nature publishing group

CLINICAL AND SYSTEMATIC REVIEWS

Khurram J. Khan, MD1, Thomas A. Ullman, MD, FACG2, Alexander C. Ford, MD, MRCP3, Maria T. Abreu, MD, FACG4, A. Abadir, MD1, John K. Marshall, MD, MSc, FRCPC1, Nicholas J. Talley, MD, PhD, FACG5 and Paul Moayyedi, MB, ChB, BSc, MPH, PhD, FACG1

The etiology of inammatory bowel disease (IBD) is unknown but may relate to an unidentied bacterial pathogen or an immunological reaction to gut microbiota. Antibiotics have therefore been proposed as a therapy for Crohns disease (CD) and ulcerative colitis (UC) to induce remission in active disease to prevent relapse. Current data are conicting and we therefore conducted a systematic review of randomized controlled trials (RCTs) evaluating antibiotics in IBD. Only parallel group RCTs were considered eligible. Studies with adult patients receiving any dose of therapy for at least 7 days and up to 16 weeks for active disease, or at least 6 months of follow-up for preventing relapse in quiescent disease were analyzed. We included any antibiotics alone or in combination using predened denitions of remission and relapse. Two reviewers independently assessed eligibility and extracted data. The primary outcome was remission or relapse using an intention-to-treat methodology. The data were summarized using relative risk (RR) and pooled using a random effects model. For active CD, there were 10 RCTs involving 1,160 patients. There was a statistically signicant effect of antibiotics being superior to placebo (RR of active CD not in remission = 0.85; 95% condence interval (CI) = 0.730.99, P = 0.03). There was moderate heterogeneity between results (I2 = 48%) and a diverse number of antibiotics were tested (anti-tuberculosis therapy, macrolides, uroquinolones, 5-nitroimidazoles, and rifaximin) either alone or in combination. Rifamycin derivatives either alone or in combination with other antibiotics appeared to have a signicant effect at inducing remission in active CD. In perianal CD stula there were three trials evaluating 123 patients using either ciprooxacin or metronidazole. There was a statistically signicant effect in reducing stula drainage (RR = 0.8; 95% CI = 0.660.98) with no heterogeneity (I2 = 0%) and an number needed to treat 5 (95% CI = 320). For quiescent CD, there were 3 RCTs involving 186 patients treated with different antibiotics combinations (all including antimycobacterials) vs. placebo. There was a statistically signicant effect in favor of antibiotics vs. placebo (RR of relapse = 0.62; 95% CI = 0.460.84), with no heterogeneity (I2 = 0%). In active UC, there were 9 RCTs with 662 patients and there was a statistically signicant benet for antibiotics inducing remission (RR of UC not in remission = 0.64; 95% CI = 0.430.96). There was moderate heterogeneity (I2 = 69%) and antibiotics used were all different single or combination drugs. Antibiotic therapy may induce remission in active CD and UC, although the diverse number of antibiotics tested means the data are difcult to interpret. This systematic review is a mandate for further trials of antibiotic therapy in IBD.
Am J Gastroenterol advance online publication, 15 March 2011; doi:10.1038/ajg.2011.72

INTRODUCTION
Crohns disease (CD) and ulcerative colitis (UC) were first described as discrete clinical entities when physicians could exclude infections as causative agents (1). Still these early pioneers thought the etiology was likely to be infectious but they could not isolate the offending organisms (1). Indeed, a Scottish surgeon, Kennedy Dalziel, is credited with the first thorough description consistent with CD in 1913 (2) and he drew parallels with Johnes disease in cattlea disease now known to be due to Mycobacterium avium subspecies Paratuberculosis (MAP). Almost 100 years later, modern science is unable to resolve the controversy with arguments for and against

the MAP hypothesis (3). MAP is not the only organism that has been implicated in CD with other mycobacteria, Yersinia enterocolitica, Chlamydia Trachomatis, listeria, and cell wall-deficient pseudomonas being postulated as the cause of CD. There is a similar long list of infectious candidates for UC (1). Evidence for any of these infectious agents as a cause for inflammatory bowel disease (IBD) is poor, but CD and UC are still believed to result from an aberrant immunological reaction to gut microbiota in a susceptible host (4). The importance of the gut microbiome is highlighted by experimental models of colitis in rats demonstrating that bowel inflammation does not occur without gut bacteria (5).

McMaster University Medical Centre, Hamilton, Ontario, Canada; 2The Mount Sinai School of Medicine, New York, New York, USA; 3Leeds Gastroenterology Institute, Leeds General Inrmary, Leeds, UK; 4Miller School of Medicine, University of Miami, Miami, Florida, USA; 5Faculty of Health, University of Newcastle, Callaghan New South Wales, Australia. Correspondence: Paul Moayyedi, MB, ChB, BSc, MPH, PhD, FACG, McMaster University Medical Centre, Room 4W8E, 1200 Main Street West, Hamilton, Ontario, Canada. L8N 3Z5, E-mail: moayyep@mcmaster.ca Received 1 December 2010; accepted 13 February 2011
2011 by the American College of Gastroenterology

The American Journal of GASTROENTEROLOGY

REVIEW

Antibiotic Therapy in Inammatory Bowel Disease: A Systematic Review and Meta-Analysis

Khan et al.

Box 1. Eligibility criteria

REVIEW

Randomized controlled trials.

Adults ( > 90% of patients aged > 16 years) with Crohns disease (CD) or ulcerative colitis (UC). Compared antibiotic therapy with placebo or no therapy. Other inammatory bowel disease therapy allowed provided this was the same in both arms of the trial. Minimum duration of therapy of 7 days in trials reporting induction of remission of active CD and UC with maximum time point to assess remission of 17 weeks. Did not specically evaluate post-operative CD patients. Minimum follow-up of 6 months in trials reporting prevention of relapse of quiescent CD and UC. Assessment of failure of remission in active CD/UC, or relapse of disease activity in quiescent CD/UC, at last time point of assessment in the trial according to a predened hierarchy (Box 2).

Despite the likely importance of gut flora in the pathogenesis of IBD, therapy has focused on suppressing the immune system rather than removing the agent that might be responsible for the aberrant response. Antibiotics have been evaluated in both CD and UC but results have been conflicting (6,7). It is important to systematically evaluate the evidence for antibiotics in IBD as this can clarify their role in inducing remission and preventing relapse. A systematic review may highlight which antibiotic class is more effective in treating either CD or UC and evaluating antimicrobial therapy overall may give an indication as to whether this is an area that needs to be explored further. There have been previous systematic reviews on antimicrobial therapy in CD (810) and UC (11), but these have focused on one antibiotic group, only evaluated specific types of disease (e.g., specifically looked at either relapse or inducing remission) and in some case need updating (8,9,11). We therefore conducted an updated, high quality, and comprehensive systematic review of all RCTs evaluating antibiotics for induction and maintenance in IBD, looking separately at CD and UC.

additional studies. Experts in the field were contacted to try to identify other studies.
Study selection and extraction

METHODS
Search strategy

An electronic search was conducted using MEDLINE (1966 December 2010), EMBASE (1984December 2010), Cochrane Central Register of Controlled Trials (Issue 4. December 2010), and Cochrane Inflammatory Bowel Disease Group Specialized Trials Register. The search strategy was not limited by language. The search was completed for all therapies in inflammatory bowel disease as part of a larger IBD review. Search terms (both free text and medical subject headings) used specific to the antibiotic and IBD systematic review include antibiotics, antimicrobial agents, antiparasitic agents, antimycobacterials agents, acetamides, penicillins, ketolides, naphthyridines, thienamycins, carbapenems, tetracyclines, polymixins, oxazolidines, glycopeptides, cephalosporins, aminoglycosides, nitroimidazoles, macrolides, fluroquinolones, -lactams, colitis, IBD, ileitis, regional enteritis, CD, and UC. Study citations were collected on a single database. All relevant abstracts and papers were retrieved from the initial screening of titles for better assessment of eligibility. A recursive search of the bibliography of all relevant trials and papers identified by the electronic search strategy was then conducted. Abstracts from DDW 20022010 and UEGW 20022010 were hand searched for
The American Journal of GASTROENTEROLOGY

Details of eligibility criteria are given in Box 1. Only parallel group randomized controlled trials or the first phase of crossover randomized trials that evaluated adult patients ( > 90% over 16 years) with IBD were eligible for inclusion in the review. The following interventions were considered: antimycobacterial drugs, traditional broad-spectrum antibiotics, antimycobacterial and antibiotics in combination. These interventions were compared with either placebo or no intervention. The minimum duration of therapy that was eligible was 10 days. Trials evaluating therapy in active CD or UC were included if they reported evidence of remission as an outcome. Our primary outcome measures were predefined according to a hierarchy for both UC and CD and are given in Box 2. Two independent reviewers evaluated each trial identified by the search. Inclusion decisions for each paper were made independently according to the pre-stated eligibility criteria (Box 1) on a standard eligibility form, and where disagreements occurred, the opinion of a third adjudicator was obtained. The data were then abstracted independently by two reviewers and collected on standard data extraction forms. Information collected included baseline demographics, patients disease history and activity, details about the experimental and control therapies, concomitant therapy, discontinued medication, adverse events, and number of patients allocated and outcomes using an intention-to-treat model (drop outs were assumed to be treatment failures).
Bias assessment

Two investigators independently assessed all studies selected for inclusion in the review for risk of bias. Where disagreements occurred, the opinion of a third investigator was obtained. Seven components of risk of bias as described in the Cochrane handbook (12) were assessed using the following six criteria: generation of the treatment allocations (truly random, pseudo-random, non-random, not stated), concealment of the treatment allocations at randomization (concealed, unconcealed, not stated), implementation of masking (patient masked/unmasked/unclear, outcome assessment masked/unmasked/unclear, clinician masked/unmasked/unclear), completeness of follow-up and use of intention-to-treat analysis, selective reporting of outcomes,
VOLUME 104 | XXX 2010 www.amjgastro.com

Antibiotic Therapy in IBD

Box 2. Data extraction methodology

Hierarchy of reporting of outcomes used: Luminal CD remission: Crohns disease activity index (CDAI) < 150 (or other validated index), endoscopic evidence of complete remission (most stringent denition available, e.g., complete mucosal healing), clinical assessment of complete remission, other author-dened criteria for remission. Luminal CD relapse: CDAI150, endoscopic/radiological evidence of relapse (most stringent denition available), other CDAI cutoff, clinical assessment as relapsed, other author-dened criteria for relapse. UC remission: endoscopic evidence of complete remission (most stringent denition available, e.g., complete mucosal healing), clinical assessment as complete remission, recognized scoring system of complete remission (e.g., Truelove and Witt), other author-dened criteria for remission. UC relapse: endoscopic evidence of relapse, clinical assessment of symptomatic relapse, recognized scoring system dening relapse, other author-dened criteria for relapse. Time of outcome assessment: last time point of assessment in the trial. Denominator used: true intention-to-treat analysis, if not available then all evaluable patients.

and other threats to validity (baseline imbalances between randomized groups, early stopping of the trial, primary end point in the study appears convoluted and is not intuitive or standard).
Statistical methods

Studies identified in search (n = 3,062)

For trials that evaluated treatment of active disease, outcomes were expressed in terms of relative risk (RR) of active disease persisting (with 95% confidence intervals (CIs)) as the summary outcome statistic, whereas trials that assessed prevention or relapse were expressed as RR of relapse. Adverse effects were summarized as RRs of overall adverse event in antibiotic vs. placebo arms. Data were synthesized using a random effects model (13). Heterogeneity will be assessed using 2 statistic and I2 (14). A value of I2 > 25% was considered to indicate heterogeneity and the reasons for this were explored using the following predefined subgroup analyses: dosage, duration of therapy, compliance, surgical resection of bowel, duration of disease, proportion with new onset disease, use of adjunct therapies, inclusion of previous treatment failures, geographical differences (North America, Latin America, South America, Australasia, Asia, Africa, Southern Europe, and Northern Europe), and high risk of bias vs. unclear vs. low risk of bias trials. Funnel plots were produced for the principal outcome for each comparison, and the Eggers test (15) of funnel plot asymmetry was used to investigate whether publication or other small study effects may have influenced the results.

Studies excluded (title and/or abstract not appropriate; n = 3,028)

Studies retrieved for evaluation (n = 34)

Excluded (n = 11) - Outcome not of interest (n = 3) - Wrong assessment period (n = 2) - Post-operative patients (n = 2) - Other reason (n = 4)

Studies eligible for inclusion (n = 23) - Treatment of Crohns (n = 14), UC (n = 9) (Active Crohns only n = 14, Quiescent Crohns n = 3)
Figure 1. Flow chart of search. UC, ulcerative colitis .

RESULTS
The initial search strategy for all therapies in IBD yielded 3,062 results. Of those, 3,028 were excluded after screening titles and abstracts for antibiotic therapies relevant to IBD management. Thirty-four papers were retrieved in full text and of those, 23 (6,7,1636) met the eligibility criteria for induction and/or maintenance of IBD (Figure 1). Eleven papers were excluded: three trials (37,3940) only gave the outcome as symptom improvement or a change in disease score, two trials (41,42) assessed remission at too late a time point, two trials (43,44) evaluated post-operative CD patients only, one trial (45) used a crossover design, one trial (46) evaluated an anti-protozoa therapy, one (47) compared antibiotic
2011 by the American College of Gastroenterology

therapy with another active treatment to induce remission, and one (38) gave therapy for < 7 days (Figure 1). Overall, the quality of the studies was generally poor with only four low risk of bias (22,26,28,31) trials. There were 18 unclear risk of bias trials (6,7,1621,2325,27,29,3136) and one high risk of bias trial (30) (Table 1). The main reasons for unclear risk of bias was that the papers did not report method of randomization or method of concealment of allocation.
Antibiotics for induction of remission of CD

We identified 10 RCTs (1625) involving 1,160 patients that evaluated the efficacy of antibiotic therapy at inducing remission in CD.
The American Journal of GASTROENTEROLOGY

REVIEW

Outcome of interest: failure of remission in active luminal Crohns disease (CD) or active ulcerative colitis (UC), relapse of disease activity in quiescent luminal CD or UC.

Khan et al.

Table 1. Quality of studies included in the review

REVIEW

Reference (16) (17) (18) (19) (20) (21) (22) (23) (24) (25) (27) (26) (28) (29) (30) (31) (6) (32) (7) (33) (34) (35)

Randomized Unclear Unclear Adequate Unclear Unclear Adequate Adequate Unclear Unclear Unclear Unclear Adequate Adequate Unclear Unclear Adequate Unclear Unclear Adequate Unclear Unclear Unclear

Conceal allocation Unclear Unclear Unclear Unclear Adequate Unclear Adequate Unclear Unclear Unclear Adequate Adequate Adequate Unclear Unclear Adequate Unclear Unclear Unclear Unclear Adequate Unclear

Blinding Double Double Double Double Double Double Double Double Double Double Double Double Double Unclear Not blind Double Double Double Double Double Double Double

Baseline table Present Present Present Absent Present Present Present Present Present Absent Absent Present Present Present Present Present Present Present Present Present Present Absent

Outcome assess Reported Reported Reported Not eported Reported Reported Reported Reported Reported Not eported Reported Reported Reported Reported Reported Reported Reported Reported Reported Reported Reported Not reported

Summary of risk of bias Unclear Unclear Unclear Unclear Unclear Unclear Low Unclear Unclear Unclear Unclear Low Low Unclear High Low Unclear Unclear Unclear Unclear Unclear Unclear

Table 2. Baseline characteristics of included active CD studies

Reference (16) (17) (18) (19) (20) (21) (22) (23) (24) (25) (27) (26) (28) Location/sites Italian, 1site Australian, 20 sites Canadian, 16 sites USA, sites NR UK, 1 site North American, 13 sites USA, 1 site Ireland, 1 site Italy, Germany, 9 sites Europe and Israel, 55 sites Netherlands, 1 site UK, USA, 11 sites Canada, USA, 12 sites Mean age (years) 34 36 32 NR 36 NR 44 28 39 NR 34 40 39 Male (%) 38 47 43 NR 42 NR 60 31 46 NR 50 50 60 Location (SBileal, LBcolon, ICboth) 28% SB, 33% LB, 40% IC 30% SB, 36% LB, 32% IC 75% SB, 25% LB NR 22% SB, 60% IC, 14% LB 40% SB, 47% IC, 12% LB 55% SB, 15% LB, 30% IC 49% SB, 51% LB 30% SB, 70% IC NR NR 8% SB, 34% LB, 36% IC, 4% perianal NR Severity/previous surgery Mean CDAI 272, surgery 40% Mean CDAI 286, surgery NR Mean CDAI 275, surgery 36% Mean CDAI 263, surgery NR Mean CDAI 264, surgery 34% Mean CDAI 291, surgery 36% Mean CDAI 204, surgery 53% CDAI NR, surgery 16% Mean CDAI 243, Surgery 49% CDAI2220 to 400, surgery NR CDAI NR surgery NR 65% > 1 stala, surgery NR Mean CDAI=137, 64% > 1 stula, 56% surgery Duration (years) 7.7 8.4 5.5 NR NR NR 13 6.7 NR NR NR 12 9.25

CD, Crohns disease; CDAI, Crohns disease activity index; NR, not reported.

The American Journal of GASTROENTEROLOGY

VOLUME 104 | XXX 2010 www.amjgastro.com

Antibiotic Therapy in IBD

Table 3. Trial design of included active CD studies

(16)

Rifampicin 600 mg, ethambutol 15 mg/kg per day, dapsone 100 mg daily for 6 days per week, clofazimine 100 mg every 2 days vs. placebo Clarithromycin 750 mg per day, rifabutin 450 mg per day, clofazamine 50 mg per day vs. placebo Ciprooxacin 500 mg b.i.d. and Metronidazole 500 mg b.i.d. vs. placebo Clarithromycin 500 mg b.i.d. vs. placebo Clarithromycin 1 g daily vs. placebo

2 Months

All patients (prednisolone 0.71 mg/kg per day tapered)

None

Clinical assessment by lead investigator

(17)

16 Weeks

All patients (prednisolone 40 mg per day tapered) All patients (budesonide 9 mg per day tapered) NR If on stable dose < 10 mg prednisolone or < 3 mg budesonide Not allowed Allowed (30% on prednisione) All patients prednisone 40 mg per day tapered None Steroids not permitted

Stable dose AZT, 6MP, 5ASA allowed Stable immunosuppressants, iniximab allowed NR AZT, 6MP, 5ASA

CDAI < 150

(18) (19) (20)

8 Weeks 3 Months 12 Weeks

CDAI < 150 CDAI < 150 CDAI < 150

(21) (22) (23) (24) (25)

Metronidazole 10 or 20 mg/kg daily vs. placebo Cipro 500 mg b.i.d. vs. placebo Clofazamine 100 mg daily vs. placebo. Rifaximin 800 mg once or twice daily vs. placebo Rifamixin-EIR 400 mg, 800 mg, or 1.2 g b.i.d. vs. placebo (antibiotic doses combined for this systematic review) Ciprooxacin 500 mg b.i.d. vs. placebo Metronidazole 10% 0.7 g ointment t.i.d. vs. placebo Metronidazole 500 mg b.i.d. vs. ciprooxacin 500 mg b.i.d. vs. placebo

16 Weeks 3 Months 3 Months 12 Weeks 12 Weeks

Not allowed 5ASA, 6MP, antidiarrheals None 5ASA, AZT, MTX Stable dose of 5-ASA, AZT, or MTX All on iniximab, 5ASA, MTX, AZT allowed 5ASA, AZT, MTX, iniximab 5-ASA (44%), AZT (20%), MTX, iniximab (16%)

CDAI < 150 CDAI < 150 Disease activity score < 5 CDAI < 150 CDAI < 150

(27) (26) (28)

12 Weeks 4 Weeks 10 Weeks

Stable dose (17% of patients) Stable dose (3% of patients) Stable dose (last 2 weeks)

50% Reduction of draining stulas PCDAI decrease by 5 Closure of at least 50% of actively draining stulas

CD, Crohns disease; CDAI, Crohns disease activity index.

Patients that were enrolled tended to have moderate CD (Table 2) and therapy was given for 416 weeks (Table 3) with only one low risk of bias trial (22) (Table 1). There was a statistically significant effect of antibiotics to induce remission in active CD compared with placebo (RR of active CD not in remission = 0.85; 95% CI = 0.73 0.99, P = 0.03) (Figure 2). There was moderate heterogeneity between results (I2 = 44%. Cochran Q = 16.06 (degree of freedom (df) = 9), P = 0.07). There was no statistically significant funnel plot asymmetry (Egger test = 0.38 (95% CI = 2.711.95), P = 0.72) suggesting no evidence of publication bias or other small study effects (Figure 3). A diverse number of antibiotics were tested either alone or in combination. It was therefore difficult to definitively evaluate whether any particular antibiotic was effective in active CD. We therefore conducted analyses that specifically looked at single antibiotics as well as analyses that evaluated all trials that contained a specific antibiotic class even if other antibiotics were also used in the combination (and therefore trials that involved combinations of antibiotics could be involved in more than one analysis). Two trials (24,25) involving 485 patients suggested that rifamixin was effective at inducing remission (RR = 0.81; 95% CI = 0.680.97)
2011 by the American College of Gastroenterology

with no statistically significant heterogeneity between the two trials (I2 = 0%. Cochran Q = 0.17 (df = 1), P = 0.68). The number needed to treat (NNT) was 9 (95% CI = 550). These were one of the few studies that did not combine antibiotics with other medical therapy such as steroids to induce remission. Data from one of the trials (25) was only available in abstract and this was a dose ranging phase II study. All doses were combined for this meta-analysis although the data suggested 800 mg b.i.d. was more effective than either 400 mg b.i.d. or 1.2 g b.i.d. for 12 weeks (25). Another approach to evaluating the data is to pool all trials that contain a rifamycin derivative. In this analysis, there were four trials (16,17,24,25) evaluating 738 patients given antibiotics for 816 weeks. There was a statistically significant effect in favor of rifamycins (RR = 0.78; 95% CI = 0.76 0.91) with no statistically significant heterogeneity between studies (I2 = 0%. Cochran Q = 2.11 (df = 3), P = 0.55) (Figure 4). The NNT was 11 (95% CI = 5100). One trial (22) did suggest ciprofloxacin was effective (RR = 0.68; 95% CI = 0.510.94) but this study evaluated only 84 patients (Figure 2). This was, however, the only trial (22) with low risk of bias. The NNT for ciprofloxacin in this trial was 4 (95% CI = 211).
The American Journal of GASTROENTEROLOGY

REVIEW

Reference

Interventions

Duration

Steroids

Co-therapy permitted

Outcome

Khan et al.

Placebo Antibiotic Study or subgroup Events Total Events Total Weight 1.1.1 Antituberculosis combination 3 Prantera et al. (16) 1 0.4% 22 18 111 Selby et al. (17) 35 102 56 11.4% Subtotal (95% CI) 124 129 11.8% Total events 57 38 Heterogeneity: 2 = 0.21; 2 = 1.33, df = 1 (P = 0.25); I 2 = 25% Test for overall effect: Z = 0.46 (P = 0.65) 1.1.2 Other combination Steinhart et al. (18) Subtotal (95% CI) 45 66 66 43 43 68 68 15.0% 15.0%

Risk ratio MH, random, 95% CI 2.45 (0.28, 21.62) 0.68 (0.49, 0.94) 0.81 (0.34, 1.96)

Risk ratio MH, random, 95% CI

REVIEW

1.08 (0.84, 1.38) 1.08 (0.84, 1.38)

45 Total events Heterogeneity: not applicable Test for overall effect: Z = 0.60 (P = 0.55) 1.1.3 Macrolides Graham et al. (19) Leiper et al. (20) Subtotal (95% CI) 2 16 7 19 26

7 16

8 22 30

1.4% 11.5% 13.0%

0.33 (0.10, 1.08) 1.16 (0.84, 1.60) 0.69 (0.17, 2.71)

Total events 18 23 Heterogeneity: 2 = 0.81; 2 = 5.05, df = 1 (P = 0.02); I 2 = 80% Test for overall effect: Z = 0.54 (P = 0.59) 1.1.4 5-Nitroimidazole Sutherland et al. (21) Subtotal (95% CI)

43

63 63

27 27

36 36

14.6% 14.6%

0.91 (0.71, 1.17) 0.91 (0.71, 1.17)

Total events 43 Heterogeneity: not applicable Test for overall effect: Z = 0.73 (P = 0.47) 1.1.5 Flouroquinilones Arnold et al. (22) 25 43 Subtotal (95% CI) 43 25 Total events Heterogeneity: not applicable Test for overall effect: Z = 2.66 (P = 0.008) 1.1.6 Rifamixin

35 35

41 41

13.2% 13.2%

0.68 (0.51, 0.90) 0.68 (0.51, 0.90)

Prantera et al. (24) 54 32 20 29 11.2% 137 301 58 101 17.0% Prantera et al. (25) 355 130 Subtotal (95% CI) 28.2% Total events 169 78 Heterogeneity: 2 = 0.00; 2 = 0.17, df = 1 (P = 0.68); I 2 = 0% Test for overall effect: Z = 2.33 (P = 0.02) 1.1.7 Clofazamine Afdhal et al. (23) 9 25 Subtotal (95% CI) 25 Total events 9 Heterogeneity: not applicable Test for overall effect: Z = 0.98 (P = 0.33)

0.86 (0.62, 1.19) 0.79 (0.64, 0.98) 0.81 (0.68, 0.97)

12 12

24 24

4.2% 4.2%

0.72 (0.37, 1.39) 0.72 (0.37, 1.39)

702 458 100.0% Total (95% CI) 347 275 Total events Heterogeneity: 2 = 0.02; 2 = 16.06, df = 9 (P = 0.07); I 2 = 44% Test for overall effect: Z = 2.16 (P = 0.03)

0.85 (0.73, 0.99)

0.05 1 0.2 Favors experimental

20 5 Favors control

Figure 2. Efcacy of antibiotics vs. placebo at inducing remission in active Crohns disease. CI, condence interval; df, degree of freedom; M-H, Mantel-Haenszel test.

The American Journal of GASTROENTEROLOGY

VOLUME 104 | XXX 2010 www.amjgastro.com

Antibiotic Therapy in IBD

1.5

2 0.05 0.2 1 5 20

RR
Subgroups Antituberculosis combination Other combination Macrolides 5-Nitroimidazole Flouroquinilones Rifamixin Clofazamine

Figure 3. Funnel plot of RCTs evaluating antibiotics against placebo in active Crohns disease patients. RR, relative risk.

In the analysis that included all trials that contained a fluoroquinolone there were two trials (18,22) evaluating 218 patients and there was no statistically significant effect of this antibiotic class inducing remission in active CD (RR = 0.86; 95% CI = 0.551.35; Figure 4). There were two trials (19,20) evaluating macrolides in 56 patients. Both trials evaluated a total daily dose of 1 g of clarithromycin and there was no statistically significant effect at inducing remission in active CD (RR = 0.69; 95% CI = 0.172.71). There was significant heterogeneity between the two studies (I2 = 80%. Cochran Q = 5.05 (df = 1), P = 0.02) with one trial (19) showing a trend toward a positive result and the other (20) being negative (Figure 2). There was another trial (17) that used the same dose of clarithromycin with other antibiotics and when this was added to the analysis there were 269 patients with no statistically significant effect of active treatment (RR = 0.76; 95% CI = 0.421.36) and the significant heterogeneity persisted (I2 = 79%. Cochran Q = 9.38 (df = 2), P = 0.009; Figure 4). One trial (21) investigated metronidazole in 99 active CD patients and reported no statistically significant benefit of treatment (RR = 0.91; 95% CI = 0.711.17; Figure 2). An additional trial (18) contained metronidazole as part of a combination of antibiotics and when these two studies were pooled there were 233 patients with no statistically significant effect of the 5-nitroimidazole (RR = 0.99; 95% CI = 0.831.18). There was no significant heterogeneity between the two studies (I2 = 0%. Cochran Q = 0.90 (df = 1), P = 0.34; Figure 4). One trial (23) reported on the efficacy of clofazamine alone in 49 active CD patients with no statistically significant effect of the antibiotic (RR = 0.72; 95% CI = 0.371.39; Figure 2). In another analysis, pooling all trials that contained clofazamine alone or in combination with other antibiotics there were three trials (16,17,23) evaluating 302 patients over 816 weeks. Different dosing schedules were used in each trial with a daily dose varying between 50 and 100 mg.
2011 by the American College of Gastroenterology

Antibiotics for induction of remission of UC

There were nine eligible trials (6,7,2935) involving 662 patients with active UC that gave remission as an outcome. Therapy was given for 7 days to 3 months (Table 4) and generally patients with moderately active UC were recruited (Table 5) with only one low risk of bias trial (31) (Table 1). There were three trials (6,7,32) that evaluated ciprofloxacin with no statistically significant benefit over placebo (Figure 6). All other trials assessed different single or combinations of antibiotics. Overall, there was a statistically significant effect in favor of antibiotics (RR of active UC not in remission = 0.64; 95% CI = 0.430.96, P = 0.03) (Figure 6). The NNT was 7 (95% CI = 425). There was moderate heterogeneity between results (I2 = 69%. Cochran Q = 25.8 (df = 8), P = 0.001). There was statistically significant funnel plot asymmetry 1.32 (95% CI = 2.56 to 0.07; P = 0.04; Figure 7), suggesting publication bias or other small study effects. Furthermore, one trial (34) reported clinical remission as a primary outcome but did not provide a definition for this nor did the authors describe the duration of follow-up with the end point being discharge from hospital. As the large majority of patients were likely to be discharged from hospital by 16 weeks (the cutoff for eligibility into the systematic review for therapy to induce remission) we included the trial. We excluded this study in a sensitivity analysis and the effect of antibiotics was not statistically significant although there was still a trend toward benefit (RR = 0.69; 95% CI = 0.471.01).
Antibiotics for maintenance of remission in quiescent UC and CD

Overall, there were 3 RCTs treating 186 patients with quiescent CD (16,17,23). There was a statistically significant effect of antibiotics in preventing CD relapse compared with placebo (RR
The American Journal of GASTROENTEROLOGY

REVIEW

0.5

There was a statistically significant effect in favor of clofazamine (RR = 0.70; 95% CI = 0.530.94) with no significant heterogeneity between studies (I2 = 0%. Cochran Q = 1.33 (df = 2), P = 0.51; Figure 4). There were two trials (16,17) evaluating 253 patients that reported on a combination of anti-tuberculosis antibiotic therapy. Overall there was no statistically significant effect of therapy (RR = 0.81; 95% CI = 0.341.96) with mild heterogeneity between studies (I2 = 25%. Cochran Q = 1.33 (df = 1), P = 0.25; Figure 2). One large trial (17) involving 213 patients was positive, whereas a small trial (16) involving 40 patients was negative (Figure 2). In addition, there were three trials (2628) that evaluated antibiotics in 123 CD patients with perianal fistulas. Therapy was given of 412 weeks (Table 3) and there were two low risk of bias trials (27,28) (Table 1). One trial evaluated metronidazole (26), one trial assessed ciprofloxacin (27) and one trial investigated both antibiotics (28) vs. placebo. Studies did not provide remission as an end point and all gave reduction in fistula drainage as an outcome so these data were synthesized. There was a trend for benefit for each antibiotic that was not statistically significant (Figure 5). When both antibiotics were combined a random effects model suggested a statistically significant benefit of antibiotics in active CD perianal fistulas (RR = 0.80; 95% CI = 0.660.98, P = 0.03) with no evidence of heterogeneity (I2 = 0%. Cochran Q = 2.9 (df = 2), P = 0.41). The NNT was 5 (95% CI = 320).

SE(log[RR])

Khan et al.

REVIEW

Antibiotic Study or subgroup Events Total 1.1.1 Contains rifamycins 3 Prantera et al. (16) 22 Prantera et al. (24) 32 54 Prantera et al. (25) 301 137 Selby et al. (17) 102 35 Subtotal (95% CI) 479

Placebo Events Total 1 20 58 56 18 29 101 111 259

Weight 0.5% 22.0% 55.0% 22.5% 100.0%

Risk ratio MH, random, 95% CI 2.45 (0.28, 21.62) 0.86 (0.62, 1.19) 0.79 (0.64, 0.98) 0.68 (0.49, 0.94) 0.78 (0.67, 0.91)

Risk ratio MH, random, 95% CI

Total events 135 207 Heterogeneity: 2 = 0.00; 2 = 2.11, df = 3 (P = 0.55); I 2 = 0% Test for overall effect: Z = 3.09 (P = 0.002) 1.1.3 Contains macrolides Graham et al. (19) Leiper et al. (20) Selby et al. (17) Subtotal (95% CI) 2 16 35 7 19 102 128 7 16 56 8 22 111 141 15.8% 42.2% 42.0% 100.0% 0.33 (0.10, 1.08) 1.16 (0.84, 1.60) 0.68 (0.49, 0.94) 0.76 (0.42, 1.36)

Total events 79 53 Heterogeneity: 2 = 0.18; 2 = 9.38, df = 2 (P = 0.009); I 2 = 79% Test for overall effect: Z = 0.93 (P = 0.35) 1.1.4 Contains clofazamine 9 25 12 24 19.3% Afdhal et al. (22) Prantera et al. (16) 1 1.8% 18 3 22 Selby et al. (17) 102 56 78.9% 111 35 149 Subtotal (95% CI) 153 100.0% 69 47 Total events Heterogeneity: 2 = 0.00; 2 = 1.33, df = 2 (P = 0.51); I 2 = 0% Test for overall effect: Z = 2.38 (P = 0.02) 1.1.5 Contains ethambutol 3 Prantera et al. (16) 22 22 Subtotal (95% CI) 3 Total events Heterogeneity: not applicable Test for overall effect: Z = 0.81 (P = 0.42) 1.1.6 Contains 5-nitroimidazole Steinhart et al. (18) Sutherland et al. (21) Subtotal (95% CI) 45 43 66 63 129 43 27 68 36 104 51.6% 48.4% 100.0% 1.08 (0.84, 1.38) 0.91 (0.71, 1.17) 0.99 (0.83, 1.18) 1 1 18 18 100.0% 100.0% 2.45 (0.28, 21.62) 2.45 (0.28, 21.62) 0.72 (0.37, 1.39) 2.45 (0.28, 21.62) 0.68 (0.49, 0.94) 0.70 (0.53, 0.94)

Total events 88 70 Heterogeneity: 2 = 0.00; 2 = 0.90, df = 1 (P = 0.34); I 2 = 0% Test for overall effect: Z = 0.08 (P = 0.94) 1.1.7 Contains flouroquinilones Arnold et al. (22) 25 43 35 41 48.7% Steinhart et al. (18) 45 66 43 68 51.3% Subtotal (95% CI) 109 109 100.0% Total events 78 70 Heterogeneity: 2 = 0.09; 2 = 5.79, df = 1 (P = 0.02); I 2 = 83% Test for overall effect: Z = 0.65 (P = 0.52) 0.68 (0.51, 0.90) 1.08 (0.84, 1.38) 0.86 (0.55, 1.35)

0.05 0.2 1 Favors experimental

20 5 Favors control

Figure 4. Efcacy of antibiotics vs. placebo at inducing remission in active Crohns disease. Analysis according to whether a specic class of antibiotic was used either alone or in combination with other antibiotics. CI, condence interval; df, degree of freedom; M-H, Mantel-Haenszel test.

The American Journal of GASTROENTEROLOGY

VOLUME 104 | XXX 2010 www.amjgastro.com

Antibiotic Therapy in IBD

Study or subgroup 1.1.1 Metronidazole

Antibiotic Events Total

Placebo Events Total

Weight

Risk ratio MH, random, 95% CI 0.79 (0.62, 1.02) 0.98 (0.66, 1.46) 0.84 (0.68, 1.04)

Risk ratio MH, random, 95% CI

Maeda et al. (26) 23 33 36 41 60.6% Thia et al. (28) 8 6 7 7 24.1% 49 40 84.7% Subtotal (95% CI) 43 Total events 29 Heterogeneity: 2 = 0.00; 2 = 0.80, df = 1 (P = 0.37); I 2 = 0% Test for overall effect: Z = 1.57 (P = 0.12) 1.1.2 Ciprofloxacin Thia et al. (28) West et al. (27) Subtotal (95% CI)

6 3

10 11 21

7 8

8 13 21

11.9% 3.5% 15.3%

0.69 (0.39, 1.21) 0.44 (0.15, 1.27) 0.62 (0.38, 1.03)

9 15 Total events Heterogeneity: 2 = 0.00; 2 = 0.63, df = 1 (P = 0.43); I 2 = 0% Test for overall effect: Z = 1.86 (P = 0.06) Total (95% CI) 61 70 100.0% 0.80 (0.66, 0.98)

38 58 Total events Heterogeneity: 2 = 0.00; 2 = 2.90, df = 3 (P = 0.41); I 2 = 0% Test for overall effect: Z = 2.17 (P = 0.03) Test for subgroup differences: not applicable

0.01 1 0.1 Favors experimental

10 100 Favors control

Figure 5. Efcacy of antibiotics vs. placebo in Crohns disease perianal stulae. CI, condence interval; df, degree of freedom; M-H, Mantel-Haenszel test.

Table 4. Baseline characteristics of included active UC studies

Reference (29) (30) (31) (6) (32) (7) (33) (34) (35) Location/sites Greece, 1site Japan, 1 site Japan, 11 sites Greece, 1 site Greece, 1 site Finland, 1 site UK, 1 site UK, 1 site Italy, 1 site Mean age (years) 41.5 NR 38.9 41.5 41.5 34.2 43.4 37.1 NR Male (%) 46 60 67 47 47 70 54 55 NR Distribution 54% Pancolitis, 38% left, 8% proctitis 70% Pancolitis, 25% left, 5% proctitis 49% Pancolitis, 37% left, 14% proctitis 30% Pancolitis, 33% left, 37% proctitis 51% Pancolitis, 40% left, 9% proctitis NR 42% Pancolitis, 32% left, 26% proctitis 61% Pancolitiis NR Severity NR NR 40% Mild, 58% moderate, 2% severe 46% Moderate NR 52% On steroids All in-patients with active UC 45% Severe, 55% moderate all in-patients 52% On steroids Duration (years) 2.1 NR NR 2.1 2.1 5.4 6.6 6.4 5.4

NR, not reported; UC, ulcerative colitis.

of relapse = 0.62; 95% CI = 0.460.84; Figure 8). The NNT was 4 (95% CI = 310). There was no significant heterogeneity between results (I2 = 0%. Cochran Q = 1.17 (df = 2), P = 0.56). All studies evaluated antimicrobials that could be considered antimycobacterial although all studied different regimens (Table 6). Follow-up was for 912 months and all trials did not report method of randomization or method of concealment of allocation and so had unclear risk of bias (Table 1). One study (36) did report long-term data in UC. This trial (36) reported on 2-year follow-up of an acute UC trial that compared 7 days of oral tobramycin vs. placebo (33). This trial evaluated 81 patients and there was no difference in relapse rates at 1 and 2 years suggesting 7 days of tobramycin did not influence long-term relapse rates.
2011 by the American College of Gastroenterology

DISCUSSION
To our knowledge, this is the first comprehensive systematic review looking at all antibacterial therapies in active and quiescent disease for both UC and CD. Our data suggest antibiotic therapy may induce remission in active UC and CD, as well as prevent relapse in patients with quiescent CD. This supports a number of studies that suggest a bacterial origin for IBD (48). The strongest effect was seen in preventing relapse of CD with an NNT of 4 (95% CI = 310) and all antibiotic combinations studied had some antimycobacterial properties. This is consistent with a systematic review (49) of 28 casecontrol studies suggesting that MAP is associated with CD. However, the antibiotic combinations were diverse and this is only based on three trials so there
The American Journal of GASTROENTEROLOGY

REVIEW

10

Khan et al.

Table 5. Trial design of included active UC studies

REVIEW

Reference (29)

Interventions Metronidazole 500 mg t.i.d., tobramycin 4 mg/kg divided q8 h vs. placebo Amoxicillin 500 mg t.i.d., tetracycline 500 mg t.i.d., metronidazole 250 mg t.i.d. vs. no treatment Amoxicillin 500 mg t.i.d., tetracycline 500 mg t.i.d., Metronidazole 250 mg t.i.d. vs. placebo Ciprooxacin 250 mg b.i.d. vs. placebo

Duration 10 days

Adjunctive therapy IV and rectal hydrocortisone (all patients) 5ASA/sulphasalazine, steroids, probiotics Sulphasalazine, steroids, immunosuppressive Prednisolone, betamethasone enema, and olsalazine 0.5 mg b.i.d. IV and rectal hydrocortisone (all patients) Steroid taper, rectal steroids, mesalamine No steroids (all patients), sulphasalazine allowed Prednisolone 40 mg per day with a tapering dose Steroids tapered

Outcomes Remission dened as 3 bowel movements per day, no blood, no systemic signs of severe colitis Lichtiger score

(30)

3 Months

(31)

3 Months

Mayo score 2

(6)

14 days, outcome at 79 weeks 10 days

Remission dened as formed stools, not bleeding, normal heme/chem prole, normal sigmoidoscopy Remission dened as 3 bowel movements per day, no blood, no systemic signs of severe colitis Global Endoscopic scoreno inammation Clinical remission ( < 3 bowel movements per day with no blood in stool). Also gave normal histology as an outcome Clinical remission (poorly dened) Remission dened as 3 bowel movements per day, no blood, no systemic signs of severe colitis

(32)

Ciprooxacin 400 mg IV b.i.d. vs. placebo

(7) (33)

Ciprooxacin 500 mg or 750 mg b.i.d. vs. placebo Tobramycin 120 mg t.i.d. vs. placebo

3 Months 34 weeks

(34) (35)

Vancomycin 500 mg q.i.d. vs. placebo Rifaximin 250 mg b.i.d. vs. placebo

7 days 10 days

UC, ulcerative colitis.

is currently insufficient data to recommend antibiotic therapy to prevent relapse of CD. Indeed, in all trials of active UC and CD a variety of single antibiotics and antibiotic combinations were evaluated, so it is not possible to recommend a specific antibiotic therapy for IBD. The exception to this is the use of either ciprofloxacin or metronidazole for treating CD perianal fistulas. Again data are sparse but this systematic review does suggest these agents may be effective in reducing fistula drainage and given the limited therapeutic options in this condition it would seem sensible to try one or other of these antibiotics in patients with perianal fistulae. In active CD, therapies that contained rifamycins appeared to be effective at inducing remission. There also seemed to be a benefit of clofazamine although this was difficult to interpret as 2/3 trials evaluating this antibiotic also used other antimicrobials in combination. As these are anti-tuberculosis drugs it is tempting to speculate that this again suggests a mycobacterial origin for CD. It is important to emphasize, however, that these antibiotics have a broad spectrum of action. Rifamycins, for example, have activity against Bacteroides, Chlamydia, Trichomonas, and Listeria species (50). Evidence supporting the possibility that these antibiotics are acting on organisms other than mycobacteria is that combinations of anti-tuberculosis therapy were not more effective than these single agents alone. Indeed, overall there was no statistically significant effect of antimycobacterial combination therapy in active CD although there were only two trials (16,17) evaluating this approach.
The American Journal of GASTROENTEROLOGY

The overall data do suggest that altering gut microbial flora may have a role in modulating IBD activity. The focus of IBD research has been mainly related to defining abnormal immune responses (51) with a MEDLINE literature search identifying 13,358 hits relating to immune function and IBD. In contrast, the evaluation of gut microbial flora in IBD has received much less attention (52) with only 1,570 hits from the same MEDLINE search. Our systematic review suggests this imbalance should be changed, as this could be a fruitful area for future research. It is not clear whether antibiotics are having an effect by acting on one bacterial species or changing the composition of gut microbiome in general (53). It is also possible that the gut bacteria have no direct effect in causing IBD, but secondary infection with gut organisms can exacerbate disease (49). This systematic review has several strengths. We used an extensive search strategy and rigorous methodology to define eligibility and treatment success. Most studies use similar outcomes for defining remission so pooling the data seems appropriate. It could be argued that it is inappropriate to pool data from different antibiotic classes. This is true if the question that is being asked is what antibiotic class should be used in CD or UC. However, if the purpose of the review is to determine whether altering the gut flora can influence IBD activity then pooling is appropriate. That was the aim of this study and this is why we pooled different classes of antibiotics. This review was conducted in the context of a wider review of all major medical therapies in
VOLUME 104 | XXX 2010 www.amjgastro.com

Antibiotic Therapy in IBD

11

7 5 89

20 10 105 135

10.7% 6.0% 19.4% 36.0%

1.05 (0.46, 2.43) 0.40 (0.10, 1.60) 0.96 (0.84, 1.08) 0.95 (0.84, 1.07)

94 101 Total events Heterogeneity: 2 = 0.00; 2 = 1.64, df = 2 (P = 0.44); I 2 = 0% Test for overall effect: Z = 0.82 (P = 0.41) 1.1.2 Ciprofloxacin Mantzaris et al. (6) Mantzaris et al. (32) Turunen et al. (7) Subtotal (95% Cl)

10 6 7

34 29 38 101

10 6 23

36 26 45 107

11.9% 9.0% 12.1% 32.9%

1.06 (0.50, 2.22) 0.90 (0.33, 2.44) 0.36 (0.17, 0.75) 0.68 (0.33, 1.39)

23 39 Total events Heterogeneity: 2 = 0.22; 2 = 4.62, df = 2 (P = 0.10); I 2 = 57% Test for overall effect: Z = 1.05 (P = 0.29) 1.1.3 Other single antibiotic Burke et al. (33) Dickison et al. (34) Gionchetti et al. (35) Subtotal (95% Cl)

11 2 5

42 18 14 74

24 7 9

42 15 14 71

14.2% 5.8% 11.1% 31.1%

0.46 (0.26, 0.81) 0.24 (0.06, 0.98) 0.56 (0.25, 1.24) 0.46 (0.29, 0.71)

Total events 18 40 Heterogeneity: 2 = 0.00; 2 = 1.07, df = 2 (P = 0.59); I 2 = 0% Test for overall effect: Z = 3.48 (P = 0.0005) Total (95% Cl) 309 313 100.0% 0.64 (0.43, 0.96)

Total events 135 180 Heterogeneity: 2 = 0.22; 2 = 25.81, df = 8 (P = 0.001); I 2 = 69% Test for overall effect: Z = 2.14 (P = 0.03)

0.1 0.2 0.5 1 Favors experimental

2 5 10 Favors control

Figure 6. Efcacy of antibiotics vs. placebo at inducing remission in active ulcerative colitis. CI, condence interval; df, degree of freedom; M-H, Mantel-Haenszel test.

IBD so the results can be placed into context with other agents. The main limitation of the review is that there were only four trials (22,26,28,31) with low risk of bias and empirical studies (54) would suggest that trials with an unclear risk of bias tend to overestimate treatment effect. It is therefore possible that the trend to overall benefit seen with antibiotics for UC and CD is an artifact due to bias. In the American College of Gastroenterology monograph of medical therapies for IBD (55), we identified 6,765 IBD patients enrolled into 5-ASA trials and 4,196 IBD patients participating in biological therapy trials and yet there are only 1,945 IBD patients included in antibiotic trials. It is important to address this discrepancy if we are to understand the role of antibiotic therapy in treating IBD. There have been major advances in delineating the gut microbiome (56). Culture-independent analysis of changes in intestinal microbiota should be included in trials of antibiotic therapy to explore what effects these interventions are having on the gut flora. This will act as surrogate marker for the effectiveness of the antibiotic as well as give mechanistic information on how the drug is acting. Our data suggest that further research into antibiotic therapy in CD and UC should be a priority. Future studies should be directed at which antibiotic or antibiotic com 2011 by the American College of Gastroenterology

0.2

SE(log[RR])

0.4

0.6

0.8

1 0.1 0.2 0.5 1 2 5 10

RR
Subgroups Multiple antibiotics Cliprofloxacin Other single antbiotic

Figure 7. Funnel plot of RCTs evaluating antibiotics against placebo in active ulcerative colitis patients. RR, relative risk.

bination is most effective. Ideally, individual antibiotics should be evaluated separately and together in a factorial design to establish whether different classes of antibiotics have a synergistic action in treating IBD.
The American Journal of GASTROENTEROLOGY

REVIEW

Antibiotic Study or subgroup Events Total 1.1.1 Multiple antibiotics Mantzaris et al. (29) 7 19 Ohkusa et al. (30) 2 10 Ohkusa et al. (31) 85 105 Subtotal (95% Cl) 134

Control Events Total

Weight

Risk ratio MH, random, 95% CI

Risk ratio MH, random, 95% CI

12

Khan et al.

Study or subgroup

Antibiotic Events Total 3 8 26 16 19 67 102

Control Events Total 6 13 31 12 17 55 84

Weight 7.0% 27.3% 65.7% 100.0%

Risk ratio MH, random, 95% CI 0.38 (0.12, 1.20) 0.55 (0.31, 0.99) 0.69 (0.47, 1.01) 0.62 (0.46, 0.84)

Risk ratio MH, random, 95% CI

REVIEW

Afdhal et al. (23) Prantera et al. (16) Selby et al. (17) Total (95% CI)

Total events 37 50 Heterogeneity: 2 = 0.00; 2 = 1.17, df = 2 (P = 0.56); I 2 = 0% Test for overall effect: Z = 3.03 (P = 0.002)

0.1 1 0.01 Favors experimental

100 10 Favors control

Figure 8. Efcacy of antibiotics vs. placebo at preventing relapse in quiescent Crohns disease. CI, condence interval; df, degree of freedom; M-H, Mantel-Haenszel test.

Table 6. Summary of studies evaluating the efcacy of antibiotics to prevent relapse in quiescent CD
Reference (23) Interventions Clofazamine 100 mg daily vs. placebo. Adjunctive steroid taper. 12 months follow-up (9 months of maintenance) Rifampicin 600 mg daily, ethambutol 15 mg/kg per day, dapsone 100 mg daily for 6 days per week, clofazimine 100 mg every 2 days vs. placebo. All patients also had prednisolone taper. Follow-up 9 months (7 months maintenance) Clarithromycin 750 mg per day, Rifabutin 450 mg per day, Clofazamine 50 mg per day vs. placebo for 52 weeks, adjunctive therapy with prednisolone taper. Azathioprine, 6MP, 5ASA allowed. Follow-up at 1 year was used. The study followed patients up for 2 years but 1-year time frame was taken as 50% drop out by second year Outcomes Disease activity score >10 Clinical assessment by lead investigator

(16)

interpreted the data. Paul Moayyedi provided statistical advice and support. Khurram J Khan and Paul Moayyedi drafted the manuscript. All authors commented on drafts of the manuscript and approved the final draft. Financial support: This study was funded by the American College of Gastroenterology. Potential competing interests: Paul Moayyedi chairperson at McMaster University partly funded by an unrestricted donation by AstraZeneca, and has received consultants and speakers bureau fees from AstraZeneca, AxCan Pharma, Nycomed, and Johnson and Johnson. The remaining authors declare no conflict of interest.
REFERENCES
1. Krisner JB. Historical aspects of inflammatory bowel disease. J Clin Gastroenterol 1988;10:28697. 2. Dalziel TK. Chronic interstitial enteritis. BMJ 1913;25:105870. 3. Sartor RB. Does Mycobacterium avium subspecies paratuberculosis cause Crohns disease? Gut 2005;54:8968. 4. Abraham C, Cho JH. Inflammatory bowel disease. N Engl J Med 2009;361:206678. 5. Sellon R, Tonkonogy S, Schultz M et al. Resident enteric bacteria are necessary for development of spontaneous colitis and immune system activation in interleukin-10-deficient mice. Infect Immun 1998;66:522431. 6. Mantzaris GJ, Archavlis E, Christoforidis P et al. A prospective randomized controlled trial of oral ciprofloxacin in acute ulcerative colitis. Am J Gastroenterol 1997;92:4546. 7. Turunen UM, Farkkila MA, Hakala K et al. Long-term treatment of ulcerative colitis with ciprofloxacin: a prospective, double-blind, placebo-controlled study. Gastroenterology 1998;115:10728. 8. Borgaonkar MR, Macintosh DG, Fardy JM et al. A meta-analysis of antimycobacterial therapy for Crohns disease. Am J Gastroenterol 2000;95:7259. 9. Rahimi R, Nikfar S, Rezaie A et al. A meta-analysis of broad-spectrum antibiotic therapy in patients with active crohns disease. Clin ther 2006, 19838. 10. Feller M, Huwiler K, Schoepfer A et al. Long-term antibiotic treatment for Crohns disease: systematic review and meta-analysis of placebo-controlled trials. Clin Infect Dis 2010;50:47380. 11. Rahimi R, Nikfar S, Rezaie A et al. A meta-analysis of antibiotic therapy for active ulcerative colitis. Dig Dis Sci 2007;52:29205. 12. Higgins JPT, Green S (eds). Cochrane Handbook for Systematic Reviews of Interventions. John Wiley & Sons, Chichester, UK (ISBN 9780470057964). 13. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:17788. 14. Higgins JPT, Thompson SG, Deeks JJ et al. Measuring inconsistency in meta-analyses. BMJ 2003;327:55760. 15. Egger M, Davey-Smith G, Schneider M et al. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315:62934. 16. Prantera C, Kohn A, Mangiarotti R et al. Antimycobacterial therapy in Crohns disease: results of a controlled, double-blind trial with a multiple antibiotic regimen. Am J Gastroenterol 1994;89:5138.

(17)

CDAI > 150 with an increase 60

CD, Crohns disease; CDAI, Crohns disease activity index.

ACKNOWLEDGMENTS

This study was funded by the American College of Gastroenterology and performed to inform their monograph on inflammatory bowel disease. We thank Jean-Paul Achkar, Charles N Bernstein, Marla C Dubinsky, Stephen B Hanauer, Sunanda V Kane, and William J Sandborn for their contributions to the discussion concerning the role of antibiotics in the treatment of inflammatory bowel disease. We are indebted to Ilya Sobolev, who cheerfully translated Russian articles that were identified as potentially eligible. We are also grateful to Racquel Simpson, Sally Kohne, and Karin Dearness for conducting the search strategies, retrieving references, and generally supporting the systematic review.
CONFLICTS OF INTEREST

Guarantor of the article: Paul Moayyedi, MB, ChB, BSc, MPH, PhD, FACG. Specific author contributions: Khurram J. Khan, Thomas A. Ullman, Alexander C. Ford, Maria T. Abreu, A. Abadir, John K. Marshall, Nicholas J. Talley, and Paul Moayyedi conceived and drafted the study. Alexander C. Ford, Khurram J. Khan, and Paul Moayyedi collected all data. J. Khan and Paul Moayyedi analyzed and
The American Journal of GASTROENTEROLOGY

VOLUME 104 | XXX 2010 www.amjgastro.com

Antibiotic Therapy in IBD

13

17. Selby W, Pavli P, Crotty B et al. Two-year combination antibiotic therapy with clarithromycin, rifabutin, and clofazimine for Crohns disease. Gastroenterology 2007;132:23139. 18. Steinhart AH, Feagan BG, Wong CJ et al. Combined budesonide and antibiotic therapy for active Crohns disease: a randomized controlled trial. Gastroenterology 2002;123:3340. 19. Graham DY, Al-Assi MT, Robinson M. Prolonged remission in Crohns disease following therapy for mycobacterium paratuberculosis infection. Gastroenterology 1995;108:A826. 20. Leiper K, Martin K, Ellis A et al. Clinical trial: randomized study of clarithromycin versus placebo in active Crohns disease. Aliment Pharmacol Ther 2008;27:12339. 21. Sutherland L, Singleton J, Sessions J et al. Double blind, placebo controlled trial of metronidazole in Crohns disease. Gut 1991;32:10715. 22. Arnold GL, Beaves MR, Pryjdun VO et al. Preliminary study of ciprofloxacin in active Crohns disease. Inflamm Bowel Dis 2002;8:105. 23. Afdhal NH, Long A, Lennon J et al. Controlled trial of antimycobacterial therapy in Crohns disease. Clofazimine versus placebo. Dig Dis Sci 1991;36:44953. 24. Prantera C, Lochs H, Campieri M et al. Antibiotic treatment of Crohns disease: results of a multicentre, double blind, randomized, placebo-controlled trial with rifaximin. Aliment Pharmacol Ther 2006;23:111725. 25. Prantera C, Lochs H, Giochetti P et al. Rifaximin-EIR (extended intestinal release) 400 mg tablets in the treatment of moderately active Crohns disease: results of the international multicentre, randomised, double-blind, placebo-controlled trial RETIC-03. Gut 2010;59 Suppl III: A1. 26. Maeda Y, Ng SC, Durdey P, et al., The Topical Metronidazole in Perianal Crohns Study Group. Randomized clinical trial of metronidazole ointment versus placebo in perianal Crohns disease. Br J Surg 2010;97:13407. 27. West RL, van der Woude CJ, Hansen BE et al. Clinical and endosonographic effect of ciprofloxacin on the treatment of perianal fistulae in Crohns disease with infliximab: a double-blind placebo-controlled study. Aliment Pharmacol Ther 2004;20:132936. 28. Thia KT, Mahadevan U, Feagan BG et al. Ciprofloxacin or metronidazole for the treatment of perianal fistulas in patients with Crohns disease: a randomized, double-blind, placebo-controlled pilot study. Inflamm Bowel Dis 2009;15:1724. 29. Mantzaris GJ, Hatzis A, Kontogiannis P et al. Intravenous tobramycin and metronidazole as an adjunct to corticosteroids in acute, severe ulcerative colitis. Am J Gastroenterol. 1994;89:436. 30. Ohkusa T, Nomura T, Terai T et al. Effectiveness of antibiotic combination therapy in patients with active ulcerative colitis: A randomized, controlled pilot trial with long-term follow-up. Scand J Gastroenterol 2005;40:133442. 31. Ohkusa T, Kato K, Terao T et al. Newly developed antibiotic combination therapy for ulcerative colitis: a double-blind placebo-controlled multicenter trial. Am J Gastroenterol 2010;105:18209. 32. Mantzaris GJ, Petraki K, Archavlis E et al. A prospective randomized controlled trial of intravenous ciprofloxacin as an adjunct to corticosteroids in acute, severe ulcerative colitis. Scand J Gastroenterol 2001;36:9714. 33. Burke DA, Axon ATR, Clayden SA et al. The efficacy of tobramycin in the treatment of ulcerative colitis. Aliment Pharmacol Ther 1990;4:1239. 34. Dickinson RJ, OConnor HJ, Pinder I et al. Double blind controlled trial of oral vancomycin as adjunctive treatment in acute exacerbations of idiopathic colitis. Gut 1985;26:13804. 35. Gionchetti P, Rizzello F, Ferrieri A et al. Rifaximin in patients with moderate or severe ulcerative colitis refractory to steroid-treatment: A double-blind, placebo-controlled trial. Dig Dis Sci 1999;44:12201.

36. Lobo AJ, Burke DA, Sobala GM et al. Oral tobramycin in ulcerative colitis: effect on maintenance of remission. Aliment Pharmacol Ther 1993;7:1558. 37. Ambrose NS, Allan RN, Keighley MR et al. Antibiotic therapy for treatment in relapse of intestinal Crohns disease. A prospective randomized study. Dis Colon Rectum 1985;28:815. 38. Chapman RW, Selby WS, Jewell DP. Controlled trial of intravenous metronidazole as an adjunct to corticosteroids in severe ulcerative colitis. Gut 1986;27:12102. 39. Goodgame RW, Kimball K, Akram S et al. Randomized controlled trial of clarithromycin and ethambutol in the treatment of Crohns disease. Aliment Pharmacol Ther 2001;15:18616. 40. Swift GL, Srivastava ED, Stone R et al. Controlled trial of anti-tuberculous chemotherapy for two years in Crohns disease. Gut 1994;35:3638. 41. Basilisco G, Ranzi T, Campanini C et al. Controlled trial of rifabutin in Crohns disease. Curr Ther Res 1989;46:24550. 42. Kelleher D, OBrien S, Weir D. Preliminary trial of clofazimine in chronic inflammatory bowel disease. Gut 1982;23:A449. 43. Rutgeerts P, Van Assche G, Vermeire S et al. Ornidazole for prophylaxis of postoperative Crohns disease recurrence: a randomized, doubleblind, placebo-controlled trial. Gastroenterology 2005;128:85661. 44. Rutgeerts P, Hiele M, Geboes K et al. Controlled trial of metronidazole treatment for prevention of Crohns recurrence after ileal resection. Gastroenterology 1995;108:161721. 45. Shaffer JL, Hughes S, Linaker BD et al. Controlled trial of rifampicin and ethambutol in Crohns disease. Gut 1984;25:2035. 46. Prantera C, Zannoni F, Scribano ML et al. An antibiotic regimen for the treatment of active Crohns disease: a randomized, controlled trial of metronidazole plus ciprofloxacin. Am J Gastroenterol 1996;91:32832. 47. Elliott PR, Burnham WR, Berghouse LM et al. Sulphadoxine-pyrimethamine therapy in Crohns disease. Digestion 1982;23:1324. 48. Pineton de Chambrun G, Colombel J-F, Poulain D et al. Pathogenic agents in inflammatory bowel diseases. Curr Opin Gastroenterol 2008;24:4407. 49. Feller M, Huwiler K, Stephan R et al. Mycobacterium avium subspecies paratuberculosis and Crohns disease: a systematic review and metaanalysis. Lancet Infect Dis 2007;7:60713. 50. Pelosini I, Scarpignato C. Rifaximin, a peculiar rifamycin derivative: established and potential clinical uses outside the gastrointestinal tract. Chemotherapy 2005;51 (Suppl 1): 12230. 51. Schoultz I, Verma D, Halfvarsson J et al. Combined polymorphisms in genes encoding the inflammasome components NALP3 and CARD8 confer susceptibility to Crohns disease in Swedish men. Am J Gastroenterol 2000;104:11808. 52. Schoepfer AM, Flogerzi B, Seibold-Schmid B et al. Low Mannanbinding lectin serum levels are associated with complicated Crohns disease and reactivity to oligomannan (ASCA). Am J Gastroenterol 2009; 104:250816. 53. Martinez C, Antolin M, Santos J et al. Unstable composition of the fecal microbiota in ulcerative colitis during clinical remission. Am J Gastroenterol 2008;103:6438. 54. Juni P, Altman DG, Egger M. Assessing the quality of controlled clinical trials. BMJ 2001;323:426. 55. Qin J, Li R, Raes J et al. A human gut microbial gene catalogue established by metagenomic sequencing. Nature 2010;464:5965. 56. Talley NJ, Abreu MT, Achkar J-P et al. An evidence-based systematic review on medical therapies for inflammatory bowel disease. Am J Gastroenterol 2011;106(Suppl 1): S2S25.

2011 by the American College of Gastroenterology

The American Journal of GASTROENTEROLOGY

REVIEW