Meta-analysis: antibiotic prophylaxis for cirrhotic patients with upper gastrointestinal bleeding -...

Meta-analysis: antibiotic prophylaxis for cirrhotic patients withupper gastrointestinal bleeding – an updated Cochrane reviewN. C. Chavez-Tapia*, T. Barrientos-Gutierrez�, F. Tellez-Avila�, K. Soares-Weiser§, N. Mendez-Sanchez*, C. Gluud– &M. Uribe*

*Medica Sur Clinic & Foundation,Mexico City, Mexico.�Tobacco Research Department,National Institute of Public Health,Mexico City, Mexico.�Endoscopy Department, InstitutoNacional de Ciencias Medicas yNutricion Salvador Zubiran, MexicoCity, Mexico.§Enhance Reviews Ltd, London, UK.–Cochrane Hepato-Biliary Group,Copenhagen Trial Unit, Center forClinical Intervention Research,Copenhagen, Denmark.

Correspondence to:Dr N. C. Chavez-Tapia, Obesity andDigestive Diseases Unit, Medica SurClinic & Foundation, Puente de Piedra150, Col. Toriello Guerra, Tlalpan14050, Mexico City, Mexico.E-mail: [email protected]

Publication dataSubmitted 3 March 2011First decision 21 March 2011Resubmitted 19 May 2011Accepted 1 June 2011EV Pub Online 27 June 2011

As part of AP&T’s peer-review process, atechnical check of this meta-analysis

was performed by Dr. P. Collins.

SUMMARY

BackgroundAntibiotic prophylaxis seems to decrease the incidence of bacterial infec-tions in patients with cirrhosis and upper gastrointestinal bleeding and isconsidered standard of care. However, there is no updated informationregarding the effects of this intervention.

AimTo assess the benefits and harms of antibiotic prophylaxis in cirrhoticpatients with gastrointestinal bleeding by performing a systematic review ofrandomised trials.

MethodsWe searched The Cochrane Hepato-Biliary Group Controlled Trials Register,The Cochrane Central Register of Controlled Trials in The Cochrane Library,MEDLINE, EMBASE and Science Citation Index EXPANDED until June2010. We statistically combined data calculating relative risk (RR) for dichot-omous outcomes and mean difference (MD) for continuous outcomes.

ResultsTwelve trials (1241 patients) evaluating antibiotic prophylaxis against pla-cebo or no antibiotic prophylaxis were included. Antibiotic prophylaxis wasassociated with reduced mortality (RR 0.79, 95% CI 0.63–0.98), mortalityfrom bacterial infections (RR 0.43, 95% CI 0.19–0.97), bacterial infections(RR 0.35, 95% CI 0.26–0.47), rebleeding (RR 0.53, 95% CI 0.38–0.74) anddays of hospitalisation (MD )1.91, 95% CI )3.80–0.02). Trials analysingrebleeding rate and hospitalisation length are still scarce, thus, cautionshould be exerted when interpreting the results.

ConclusionsAntibiotic prophylaxis in patients with cirrhosis and upper gastrointestinalbleeding significantly reduced bacterial infections, and reduce all-causemortality, bacterial infection mortality, rebleeding events and hospitalisationlength. Novel clinically significant outcomes were included in this meta-analysis. Some benefits are biased and the risks are not yet properlyassessed, this encourages future research in this field.

Aliment Pharmacol Ther 2011; 34: 509–518

ª 2011 Blackwell Publishing Ltd 509

doi:10.1111/j.1365-2036.2011.04746.x

Alimentary Pharmacology and Therapeutics

INTRODUCTIONAround one-third of cirrhotic patients will experiencebleeding from oesophageal or gastro-oesophageal varices,with 70% recurrence rate and 20% fatality rate.1 Thehighest mortality peak is observed during the first6 weeks after the bleeding episode, being mainly associ-ated with hepatic functional status, renal dysfunctionand bacterial infections.2, 3 Consequently, guidelines fortreatment of patients suffering variceal gastrointestinalbleeding include volume expansion, haemorrhage con-trol, use of vasoconstrictors and short-term antibioticprophylaxis.4

The prophylactic use of oral or intravenous antibioticshas been recommended in several consensus guidelines.The recommended drugs are mainly oral quinolones(norfloxacin 400 mg b.i.d. for 7 days) or intravenouscephalosporins (ceftriaxone 1 g ⁄ day for 7 days).5 A pre-vious systematic review6 assessed the benefits and harmsof antibiotic prophylaxis in patients with cirrhosis andgastrointestinal haemorrhage. Since then, new trials havebecome available providing novel evidence about theeffects of this intervention.

The aim of this systematic review was to assess thebenefits and harms of antibiotic prophylaxis in cirrhoticpatients with upper gastrointestinal bleeding. Specificallythis review was designed to: a) compare all-cause mortal-ity and infection mortality between cirrhotic patientswith gastrointestinal bleeding receiving antibiotic pro-phylaxis or no intervention ⁄ placebo; and b) compare thefrequency of bacterial infections in patients with gastro-intestinal bleeding receiving antibiotic prophylaxis vs. nointervention ⁄ placebo.

METHODS

Eligibility criteriaThis review included randomised clinical trials compar-ing different types of antibiotic therapy against no inter-vention, placebo, or another antibiotic, in theprophylaxis of bacterial infections in cirrhotic patientswith upper gastrointestinal bleeding. Trials were includedirrespective of publication status, language, or blinding.

OutcomesThe primary outcomes considered in this review were:number of deaths (overall and associated with bacterialinfections); and number of patients that developedbacterial infections. Secondary outcomes were: numberof patients who dropped out from the trial after

randomisation; number of patients with rebleeding, andnumber of days of hospitalisation.

Data sources and searchesRelevant randomised trials were identified by searchingThe Cochrane Hepato-Biliary Group Controlled TrialsRegister, the Cochrane Central Register of ControlledTrials (Issue 2, 2010) in The Cochrane Library, MED-LINE (1950 to 21 June 2010), EMBASE (1980 to 21 June2010) and Science Citation Index EXPANDED (1945 to21 June 2010).7 Search strategies and time span of thesearches are provided in Table S1.

The reference list from all identified studies wasinspected for more trials. Moreover, the first or corre-sponding author of each included trial, as well asresearchers active in the field, were contacted for infor-mation regarding unpublished trials and additional infor-mation on their own trial.

Study selection and data collectionThree authors (NC, FT, TB) independently inspectedeach identified reference and applied the inclusion crite-ria. For potentially relevant articles, or in cases of dis-agreement between the reviewers, the full text article wasobtained and inspected independently. If necessary theoriginal authors were contacted and, in the event of noreply, MU or KSW reviewed the article to solve the dis-agreement.

Two authors (NC and TB) independently extractedthe data from included trials. In case of disagreementbetween the two authors, a third author (FT) extractedthe data. Data extraction was discussed, decisions docu-mented, and, when necessary, the authors of the originalstudies were contacted for clarification.

Methodological quality assessmentThe risk of bias was assessed following the instructionsgiven in the Cochrane Handbook for Systematic Reviewsof Interventions.8 The methodological quality of the trialswas assessed based on sequence generation, allocationconcealment, blinding of outcome assessors, incompleteoutcome data and early stopping for benefit.

Data synthesis and analysisRelative risk (RR) and its respective 95% CI were used assummary measure of association for trials with dichoto-mous primary outcomes. Continuous data were analysedcalculating mean differences (MD) and 95% CI betweentrials evaluating antibiotic prophylaxis against no

NN.. CC.. CChhaavveezz--TTaappiiaa eett aall..

510 Aliment Pharmacol Ther 2011; 34: 509–518

ª 2011 Blackwell Publishing Ltd

intervention or placebo, and trials comparing differentantibiotic regimens.

Heterogeneity of effects across trials was evaluated byvisual inspection of the forest plots and v2 and I2 testsfor heterogeneity.8 Statistical heterogeneity was definedas a P value £0.10 (v2) or I2 > 25%. A funnel plotestimating the precision of trials (plot of logarithm ofthe RR against the sample size) was examined to evaluatethe potential for publication bias.

We analysed data using both fixed and random-effectsmodels. When both models produced similar estimates,the fixed-effect result was reported. To examine theinfluence of drop-outs, we performed both worst-best-case (assigning bad outcomes to all of the missingexperimental group patients and good outcomes to all ofthe missing control group patients) and best-worst-case(assigning good outcomes to all of the missing experi-mental group patients and bad outcomes to all the miss-ing control group patients) analyses.

To assess the reliability of the meta-analyses on mor-tality, mortality from bacterial infections and bacterialinfections, the required information size (RIS) was calcu-lated by trial sequential analysis (TSA). We assumed anaverage event proportion of 22%, 5% and 36% in thecontrol group of the three analyses; a 20% relative riskreduction of the experimental intervention and statisticalerror levels of 5% alpha and 20% beta (80% power)respectively. Whenever the cumulative information sizein the meta-analysis was smaller than the RIS, the

threshold to maintain statistical significance was calcu-lated with the O’Brien-Fleming boundaries.9

For the statistical analyses, we used RevMan Analy-ses.10 To compare the RR from different antibioticgroups and detect differences among the antibioticstested vs. no intervention or placebo, a test for inter-action was calculated.11

RESULTS

Study characteristicsSeventeen trials were included (1891 participants) foranalyses12–28; Figure 1 shows the study screening work-flow. Table 1 presents twelve trials comparing antibioticprophylaxis using quinolones (five trials), quinolonesplus beta-lactams (two trials), cephalosporins (three tri-als), carbapenems (one trial) and non-absorbable antibi-otics (one trial) against no intervention or placebo.12–

18, 20, 23–25, 28 Head-to-head antibiotic comparisons wereexplored in five trials (Table 1),19, 21, 22, 26, 27 as follows:combination of antibiotics vs. a single antibiotic,21 twoantibiotics from the same group,19, 22 and differentgroups of antibiotics in each intervention group.26, 27

Effects of interventionsThe effect of antibiotic prophylaxis on all-cause mortalitywas significant (RR 0.79, 95% CI 0.63–0.98) (Figure 2).This association was sensitive to the drop-out rate, asrevealed by the best-worst-case (RR 0.48, 95% CI 0.38–

184 records identified through database searching

19 additional records identified through other sources

55 records after duplicates removed

55 records screened 34 records excluded

21 full-text articles assessed for eligibility

12 studies included in qualitative synthesis

9 full-text articles excluded, with reasons

7 studies included in quantitative synthesis

(meta-analysis)

Iden

tific

atio

nS

cree

ning

Elig

ibili

tyIn

clud

ed

Figure 1 | Study screening flowchart.

MMeettaa--aannaallyyssiiss:: aannttiibbiioottiiccss aanndd cciirrrrhhoottiicc ggaassttrrooiinntteessttiinnaall bblleeeeddiinngg

Aliment Pharmacol Ther 2011; 34: 509–518 511


Table 1 | Characteristics of included trials

Study Country Intervention Control Outcome

Trials comparing with placebo or no intervention

Rimola et al.12 Spain Group Ia – oral gentamicin(200 mg) + vancomycin(500 mg) + nystatin (106 UI) every6 h, until 2 days after cessation ofhaemorrhage.Group Ib – neomycin (1 g) + colistin(1.5 · 106 UI) + nystatin (106 UI)every 6 h, until 2 days saftercessation of haemorrhage.

No antibiotic prophylaxis Bacterial infections

Soriano et al.13 Spain Oral norfloxacin 800 mg ⁄ day during7 days.


Rolando et al.14 UK Intravenous imipenem + cilastin,500 mg before and after thesclerotherapy.

Intravenous dextrose-saline solution

Bacterial infections

Selby et al.16 Australia Intravenous cefotaxime, 1 gimmediately before sclerotherapy.


Blaise et al.15 France Intravenous + oral ofloxacin,400 mg ⁄ day, 10 days;amoxicillin + clavulanic acid(bolus, 1 g) before each endoscopyprocedure.


Pauwels et al.17 France Intravenous + oral ciprofloxacin400 mg ⁄ day, amoxicillin-clavulanicacid 3 g ⁄ day, until 3 days aftercessation of haemorrhage.


Zacharof et al.18 Greece Oral ciprofloxacin 500 mg ⁄ day during7 days.


Hsieh et al.20 Taiwan Oral ciprofloxacin, 1 g ⁄ day, 7 days. Placebo Bacterial infections

Hong et al.23 Korea Intravenous ciprofloxacin 200 mg b.d.for 3 days.


Lin et al.24 Taiwan Intravenous cefazolin 1 g t.d.s. during3 days and then shift to oralcephalexin 500 mg q.d.s. for 4 days.


Hou et al.25 Taiwan Intravenous ofloxacin 200 mg b.d. for2 days followed by oral ofloxacin200 mg b.d. for 5 days.

No antibiotic prophylaxis Rebleeding

Jun et al.28 Korea Intravenous cefotaxime 2 g t.d.s. for7 days.

No antibiotic prophylaxis Rebleeding rate

Trials comparing with another antibiotic regimen

Sabat et al.21 Spain Oral norfloxacin 800 mg ⁄ day, during7 days plus intravenous ceftriaxone2 g ⁄ day the first 3 days.

Oral norfloxacin800 mg ⁄ day, 7 days


Spanish Group19 Spain Oral norfloxacin, 800 mg ⁄ day, 5 days. Oral ofloxacin,400 mg ⁄ day, 5 days


Gulberg et al.22 Germany Intravenous ceftriaxone, 1 g, singledose before TIPS.

Intravenous ceftriaxone,2 g, single dose beforeTIPS


Lata et al.26 CzechRepublic

Intravenous ampicillin ⁄ sulbactam 1.5 gb.d. for 7 days.

Oral or through nasogastrictube norfloxacin 400 mgb.d. for 7 days

Early and latemortality

Fernandez et al.27 Spain Intravenous ceftriaxone 1 g ⁄ day for7 days.

Oral norfloxacin 400 mgb.d. for 7 days





0.60) and worst-best-case analyses (RR 1.45, 95% CI1.04–2.02). The TSA showed a trend towards beneficialeffects of the intervention to reduce mortality, but thecumulative Z-score did not cross the O’Brien-Flemingboundaries (data not shown).

Antibiotic prophylaxis was associated with a signifi-cant decrease in mortality from bacterial infections (RR0.43, 95% CI 0.19–0.97) (Figure 3). The sensitivity analy-sis showed this estimation could have been biased by dif-ferential drop-out rates (worst-best-case analysis RR 3.30,95% CI 1.43–7.62; best-worst-case analysis RR 0.14, 95%CI 0.06–0.31). The TSA demonstrated that the few num-ber of trials included in the analysis were not enough toconclude a beneficial effect of prophylaxis over mortalityfrom bacterial infections (data not shown).

Antibiotic prophylaxis significantly reduced bacterialinfections when all trials with confirmed bacterialinfections were considered (RR 0.35, 95% CI 0.26–0.47)

(Figure 4a). Although heterogeneity of effects across tri-als was detected, effect estimators remained statisticallysignificant with either fixed or random effects models,strengthening the evidence for the proposed effect. Esti-mators for bacterial infections were not affected by thesensitivity analysis (worst-best-case analysis RR 0.70,95% CI 0.57–0.86; best-worst-case analysis RR 0.27, 95%CI 0.17–0.45), also, the TSA showed significant benefit ofantibiotic prophylaxis over no intervention or placebo(Figure 4b).

Considering specific groups of infections, bactera-emia was reported in nine trials with a significant riskreduction in patients under antibiotic prophylaxis (RR0.25, 95% CI 0.15–0.40). Similarly, other infectiousoutcomes were significantly reduced with the use ofantibiotic prophylaxis: pneumonia in nine trials (RR0.45, 95% CI 0.27–0.75), spontaneous bacterial perito-nitis in eight trials (RR 0.29, 95% CI 0.15–0.57) and

AntibioticEvents EventsStudy or Subgroup Total Total

ControlWeight

Risk ratio Risk ratioM-H, Fixed, 95% CI M-H, Fixed, 95% CIYear

Rimola 1985Soriano 1992Rolando 1993Blaise 1994Selby 1994Pauwels 1996Zacharof 1997Hsieh 1998Hong 2002Lin 2002Hou 2004Jun 2006

18 74 23 75 16.3%5.0%

10.0%11.3%3.6%5.1%6.2%

12.9%0.7%2.1%9.1%

17.7%

4 6450

64505919

107

58194035 30

6020506158

6020475962

1011248

1312

1620

714165

138

1813

1324

0.79 [0.47, 1.34]0.57 [0.18, 1.86]0.71 [0.35, 1.45]0.70 [0.36, 1.38]0.40 [0.09, 1.81]0.82 [0.29, 2.38]0.86 [0.37, 2.01]0.72 [0.39, 1.34]

1.00 [0.07, 14.90]

1.27 [0.67, 2.41]0.78 [0.49, 1.25]

0.79 [0.63, 0.98]

0.71 [0.12, 4.06]

198519921993199419941996199719982002200220042006

0.01 0.1 1 10 100Favours antibiotic Favours control

Total (95% CI) 588 653 100.0%Total eventsHeterogeneity: = 3.58, df = 11 (P = 0.98); I22 = 0%Test for overall effect: Z = 2.16 (P = 0.03)

109 145χ

Figure 2 | Forest plot of comparison: antibiotic vs. placebo ⁄no intervention. Outcome: overall mortality.

Rimola 1985Soriano 1992Pauwels 1996Lin 2002Hou 2004Jun 2006


ControlWeight


111022

7

7 20

24223

74 7564 64

10740506158

475962

Total eventsHeterogeneity: Test for overall effect: Z = 2.02 (P = 0.04)

= 2.43, df = 5 (P = 0.79); I 22 = 0%

Total (95% CI) 346 415 100.0% 0.43 [0.19, 0.97]

0.14 [0.02, 1.15]0.50 [0.05, 5.38]0.67 [0.08, 5.80]0.21 [0.01, 4.31]1.03 [0.15, 7.10]0.62 [0.11, 3.60]

198519921996200220042006


37.3%10.7%11.7%13.0%10.6%16.6%

χ

Figure 3 | Forest plot of comparison: antibiotic vs. placebo ⁄no intervention. Outcome: mortality due to bacterial infec-tions.




urinary tract infections in nine trials (RR 0.23, 95% CI0.13–0.41).

All antibiotics tested provided a beneficial effect overbacterial infections, although the protective effect wasstronger with cephalosporins (RR 0.16, 95% CI 0.05–0.48),quinolones (RR 0.27, 95% CI 0.18–0.39), and quinolonesplus beta-lactams (RR 0.38, 95% CI 0.23–0.62), than withother antibiotics (RR 0.57, 95% CI 0.41–0.81), compared

with placebo or no intervention (Figure 5). However, thetest for interaction demonstrated that only the group of‘other antibiotics’ significantly differed from all otherdrugs. No significant difference between quinolones andcephalosporins was observed. In the head-to-headantibiotic comparison trials no significant differencesbetween regimens were observed for the outcomes understudy (Table 2).

Rimola 1985Soriano 1992Rolando 1993

Blaise 1994Selby 1994

Pauwels 1996Zacharof 1997Hsieh 1998

Hong 2002Lin 2002

Hou 2004Jun 2006

Total (95% CI)Total eventsHeterogeneity: τTest for overall effect: Z = 6.82 (P < 0.00001)

= 0.08; 2 χ = 16.17, df = 11 (P = 0.14); I 22 = 32%75 233

574 607 100.0% 0.35 [0.26, 0.47]

0.21 [0.05, 0.92]0.13 [0.03, 0.54]0.22 [0.05, 0.90]0.08 [0.00, 1.41]0.22 [0.10, 0.50]0.21 [0.08, 0.57]0.34 [0.13, 0.86]0.31 [0.16, 0.59]0.17 [0.02, 1.26]0.65 [0.41, 1.05]0.43 [0.24, 0.77]0.51 [0.31, 0.84] 1985

19921993199419941996199719982002200220042006

16 74 7564 6450 5019 1958 592635 30

6050206158

3228266

302816

61

2769

169

6047205962

1217

194460222

16.3%14.3%

3.7%

2.1%

9.6%

12.6%7.7%7.3%

1.1%4.1%4.0%

17.4%



ControlWeight

Risk ratio Risk ratioM-H, Random, 95% CI M-H, Random, 95% CIYear

Number ofpatients

(linear scaled)

Fav

ors

antib

iotic

sF

avor

spl

aceb

o/no

inte

rven

tion

–8

–7

–6

–5

–4

–3

–2

–1

8

7

6

5

4

3

2

11241

TSMB Pc 36%; RR 20%; a 5%, b20% is a Two-sided graph

TSMB Pc 36%; RR 20%; a 5%, b20% = 1980

CumulativeZ-Score

Z-curve

(a)

(b)

Figure 4 | Forest plot of comparison: antibiotic vs. placebo ⁄no intervention. Outcome: bacterial infections (Panel a).O’Brien-Fleming monitoring boundaries for assessing statistical significance for bacterial infections (Panel b). The solidcurve presents the cumulative meta-analysis test-score and the inward sloping curves present the adjusted thresholdfor statistical significance – the two-sided O’Brien-Fleming boundaries.




A significant reduction in overall rebleeding episodeswas observed among patients under antibiotic prophy-laxis (RR 0.53, 95% CI 0.38–0.74), also, rebleeding afterup to 7 days of follow-up was significantly reduced (RR0.24, 95% CI 0.12–0.50).

Trials evaluating hospitalisation length showedpatients receiving antibiotic prophylaxis tended to haveshorter hospital stays (MD )1.91 days, 95% CI )3.80 to)0.02), however, this finding was not replicated in lengthof stay in the intensive care unit (MD )0.27 days, 95%CI )1.55–1.00).

Quality of the evidenceThis review included 1891 cirrhotic patients withupper gastrointestinal bleeding; 1241 of them partici-

pated in randomised trials comparing antibiotic pro-phylaxis vs. no intervention or placebo, and theremaining 650 participants in trials comparing differentantibiotic prophylactic regimens. All trials presentedmethodological weaknesses and should be consideredat risk of bias. Lack of blinding and lack of propersample size calculations were the most commonsources of bias (Table 3). Despite that the primary andsecondary outcomes were included in the final analysisit is clear that the absence of mortality for bacterialinfections, a clear and available outcome in the trialsincluded could be considered an important source ofbias.

The manuscripts included did not report adverseevents, including resistance patterns.


ControlWeight


1.20.1 Quinolones

1.20.2 Quinolones + beta-lactams

1.20.3 Cephalosporins

1.20.4 Others

Soriano 1992

Hsieh 1998Hong 2002Hou 2004

Zacharof 1997

Subtotal (95% CI)

Subtotal (95% CI)

Subtotal (95% CI)

Subtotal (95% CI)

12 6435602059

6430602061

235238

2816279

16

4622

12.5%7.7%

12.1%4.0%7.0%

7.8%21.1%

13.3%

43.4%26 96

1992

19941996

1994

19851993

20022006

1997199820022004

0.43 [0.24, 0.77]0.21 [0.08, 0.57]0.22 [0.10, 0.50]0.22 [0.05, 0.90]0.13 [0.03, 0.54]0.27 [0.18, 0.39]

0.31 [0.16, 0.59]0.50 [0.23, 1.11]0.38 [0.23, 0.62]

0.17 [0.02, 1.26]0.08 [0.00, 1.41]0.21 [0.05, 0.92]0.16 [0.05, 0.48]

0.51 [0.31, 0.84]0.65 [0.41, 1.05]0.57 [0.41, 0.81]

0.36 [0.29, 0.45]

Blaise 1994

Selby 1994Lin 2002Jun 2006

Rimola 1985Rolando 1993

Pauwels 1996

Total (95% CI)Total events

Total events

Total events

Total events

Total events

Test for overall effect: Z = 3.26 (P = 0.001)



Test for overall effect: Z = 6.65 (P < 0.00001)

Test for overall effect: Z = 8.92 (P < 0.00001) Heterogeneity: χ = 16.30, df = 11 (P = 0.13); I 22

χ2

= 33%

Heterogeneity: χ = 0.33, df = 2 (P = 0.85); I22 = 0%

Heterogeneity: χ = 0.53, df = 1 (P = 0.47); I 22 = 0%



Test for subgroup differences: = 11.01, df = 3 (P = 0.01), I2 = 72.8%

77 237


588 653 100.0%

96

15 62

323058

4098

59107166

2.7%2.8%4.2%9.7%

19194762

128

5058

127

669

102

3 21

1617

33 58

74 3226

755050

124 125

14.2%11.6%25.9%

Figure 5 | Forest plot of comparison: antibiotic vs. placebo ⁄no intervention. Outcome: bacterial infections according tothe antibiotic used.




DISCUSSIONThis review aimed to evaluate the effects of antibiotic pro-phylaxis over bacterial infections and mortality in cirrhoticpatients with gastrointestinal bleeding. A significant reduc-tion in bacterial infections was observed in patients receiv-ing antibiotic prophylaxis as reported previously.29 Also,although still not yet overwhelming, antibiotic prophylaxiswas associated with reduced all-cause mortality, bacterialinfection mortality, incidence of rebleeding events andlength of hospitalisation.

This systematic review is an update of a systematicreview and meta-analysis published in 2002.6 The inclu-sion of new trials did not modify a previously observedbeneficial effect of antibiotic prophylaxis on all-causemortality and bacterial infections. New clinically relevantoutcomes such as rebleeding and hospitalisation lengthwere also included. Both rebleeding episodes and hospi-talisation length were significantly reduced with anti-biotic prophylaxis, although no association was observedwith the time spent in critical care.

The evolution of the antibiotic prophylaxis goes fromnon-absorbable antibiotics, to quinolones, and, morerecently, to cephalosporins. In the most recent update

from the Baveno V scientific committee suggests as anarea of further study the best antibiotic treatment.30

Studies comparing antibiotic regimens against placebo orno intervention provided no solid evidence to prefer oneantibiotic regimen over the other. A similar finding wasobserved in trials conducting head to head antibioticcomparisons. In general prophylaxis benefits wereobserved indistinctly of the antibiotic used, therefore, nospecific regimen can be recommended over another, andantibiotic selection should be made considering localconditions such as bacterial resistance profile and treat-ment cost. Use of quinolones was first explored bySoriano et al.13 and quinolones have been broadlyused since then, despite rising concerns of a potentialreduction of their effects due to bacterial resistance.However, considering that bacterial resistance patternvary by location, use of quinolones for antibiotic prophy-laxis will have to be assessed in specific local settings.The available information does not allow establishingany conclusion about the best option for antibioticprophylaxis.

The current evidence to support antibiotic prophylaxisis based on twelve randomised trials and, except for

Table 2 | Data on trials comparing head-to-head two different regimens of antibiotic prophylaxis

Author Outcome Experimental (n ⁄N) Control (n ⁄N) Relative risk 95% CI

Sabat et al.21 Mortality Norfloxacin + ceftriaxone(1 ⁄ 24)

Norfloxacin (2 ⁄22) 0.46 0.04–4.71

Sabat et al.21 Bacterial infections Norfloxacin + ceftriaxone(3 ⁄24)

Norfloxacin (4 ⁄22) 0.69 0.17–2.73

Sabat et al.21 Cost Norfloxacin + ceftriaxone –US $99.3–220.1

Norfloxacin – US $1.9–745.5 – –

Sabat et al.21 Drop-outs Norfloxacin + ceftriaxone(4 ⁄28)

Norfloxacin (6 ⁄28) 0.67 0.21–2.11

Gulberg et al.22 Mortality Ceftriaxone 1 g (0 ⁄40) Ceftriaxone 2 g (0 ⁄42) Riskdifference:0.00

)0.05–0.05

Gulberg et al.22 Bacterial infections Ceftriaxone 1 g (1 ⁄40) Ceftriaxone 2 g (1 ⁄42) 1.05 0.11–9.80

Spanish Group19 Bacterial infections(proven)

Norfloxacin 800 mg(26 ⁄ 183)

Ofloxacin 400 mg (27 ⁄ 182) 0.96 0.58–1.58

Spanish Group19 Bacterial infections(suspected)

Norfloxacin 800 mg(51 ⁄ 183)

Ofloxacin 400 mg (53 ⁄ 182) 0.96 0.69–1.32

Lata et al.26 Mortality Ampicillin and sulbactam3 g (12 ⁄21)

Norfloxacin 800 mg (7 ⁄25) 2.04 0.98–4.23

Fernandez et al.27 Mortality Ceftriaxone 1 g (8 ⁄54) Norfloxacin 800 mg (6 ⁄57) 1.41 0.52–3.79

Fernandez et al.27 Mortality frombacterial infections

Ceftriaxone 1 g (1 ⁄54) Norfloxacin 800 mg (1 ⁄57) 1.06 0.07–16.46

Fernandez et al.27 Bacterial infections Ceftriaxone 1 g (6 ⁄54) Norfloxacin 800 mg (5 ⁄57) 1.27 0.41–3.94




America and Africa, the intervention has been assessedin heterogeneous populations, providing external validityto this review. The effects observed were more robust forprevention of bacterial infections than for mortalityreduction, which could be explained by the fact that alltrials included were designed and powered to evaluatethis outcome. The bias associated with the lack of infor-mation regarding mortality for bacterial infections is nota minor issue. This bias limits the strength of the conclu-sion about the beneficial effects of antibiotics in this out-come, and in consequence it is not clear exactly how thebeneficial effect on overall mortality is reached. The logi-cal argument indicates that the most important benefitfrom an antibiotic intervention must be related withtheir effects on infections. However, the indirect effectsof antibiotics31 could influence the overall mortality, butthis has not been properly addressed.

Trials analysing rebleeding rate and hospitalisationlength are still scarce, thus, caution should be exertedwhen interpreting the results.

In conclusion, this up-dated systematic review and meta-analysis, enhances the beneficial effects of antibiotic prophy-laxis in patients with cirrhosis and gastrointestinal bleeding.Novel, clinically significant outcomes were included amongthe benefits from antibiotic prophylaxis. This increases thesize of treatment effect, promoting this intervention as the

standard of care. However, some benefits remains unclearand the risks are not yet properly assessed.

ACKNOWLEDGEMENTSWe would like to thank M Brezis, R Tur-Kaspa, and LLeibovici who participated in the previous version of thisreview. We thank Dimitrinka Nikolova and Sarah LouiseKlingenberg of The Cochrane Hepato-Biliary Group forongoing support for this review. Thanks to J. Fernandez,P. Gines, A. Rimola, N. Rolando and A. Zacharof, whokindly supplied additional information on their own trials.Declaration of personal interests: All the authors wereinvolved in the analysis and interpretation of data, draftwriting and in approving the final version. Declaration offunding interests: The preparation of this paper wasfunded in part by an educational grant from Medica SurClinic & Foundation.

SUPPORTING INFORMATIONAdditional Supporting Information may be found in theonline version of this article:

Table S1. Search strategy and time span.Please note: Wiley-Blackwell are not responsible for the

content or functionality of any supporting materials suppliedby authors. Any queries (other than missing material) shouldbe directed to the corresponding author for the article.

Table 3 | Risk of bias summary, review authors’ judgments about each risk of bias item for each included study

AuthorSequencegeneration

Allocationconcealment Blinding

Incompleteoutcome

Selectivereporting

Otherbias

Intentionto treat

Samplecalculation

Rimola et al.12 + + ) ) + ) ) )

Soriano et al.13 + ? ) + + + ) )

Rolando et al.14 + + ) + + ? ) )

Selby et al.16 + + ) + + ? ) )

Pauwels et al.17 + ? ) + + + ) )

Zacharof et al.18 + + ) + ? ? ? ?

Hsieh et al.20 ? ? ? + + + + )

Sabat et al.21 ? ? ) + + ) ) ?

Spanish Group19 ? ? ? ? ? ? ? ?

Gulberg et al.22 ? ? ) + + ) ? )

Hong et al.23 ? ? ) ? + ? ? ?

Lin et al.24 ? ? ) + + + ) )

Blaise et al.15 ? ? ) + + + ) )

Hou et al.25 + ? ) + + + ) +

Lata et al.26 + ? ) + + + ? )

Fernandez et al.27 + ? ) + + + ) +

Jun et al.28 + + ) ) + + ) +




REFERENCES1. The North Italian Endoscopic Club for

the Study and Treatment of EsophagealVarices. Prediction of the first varicealhemorrhage in patients with cirrhosis ofthe liver and esophageal varices. A pro-spective multicenter study. N Engl J Med1988; 319: 983–9.

2. Burroughs AK, Triantos CK, O’Beirne J,Patch D. Predictors of early rebleedingand mortality after acute variceal hemor-rhage in patients with cirrhosis. Nat ClinPract Gastroenterol Hepatol 2009; 6:72–3.

3. Augustin S, Muntaner L, Altamirano JT,et al. Predicting early mortality afteracute variceal hemorrhage based on clas-sification and regression tree analysis.Clin Gastroenterol Hepatol 2009; 7:1347–54.

4. Garcia-Tsao G, Sanyal AJ, Grace ND,Carey W. Prevention and managementof gastroesophageal varices and varicealhemorrhage in cirrhosis. Hepatology2007; 46: 922–38.

5. Garcia-Tsao G, Lim JK. Managementand treatment of patients with cirrhosisand portal hypertension: recommenda-tions from the Department of VeteransAffairs Hepatitis C Resource Center Pro-gram and the National Hepatitis C Pro-gram. Am J Gastroenterol 2009; 104:1802–29.

6. Soares-Weiser K, Brezis M, Tur-KaspaR, Leibovici L. Antibiotic prophylaxis forcirrhotic patients with gastrointestinalbleeding. Cochrane Database Syst Rev2002; 9: CD002907.

7. Royle P, Milne R. Literature searchingfor randomized controlled trials used inCochrane reviews: rapid versus exhaus-tive searches. Int J Technol Assess HealthCare 2003; 19: 591–603.

8. Higgins J, Green S. Cochrane Handbookfor Systematic Reviews of InterventionsVersion 5.0.2. West Sussex, UK: TheCochrane Collaboration, 2009.

9. Thorlund K, Devereaux PJ, Wetterslev J,et al. Can trial sequential monitoringboundaries reduce spurious inferencesfrom meta-analyses? Int J Epidemiol2009; 38: 276–86.

10. Centre TNC. Review Manager (RevMan).5th ed. Copenhagen: The Cochrane Col-laboration, 2010.

11. Altman DG, Bland JM. Interaction revis-ited: the difference between two esti-mates. BMJ 2003; 326: 219.

12. Rimola A, Bory F, Teres J, Perez-AyusoRM, Arroyo V, Rodes J. Oral, nonab-sorbable antibiotics prevent infection incirrhotics with gastrointestinal hemor-rhage. Hepatology 1985; 5: 463–7.

13. Soriano G, Guarner C, Tomas A, et al.Norfloxacin prevents bacterial infectionin cirrhotics with gastrointestinal hemor-rhage. Gastroenterology 1992; 103:1267–72.

14. Rolando N, Gimson A, Philpott-HowardJ, et al. Infectious sequelae after endo-scopic sclerotherapy of oesophageal vari-ces: role of antibiotic prophylaxis. JHepatol 1993; 18: 290–4.

15. Blaise M, Pateron D, Trinchet JC, Lev-acher S, Beaugrand M, Pourriat JL. Sys-temic antibiotic therapy preventsbacterial infection in cirrhotic patientswith gastrointestinal hemorrhage. Hepa-tology 1994; 20(1 Pt 1): 34–8.

16. Selby WS, Norton ID, Pokorny CS, BennRA. Bacteremia and bacterascites afterendoscopic sclerotherapy for bleedingesophageal varices and prevention byintravenous cefotaxime: a randomizedtrial. Gastrointest Endosc 1994; 40: 680–4.

17. Pauwels A, Mostefa-Kara N, Debenes B,Degoutte E, Levy VG. Systemic antibi-otic prophylaxis after gastrointestinalhemorrhage in cirrhotic patients with ahigh risk of infection. Hepatology 1996;24: 802–6.

18. Zacharof A, Petrogiannopoulos C, Sou-tos D, Katsaros D, Zacharof H. Bacterialinfection is prevented by ciprofloxacin incirrhotics with gastrointestinal hemor-rhage. Gut 1997; 41(Suppl 3): A189.

19. Spanish Group for the Study of BacterialInfections in Cirrhosis. Norfloxacin ver-sus ofloxacin in the prophylaxis of infec-tion in cirrhotic patients withgastrointestinal hemorrhage. J Hepatol1998; 28(Suppl 1): S80.

20. Hsieh WJ, Lin HC, Hwang SJ, et al. Theeffect of ciprofloxacin in the preventionof bacterial infection in patients withcirrhosis after upper gastrointestinalbleeding. Am J Gastroenterol 1998; 93:962–6.

21. Sabat M, Kolle L, Soriano G, et al. Par-enteral antibiotic prophylaxis of bacterialinfections does not improve cost-efficacyof oral norfloxacin in cirrhotic patientswith gastrointestinal bleeding. Am JGastroenterol 1998; 93: 2457–62.

22. Gulberg V, Deibert P, Ochs A, Rossle M,Gerbes AL. Prevention of infectiouscomplications after transjugular intrahe-patic portosystemic shunt in cirrhoticpatients with a single dose of ceftriaxone.Hepatogastroenterology 1999; 46:1126–30.

23. Hong SN, Kim BJ, Lee SY, et al. Pro-spective randomized trial of intravenousciprofloxacin for prevention of bacterialinfection in cirrhotic patients withesophageal variceal bleeding. TaehanKan Hakhoe Chi 2002; 8: 288–96.

24. Lin YT, Lo GH, Lai KH, Chen TA, LinWJ. Prophylactic antibiotics in cirrhoticswith upper gastrointestinal hemorrhage:a prospective, controlled trial. ZhonghuaYi Xue Za Zhi (Taipei) 2002; 65: 365–71.

25. Hou MC, Lin HC, Liu TT, et al. Antibi-otic prophylaxis after endoscopic therapyprevents rebleeding in acute varicealhemorrhage: a randomized trial. Hepatol-ogy 2004; 39: 746–53.

26. Lata J, Jurankova J, Husova L, et al.Variceal bleeding in portal hypertension:bacterial infection and comparison ofefficacy of intravenous and per-oralapplication of antibiotics–a randomizedtrial. Eur J Gastroenterol Hepatol 2005;17: 1105–10.

27. Fernandez J, Ruiz del Arbol L, Gomez C,et al. Norfloxacin vs ceftriaxone in theprophylaxis of infections in patients withadvanced cirrhosis and hemorrhage. Gas-troenterology 2006; 131: 1049–56.

28. Jun CH, Park CH, Lee WS, et al. Antibi-otic prophylaxis using third generationcephalosporins can reduce the risk ofearly rebleeding in the first acute gastro-esophageal variceal hemorrhage: a pro-spective randomized study. J KoreanMed Sci 2006; 21: 883–90.

29. Bernard B, Grange JD, Khac EN, AmiotX, Opolon P, Poynard T. Antibiotic pro-phylaxis for the prevention of bacterialinfections in cirrhotic patients with gas-trointestinal bleeding: a meta-analysis.Hepatology 1999; 29: 1655–61.

30. de Franchis R. Revising consensus inportal hypertension: report of the BavenoV consensus workshop on methodologyof diagnosis and therapy in portal hyper-tension. J Hepatol 2010; 53: 762–8.

31. Hamilton-Miller JM. Immunopharmacol-ogy of antibiotics: direct and indirect im-munomodulation of defence mechanisms.J Chemother 2001; 13: 107–11.




Meta-analysis: antibiotic prophylaxis for cirrhotic patients with upper gastrointestinal bleeding -...

Documents

Transcript of Meta-analysis: antibiotic prophylaxis for cirrhotic patients with upper gastrointestinal bleeding -...