69.

[22] Laupland KB, Kirkpatrick AW, Church DL, et al. Intensive-care-unit-acquired bloodstream infections in a regional critically ill population.J Hosp Infect 2004;58:137-45.

[23] Park DR. The microbiology of ventilator-associated pneumonia.Respir Care 2005;50:742-63 [discussion 763-745].

[24] Kim SI, Kim YJ, Jun YH, et al. Epidemiology and risk factors forbacteremia in 144 consecutive living-donor liver transplant recipients.Yonsei Med J 2009;50:112-21.

[25] Smith JO, Shiffman ML, Behnke M, et al. Incidence of prolongedlength of stay after orthotopic liver transplantation and its influence onoutcomes. Liver Transpl 2009;15:273-9.