Antibiotics for the induction and maintenance of remission in ulcerative colitis

Abstract

Background

Antibiotics have been considered to treat ulcerative colitis (UC) due to their antimicrobial properties against intestinal bacteria linked to inflammation. However, there are concerns about their efficacy and safety.

Objectives

To determine whether antibiotic therapy is safe and effective for the induction and maintenance of remission in people with UC.

Search methods

We searched five electronic databases on 10 December 2021 for randomised controlled trials (RCTs) comparing antibiotic therapy to placebo or an active comparator.

Selection criteria

We considered people with UC of all ages, treated with antibiotics of any type, dose, and route of administration for inclusion. Induction studies required a minimum duration of two weeks for inclusion. Maintenance studies required a minimum duration of three months to be considered for inclusion.

Data collection and analysis

We used standard methodological procedures expected by Cochrane. Our primary outcome for induction studies was failure to achieve remission and for maintenance studies was relapse, as defined by the primary studies.

Main results

We included 12 RCTs (847 participants). One maintenance of remission study used sole antibiotic therapy compared with 5‐aminosalicylic acid (5‐ASA). All other trials used concurrent medications or standard care regimens and antibiotics as an adjunct therapy or compared antibiotics with other adjunct therapies to examine the effect on induction of remission.

There is high certainty evidence that antibiotics (154/304 participants) compared to placebo (175/304 participants) result in no difference in failure to achieve clinical remission (risk ratio (RR) 0.88, 95% confidence interval (CI) 0.74 to 1.06). A subgroup analysis found no differences when steroids, steroids plus 5‐ASA, or steroids plus 5‐ASA plus probiotics were used as additional therapies to antibiotics and placebo.

There is low certainty evidence that antibiotics (102/168 participants) compared to placebo (121/175 participants) may result in no difference in failure to achieve clinical response (RR 0.75, 95% CI 0.47 to 1.22). A subgroup analysis found no differences when steroids or steroids plus 5‐ASA were used as additional therapies to antibiotics and placebo.

There is low certainty evidence that antibiotics (6/342 participants) compared to placebo (5/349 participants) may result in no difference in serious adverse events (RR 1.19, 95% CI 0.38 to 3.71). A subgroup analysis found no differences when steroids were additional therapies to antibiotics and placebo.

There is low certainty evidence that antibiotics (3/342 participants) compared to placebo (1/349 participants) may result in no difference in withdrawals due to adverse events (RR 2.06, 95% CI 0.27 to 15.72). A subgroup analysis found no differences when steroids or steroids plus 5‐ASA were additional therapies to antibiotics and placebo.

It is unclear if there is any difference between antibiotics in combination with probiotics compared to no treatment or placebo for failure to achieve clinical remission (RR 0.68, 95% CI 0.39 to 1.19), serious adverse events (RR 1.00, 95% CI 0.07 to 15.08), or withdrawals due to adverse events (RR 1.00, 95% CI 0.07 to 15.08). The certainty of the evidence is very low.

It is unclear if there is any difference between antibiotics compared to 5‐ASA for failure to achieve clinical remission (RR 2.20, 95% CI 1.17 to 4.14). The certainty of the evidence is very low.

It is unclear if there is any difference between antibiotics compared to probiotics for failure to achieve clinical remission (RR 0.47, 95% CI 0.23 to 0.94). The certainty of the evidence is very low.

It is unclear if there is any difference between antibiotics compared to 5‐ASA for failure to maintain clinical remission (RR 0.71, 95% CI 0.47 to 1.06). The certainty of the evidence is very low.

It is unclear if there is any difference between antibiotics compared to no treatment for failure to achieve clinical remission in a mixed population of people with active and inactive disease (RR 0.56, 95% CI 0.29 to 1.07). The certainty of the evidence is very low.

For all other outcomes, no effects could be estimated due to a lack of data.

Authors' conclusions

There is high certainty evidence that there is no difference between antibiotics and placebo in the proportion of people who achieve clinical remission at the end of the intervention period. However, there is evidence that there may be a greater proportion of people who achieve clinical remission and probably a greater proportion who achieve clinical response with antibiotics when compared with placebo at 12 months. There may be no difference in serious adverse events or withdrawals due to adverse events between antibiotics and placebo.

No clear conclusions can be drawn for any other comparisons.

A clear direction for future research appears to be comparisons of antibiotics and placebo (in addition to standard therapies) with longer‐term measurement of outcomes. Additionally. As there were single studies of other head‐to‐head comparisons, there may be scope for future studies in this area.

Plain language summary

The use of antibiotics in the treatment of ulcerative colitis

What was the aim of this review?

This Cochrane Review aimed to determine whether antibiotics are useful and safe for people with ulcerative colitis (UC) in achieving or maintaining remission.

Key messages

We found that antibiotics, when added to standard therapies, are similar to placebo in how often they fail to achieve remission or disease improvement for people with UC, measured up to three months from taking them.

We found that antibiotics when compared to placebo may be useful in achieving remission or some improvement of UC symptoms, measured at 12 months from taking them.

In terms of safety, the evidence is of low certainty, but antibiotics may be as safe as placebo.

We have confidence in the findings up to three months, but have less confidence in the evidence at 12 months due to low numbers of study participants.

What is ulcerative colitis?

UC is a chronic (long term) inflammatory disease that affects the large bowel. The most common symptoms include bloody diarrhoea, abdominal pain, and a sudden almost uncontrollable urge to pass stool. Some people have other symptoms including sore joints, mouth sores, and inflammation in their eyes. When someone is experiencing symptoms of UC, they are said to have 'active' disease. When symptoms improve with treatment, the disease is said to be responding to therapy. When symptoms stop, the disease is said to be in 'remission'. People with UC in remission are often given therapy with medicines to try and prolong (maintain) their remission or response to therapy.

What did the review study?

This review studied the use of antibiotics in the treatment of UC. Eleven studies considered antibiotics as a therapy for achieving remission and one study as therapy for maintaining remission.

Antibiotics could theoretically be helpful as they might be able to fight harmful bacteria in the gut that are thought to be a partial cause for the presence of UC symptoms.

Currently, most doctors do not recommend the use of antibiotics unless the patient has an infection.

What were the main results of the review?

We searched for randomised controlled trials (RCTs; clinical studies where people are randomly put into one of two or more treatment groups) comparing antibiotics with another treatment or dummy treatment. We found 12 RCTs including 847 participants. The trials looked at people of all ages; using different types and doses of antibiotics; and at antibiotics taken as tablets, injections, or any other way. The main results from our review are as follows.

1. Antibiotics, when added to standard therapies, are no different from placebo in failing to achieve remission or improve disease symptoms for people with UC when measured up to three months from taking them.

2. Antibiotics probably fail less than placebo in achieving remission or some improvement of UC symptoms measured at 12 months from taking them.

3. In terms of safety, antibiotics may be similar to placebo, but the evidence is of low certainty.

4. We are uncertain about how antibiotics compare to other treatments because there is not enough good evidence to make a judgement.

Conclusion

There is no difference between antibiotics and placebo when added to standard therapies to achieve remission or improvement of symptoms in the short term. However, they may help to lessen the number of people who do not achieve remission, and they probably lead to fewer numbers of people who have not achieved improvement of symptoms at 12 months.

No conclusions can be drawn on other treatments or safety due to lack of evidence, even though there may be a possibility that antibiotics are as safe as placebo. Further research would be needed to clarify the usefulness of antibiotics in the treatment of UC; however, such studies should focus on longer‐term outcomes given the findings of this review.

How up‐to‐date is this review?

This review is up‐to‐date as of December 2021.

Background

Description of the condition

Ulcerative colitis (UC) is a chronic inflammatory condition affecting the colon, characterised by periods of relapse and remission (Ordás 2012). Symptoms can include persistent diarrhoea, which may often be bloody, problems with defecation, abdominal pain, rectal pain and bleeding, weight loss, and fatigue. It starts in the rectum and extends proximally throughout the colon. UC and Crohn's disease (CD) are the two most common types of inflammatory bowel diseases (IBDs). UC tends to be more common than CD, with an estimated prevalence of 90 to 505 cases per 100,000 people in North America and Northern Europe (Conrad 2014). While the prevalence of UC has been historically higher in Western countries, its incidence in industrialised parts of Asia and Latin America is on the rise. The cause of UC is unknown but is believed to be associated with specific genetic and environmental factors. The risk of developing UC is higher in Ashkenazi Jews, people with a family history of the disease, and those who live in Western countries (da Silva 2014).

Description of the intervention

It has been proposed that there is a link between increased intestinal bacterial concentrations and chronic inflammation (Swidsinksi 2002; Vrakas 2017). Studies have suggested bacterial pathogens, such as Escherichia coli, Bacteroides spp, and Mycobacterium avium, are involved in the pathogenesis of UC. Antibiotics may influence the course of UC by decreasing the number of bacteria in the gut and altering the composition of the microbiome. Specific antibiotics that have been used for this purpose include ciprofloxacin, metronidazole, rifaximin, anti‐tuberculous regimens, and antibiotic combinations (Nitzan 2016). Oral administration of antibiotics is the most common route. For induction of remission, antibiotics can be taken for a few days up to one month. For maintenance of remission, antibiotics can be taken for any length, ranging from a few months to over one year.

How the intervention might work

Several antibiotics have been evaluated for the treatment of UC. Reducing the bacterial load in the intestinal mucosa may reduce the pathogenic immune response. Furthermore, antibiotics also act to limit bacterial translocation and reduce the concentration of bacteria adherent to the lumen or mucosa (Scribano 2013). In people with high levels of E coli in their microbiome, treatment with mesalamine showed a decrease in intestinal inflammation. This further suggests the crucial role the gut microbiome may have in IBD pathophysiology and the potential for antimicrobial agents may have as a treatment (Kostic 2014). There is a possibility that alteration of the mucosal flora may have a therapeutic role in UC by inhibiting the stimulus for pathogenic immune responses (Ott 2004; Swidsinksi 2002).

Why it is important to do this review

Given the possible role of the bacterial load in the pathogenesis of UC, it is reasonable to suggest that antibiotic therapy may be effective for the induction or maintenance (or both) of remission in UC. However, several potential problems exist with this approach. First, the use of broad‐spectrum antibiotics is a very generalised strategy that may aggravate dysbiosis. Second, the resident flora is determined by both genetic and dietary factors that may be difficult or impossible to modify on a chronic basis; therefore, treatment (if effective), may have to be continued indefinitely. Finally, broad‐spectrum antibiotic therapy is associated with important adverse effects, notably an increased risk of C diff (Clostridioides difficile, previously known as Clostridium difficile) infection. For these reasons, evidence from high‐quality randomised controlled trials (RCTs) is necessary before antibiotic therapy is accepted as an effective and safe treatment for UC.

No evidence‐based recommendations exist regarding the antibiotic of choice, dose, or duration of treatment for UC. Despite limited supporting evidence, the most recent guidelines published by the World Gastroenterology Organisation support the use of antibiotics in perianal disease, fistulising disease, and bacterial overgrowth secondary to stricturing disease (Bernstein 2016). One previous meta‐analysis on this topic suggested that adjunctive antibacterial therapy is effective for inducing clinical remission in UC (Rahimi 2007). However, recently published international guidelines do not recommend the routine use of antibiotics for induction and maintenance of remission in the management of UC, since trials of oral or intravenous antibiotics have not demonstrated any consistent benefit compared to conventional therapies, and a possible role is only suggested in cases of infection or prior to surgery (Feuerstein 2020; Harbord 2017; Rubin 2019; Singh 2020). This Cochrane Review aims to examine the current evidence from RCTs that investigate the role of antibiotics in the induction and maintenance of remission in UC.

Objectives

To determine whether antibiotic therapy is safe and effective for the induction and maintenance of remission in people with UC.

Methods

Criteria for considering studies for this review

Types of studies

We included only studies with a randomised study design (RCTs) and cluster RCTs.

Cross‐over RCTs were only considered for inclusion if the results were presented separately for each stage of the trial, or if we could obtain them after contacting the trial authors.

Types of participants

We considered participants of any age, with a diagnosis of UC, confirmed by any established method, who had disease at any stage or any extent for inclusion. We considered participants with active or quiescent UC (as defined by the original studies). Trials were in any setting (e.g. single‐centre or multicentre).

Types of interventions

We considered any class of antibiotic, delivered via any route (i.e. oral, intravenous). Trials that compared antibiotic therapy to a placebo or an active comparator were suitable for inclusion. We considered antibiotics used as adjunctive therapy or as monotherapy.

Studies on induction of remission must have provided outcome measurements of at least two weeks from the beginning of the intervention to be considered for inclusion.

Studies on the maintenance of remission must have provided outcome measurements of at least three months from the beginning of the intervention to be considered for inclusion.

We excluded studies in which placebo or comparator groups had recent antibiotic exposure (i.e. within the past six months).

Types of outcome measures

Both dichotomous and continuous outcomes were considered for inclusion.

Primary outcomes

Studies on the induction of remission

1. Proportion of participants who failed to achieve remission, as defined by the original studies.

Studies on the maintenance of remission

1. Proportion of participants who relapsed, as defined by the included studies.

Secondary outcomes

Efficacy outcomes

1. Proportion of participants who failed to achieve clinical response (as defined by the original studies).

2. Proportion of participants who failed to achieve endoscopic remission (as defined by the original studies).

3. Proportion of participants who failed to achieve histological remission (as defined by the original studies).

4. Proportion of participants who had a clinical relapse (as defined by the original studies).

5. Proportion of participants who had an endoscopic relapse (as defined by the original studies).

6. Health‐related quality of life (as measured by a validated quality‐of‐life instrument).

Safety outcomes

1. Proportion of participants with any adverse event (AE).

2. Proportion of participants with serious adverse events (SAEs), as defined by the original studies.

3. Proportion of participants who withdrew from the study due to AEs.

Search methods for identification of studies

Electronic searches

We searched the following databases for relevant studies. The search was performed on 10 December 2021.

1. Cochrane Central Register of Controlled Trials (CENTRAL) and Cochrane Database of Systematic Reviews (CDSR) (via Ovid Evidence‐Based Medicine Reviews Database (EBMR), inception to present) (Appendix 1).

2. MEDLINE (via Ovid, 1946 to present) (Appendix 2).

3. Embase (via Ovid, 1974 to present) (Appendix 3).

4. ClinicalTrials.gov (www.clinicaltrials.gov).

5. World Health Organization International Clinical Trials Registry Platform (ICTRP; trialsearch.who.int/).

Searching other resources

We searched the references listed within relevant studies and review articles for additional citations not identified in the search. Furthermore, we searched the conference proceedings from major meetings (Digestive Disease Week, the European Crohn's and Colitis Organisation (ECCO) congress, and the United European Gastroenterology Week conference) from 2017 to 2021, for studies published in abstract form only.

When the search identified a relevant abstract, we requested details of the full study methodology and results from the authors to allow a thorough assessment of the quality of identified studies. We excluded abstracts for which this information could not be obtained.

Data collection and analysis

Selection of studies

Two review authors (CGC and VS) independently reviewed each article at each stage of selection including title screening, abstract screening, and full‐text review.

Step 1. Two review authors screened and selected titles using the described search strategy that appeared to be relevant to this review. Adjudication did not occur at the title screening stage, thus any study that was ambiguous at this stage proceeded to step 2 by default.

Step 2. Abstract screening involved the selection of articles that reported studies (or abstracts) with a reasonable possibility of inclusion. We resolved differences in assessment for inclusion by discussion between the two review authors. Adjudication did not occur at the abstract screening stage.

Step 3. Full‐text review involved the selection of articles based on careful examination of the full‐report against the inclusion criteria. We resolved differences in assessment for inclusion by discussion between the two review authors. One review author (AG) performed adjudication as needed.

Data extraction and management

Two review authors (CGC and VS) independently extracted and recorded the data using a data extraction form. Extracted data included the following items.

1. Study design: type of RCT, setting, number of interventions, year, author contact details.

2. Population characteristics: age, sex, disease distribution, disease duration, site of disease, the total number of participants originally assigned to each treatment group.

3. Intervention: class of antibiotic, route of administration, dose of antibiotic, and length of regimen.

4. Control: no active treatment, placebo, active comparator (type/class, route, dose, length).

5. Concurrent medications.

6. Outcomes: time of assessment, length of follow‐up, the definitions of clinical remission, endoscopic remission, histological remission, the proportion of participants failing to achieve remission (for remission RCTs); the definitions of clinical and endoscopic relapse, the proportion of participants relapsing (for maintenance RCTs).

Assessment of risk of bias in included studies

Two review authors (CGC and VS) independently assessed the risk of bias using the Cochrane RoB 1 tool, as outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). A third review author (MG) performed adjudication as needed. We assessed the following domains as having a low, moderate, high, or unclear risk of bias.

1. Sequence generation (i.e. was the allocation sequence adequately generated?).

2. Allocation concealment (i.e. was allocation concealment concealed?).

3. Blinding of participants and personnel and outcome assessors (i.e. was knowledge of the allocated intervention adequately prevented during the study?).

4. Incomplete outcome data (i.e. were incomplete outcome data adequately addressed?).

5. Selective outcome reporting (i.e. were reports of the study free of suggestion of selective outcome reporting?).

6. Other potential sources of bias (i.e. was the study apparently free of other problems that could have put it at a high risk of bias?).

We reviewed each outcome to determine the overall certainty of evidence supporting the outcome, classified as high certainty (the estimate of effect is very unlikely to be changed despite further research); moderate certainty (the estimate of effect is unlikely to be changed despite further research); low certainty (the estimate of effect may be changed despite further research), or very low certainty (the estimate of effect likely will be changed with further research).

Measures of treatment effect

We used Review Manager 5 to analyse the data on an intention to treat (ITT) basis (Review Manager 2020). For dichotomous outcomes, we calculated the risk ratio (RR) estimates and associated two sided 95% confidence intervals (CI). For continuous outcomes, we calculated the mean difference (MD) with 95% CIs when all studies reported outcomes using the same scale, and standardised mean difference (SMD) with 95% CIs when studies used different scales to report their outcomes.

Unit of analysis issues

We included cross‐over trials if data were available for the first phase of the trial prior to cross‐over. To deal with events that may re‐occur (e.g. AEs), we reported the proportion of participants who experienced at least one AE. We performed separate comparisons for studies that compared antibiotics to placebo and for studies that compared antibiotics to other active therapies. We also performed separate comparisons for each type of antibiotic. Where we encountered multiple treatment groups (e.g. for different doses of antibiotics), we divided the placebo group across the treatment groups or combined groups to create a single pair‐wise comparison, as appropriate.

Dealing with missing data

We used An ITT analysis for dichotomous outcomes whereby participants with missing outcomes were assumed to be treatment failures. Sensitivity analyses assessed the impact of this assumption on the effect estimate.

Assessment of heterogeneity

We assessed heterogeneity using the Chi² test (a P value of 0.10 was statistically significant) and the I² statistic. We considered an I² statistic greater than 75% to indicate high heterogeneity among study data, greater than 50% indicated moderate heterogeneity, and greater than 25% indicated low heterogeneity (Higgins 2003). We conducted sensitivity analyses to explore possible explanations for heterogeneity.

Assessment of reporting biases

We initially compared the outcomes listed in the study protocol to those reported in the published manuscript. If we did not have access to the protocol, we used the outcomes listed in the methods sections of the published manuscript and compared this to what was reported in the results section. If any pooled analyses included 10 or more studies, we investigated potential publication bias using funnel plots (Egger 1997).

Data synthesis

We combined data for meta‐analysis from individual trials when the interventions, patients groups, and outcomes were similar, as deemed by review author consensus. We used a fixed‐effect model to pool data unless there was significant heterogeneity between the studies. We used a random‐effects model if heterogeneity existed (I² = 50% to 75%). We did not pool data for meta‐analysis if there was a high degree of heterogeneity (I² of 75% or greater).

Subgroup analysis and investigation of heterogeneity

We planned subgroup analysis, provided there were adequate data, for the following factors.

1. Age of participants (children versus adults).

2. Different antibiotic dose.

3. Duration of the intervention.

4. Sex (male versus female).

5. Route of administration (oral versus intravenous).

Sensitivity analysis

Where possible, we planned to undertake a sensitivity analysis for the primary outcome of 'failure to achieve remission' for induction studies and 'relapse' for maintenance studies, to assess whether the findings of the review are robust to the decisions made during the review process. In particular, we planned to exclude from the analyses studies at high or unclear risk of selection bias (due to the method of allocation concealment) and performance bias. Where data analyses included studies with reported and estimated standard deviations (SD), we planned to exclude those with estimated SDs to assess whether this affected the findings of the review. We also planned to investigate whether the choice of model (fixed‐effect versus random‐effects) may have affected the results. However, due to the heterogeneous studies presented that limited scope for meta‐analysis, these were not completed.

Summary of findings and assessment of the certainty of the evidence

We presented the main results (clinical remission at study end, clinical response at study end, SAEs and withdrawals due to AEs) in summary of findings tables. Each comparison and primary outcome was exported to GRADEpro GDT software (developed by the GRADE Working Group) for quality assessment. Based on the risk of bias, inconsistency, imprecision, indirectness, and publication bias, we graded the certainty of the evidence for each outcome as high, moderate, low, or very low. These ratings were as follows.

1. High certainty. Further research is very unlikely to change our confidence in the estimate of effect.

2. Moderate certainty. Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.

3. Low certainty. Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.

4. Very low certainty. Any estimate of effect is very uncertain.

Results

Description of studies

Results of the search

The literature search identified 6279 records through database searching and alternative sources. After removal of duplicates, 6102 unique records remained. Examination of the titles and abstracts left 71 records for full text screening. After assessing all 71 records, we identified 14 records of 12 studies that met the inclusion criteria and these were included in the review (Characteristics of included studies table).

We excluded 44 studies (46 records; Characteristics of excluded studies table).

We identified one ongoing study (Characteristics of ongoing studies table) and a further 10 records (eight RCTs) are awaiting classification (Characteristics of studies awaiting classification table).

The results of the search are presented in the PRISMA flow diagram (Figure 1).

1.

Study flow diagram.

Included studies

Setting

We identified 12 RCTs meeting our inclusion criteria (Bataga 2015; Burke 1990; Gilat 1987; Gilat 1989; Mantzaris 1994; Mantzaris 1997; Mantzaris 2001; Mishra 2020; Ohkusa 2005; Ohkusa 2010; Petersen 2014; Turunen 1998). Nine were single‐centre studies (Bataga 2015; Burke 1990; Mantzaris 1994; Mantzaris 1997; Mantzaris 2001; Mishra 2020; Ohkusa 2005; Petersen 2014; Turunen 1998); and three were multicentre studies (Gilat 1987; Gilat 1989; Ohkusa 2010). The studies were conducted in Romania (Bataga 2015), the UK (Burke 1990), Israel (Gilat 1987; Gilat 1989), Greece (Mantzaris 1994; Mantzaris 1997; Mantzaris 2001), India (Mishra 2020), Japan (Ohkusa 2005; Ohkusa 2010), Denmark (Petersen 2014), and Finland (Turunen 1998).

Participants

The 12 RCTs included 847 participants. The mean age of participants ranged from 33.0 years in Mishra 2020 to 53.2 years in Bataga 2015.

All studies except Turunen 1998 described the extent of disease in participants, which included left sided disease, proctitis, total colitis, and proctosigmoiditis.

Five studies reported that they included participants with mild disease (Bataga 2015; Burke 1990; Mantzaris 1997; Ohkusa 2005; Ohkusa 2010), five with moderate disease (Burke 1990; Mantzaris 1997; Ohkusa 2005; Ohkusa 2010; Turunen 1998), and six with severe disease (Burke 1990; Mantzaris 1994; Mantzaris 2001; Mishra 2020; Ohkusa 2010; Turunen 1998). Gilat 1987 mentioned that they included people with active non‐severe UC. Petersen 2014 described their included participants as having active UC with a Clinical Activity Index (CAI) of 6 or greater.

Gilat 1989, as a maintenance study, only included people in remission.

Mean disease duration ranged from two years in Mishra 2020 to 12 years in Ohkusa 2005.

Eight of the induction studies included people experiencing a relapse and people presenting with a first episode of UC (Burke 1990; Gilat 1987; Mantzaris 1994; Mantzaris 1997; Mantzaris 2001; Mishra 2020; Petersen 2014; Turunen 1998), while the other three only included people experiencing a relapse of their UC (Bataga 2015; Ohkusa 2005; Ohkusa 2010).

Induction/maintenance of remission

Eight studies examined the effect of antibiotic therapy on the induction of remission (Bataga 2015; Burke 1990; Gilat 1987; Mantzaris 1994; Mantzaris 1997; Mantzaris 2001; Mishra 2020; Petersen 2014).

One study examined the effect of antibiotic therapy on the maintenance of remission (Gilat 1989).

Three studies reported that they examined both induction and maintenance (Ohkusa 2005; Ohkusa 2010; Turunen 1998). However, as these studies followed up the same participant cohorts from the beginning to the end of the study, without re‐randomising after remission was induced for the maintenance part of their studies, we could not consider the effect of their interventions as therapy for maintenance of remission. Therefore, we only considered Ohkusa 2010 and Turunen 1998 as induction studies, extracting our outcomes as per our protocol (Gordon 2020).

Ohkusa 2005 included a mix of participants with active and inactive disease, as per our communication with the author. Our team considered whether we should categorise this study as an induction study, maintenance study, or whether we should exclude it from our review. We decided that we could not exclude it, as it was not explicitly violating our inclusion criteria; however, we could neither consider it as a study on the induction of remission nor the maintenance of remission, as the authors could not provide us with separate data for the study participants in remission and those with active disease. Therefore, we presented this study in a separate category of studies with a mix of participants with active and inactive disease.

Add‐on therapies

All studies except two did not examine the use of antibiotics as the sole therapy for the induction or maintenance of remission. Instead, antibiotics were used as an add‐on therapy. The two exceptions to this were Gilat 1987 and Gilat 1989.

Interventions

For induction of remission

1. Antibiotics plus steroids versus placebo plus steroids (Burke 1990; Mantzaris 1994; Mantzaris 2001; Mishra 2020).

2. Antibiotics plus steroids plus 5‐aminosalicylic acid (5‐ASA) versus placebo plus steroids plus 5‐ASA (Mantzaris 1997; Ohkusa 2010; Petersen 2014 (in one arm of the four‐arm intervention); Turunen 1998).

3. Antibiotics plus probiotics plus 5‐ASA versus steroids plus 5‐ASA (Bataga 2015).

4. Antibiotics versus 5‐ASA (Gilat 1987).

5. Antibiotics plus probiotics plus steroids plus 5‐ASA versus antibiotics plus placebo plus steroids plus 5‐ASA versus probiotics plus placebo plus steroids plus 5‐ASA versus placebo plus placebo plus steroids plus 5‐ASA (Petersen 2014).

For maintenance of remission

1. Antibiotics versus 5‐ASA (Gilat 1989).

Studies on a mixed population of participants with active and inactive disease

1. Antibiotics plus steroids plus 5‐ASA versus steroids plus 5‐ASA (Ohkusa 2005).

Therapy details for induction studies

Four studies compared antibiotics plus steroids versus placebo plus steroids.

1. Burke 1990 compared oral tobramycin as an adjunct to steroid therapy (prednisone or hydrocortisone) versus placebo as an adjunct to steroid therapy (participants receiving sulfasalazine were allowed to continue receiving it).

2. Mantzaris 1994 compared oral metronidazole plus hydrocortisone therapy versus placebo plus hydrocortisone therapy. Participants also received parenteral nutrition and were not allowed to eat or drink anything except water. Electrolytes losses were substituted and blood was transfused to maintain haemoglobin levels.

3. Mantzaris 2001 compared intravenous ciprofloxacin plus hydrocortisone therapy versus placebo plus hydrocortisone therapy. Participants received parenteral nutrition and were not allowed to eat or drink anything except water and oral glucose solutions. Care was taken to substitute for electrolyte losses. Blood was transfused to maintain haemoglobin levels.

4. Mishra 2020 compared intravenous ceftriaxone plus metronidazole plus standard care versus placebo plus standard care. Standard care consisted of steroids, prophylactic anticoagulation, and oral nutrition for both groups.

Four studies compared antibiotics plus steroids plus 5‐ASA versus placebo plus steroids plus 5‐ASA.

1. Mantzaris 1997 compared oral ciprofloxacin plus olsalazine versus placebo plus olsalazine. Participants with a mild or moderate exacerbation of UC received oral prednisolone. All participants received betamethasone enemas.

2. Ohkusa 2010 compared an oral combination of amoxicillin plus tetracycline plus metronidazole versus placebo. Doses of any concomitant oral or rectal medication for UC remained constant, except for corticosteroids, which were tapered until discontinuation.

3. Petersen 2014 compared oral ciprofloxacin plus prednisolone plus mesalamine versus placebo plus prednisolone plus mesalamine, as one part of their four‐arm intervention. Prednisolone was administered to all participants for up to 12 weeks. If disease activity was limited to the rectum, rectal administration of corticosteroids and mesalamine was used when steroids were tapered off. Hospitalisation and intravenous treatment were used at entry if necessary. Mesalamine administration was started or continued at entry and used as a maintenance treatment during the entire study period.

4. Turunen 1998 compared oral ciprofloxacin plus prednisolone plus mesalamine therapy versus placebo plus prednisolone plus mesalamine therapy. Prednisolone was administered to all participants and tapered, if possible, meaning that in cases of partial response or recurrence of the symptoms, prednisolone was reduced more slowly or increased moderately. If disease activity was limited to the rectum, rectal administration of corticosteroids and mesalamine was used when steroids were tapered to zero. Mesalamine administration was started or continued at entry and used as a maintenance treatment during the entire study period.

One study compared antibiotics plus probiotics plus 5‐ASA versus steroids plus 5‐ASA (Bataga 2015). More specifically, it compared oral rifaximin plus probiotics (Lactobacillus acidophilus, Bifidobacterium infantis, and Enterococcus faecium) versus budesonide. All participants also received mesalazine 3 g/day.

One study compared antibiotics versus 5‐ASA (Gilat 1987). More specifically, it compared oral metronidazole versus sulfasalazine.

One study compared oral antibiotics plus probiotics plus steroids plus 5‐ASA versus antibiotics plus placebo plus steroids plus 5‐ASA versus probiotics plus placebo plus steroids plus 5‐ASA versus placebo plus placebo plus steroids plus 5‐ASA (Petersen 2014). More specifically, it compared ciprofloxacin plus EcN (E coli Nissle) versus ciprofloxacin plus placebo versus probiotic plus placebo versus placebo plus placebo. Prednisone was administered to all participants for up to 12 weeks. If disease activity was limited to the rectum, rectal administration of corticosteroids and mesalamine was used when steroids were tapered to zero. Hospitalisation and intravenous treatment were used at entry if necessary. Mesalamine administration was started or continued at entry and used as a maintenance treatment during the entire study period.

Therapy details for maintenance studies

One study compared antibiotics (metronidazole) versus 5‐ASA (sulfasalazine) (Gilat 1989).

More details about the numbers of participants, length of the interventions, and participants' disease activity for the included studies can be found in Table 7, Table 8, and Table 9, and a more detailed description of the interventions is presented in the Characteristics of included studies tables.

1. Characteristics of included studies for induction of remission.

Comparison	Study ID	Intervention	Number of participants	Length of intervention	Disease activity at beginning of study	Definition of remission/relapse	Conflicts of interest
Antibiotics vs placebo (antibiotics + steroids vs placebo + steroids)	Burke 1990	Oral tobramycin + steroids vs oral placebo + steroids	84	7 days of therapy and follow‐up at 21 or 28 days depending on subgroup.	Mild to severe	Participants were considered to be in symptomatic remission if their bowel actions were < 3/day without blood, there was no malaise, abdominal pain, anorexia, vomiting, abdominal tenderness, fever, or any complication or active extraintestinal manifestation of ulcerative colitis, based on the Truelove and Witts criteria (Truelove 1955).	None
Antibiotics vs placebo (antibiotics + steroids vs placebo + steroids)	Mantzaris 1994	Oral metronidazole + hydrocortisone therapy vs oral placebo + hydrocortisone therapy. Participants also received parenteral nutrition and electrolytes.	39	10 days. Follow‐up at 8 months to check how many participants were in remission.	Severe	Participants were classified as responders or non‐responders by the following method: in brief, participants passing ≤ 3 formed stools or forming stool daily without rectal bleeding or systemic signs of severe colitis were considered to be improved (responders).	None
Antibiotics vs placebo (antibiotics + steroids vs placebo + steroids)	Mantzaris 2001	Intravenous ciprofloxacin + hydrocortisone therapy vs intravenous placebo + hydrocortisone therapy. Participants also received parenteral nutrition and electrolytes.	55	10 days. Follow‐up at 3 months to check how many participants were in remission.	Severe	Participants passing ≤ 3 forming or formed stools daily without any visible blood or systemic signs of severe colitis, were considered to be improved (responders) and were given food orally and oral corticosteroids. Participants who either did not improve or deteriorated progressively were considered as non‐responders.	None
Antibiotics vs placebo (antibiotics + steroids vs placebo + steroids)	Mishra 2020	Intravenous ceftriaxone + metronidazole + steroids vs intravenous placebo + steroids. Participants also received oral nutrition and prophylactic anticoagulation.	50	5 days of therapy and follow‐up 28 days	Severe	Clinical response as defined by the Oxford criteria (Travis 1996) at day 3 of therapy: complete response = stool frequency < 3; partial response = stool frequency 3–8 and CRP < 45 mg/L; no response = stool frequency > 8 or stool frequency 3–8 with CRP > 45 mg/L.	None
Antibiotics vs placebo (antibiotics + steroids + 5‐ASA vs placebo + steroids + 5‐ASA)	Mantzaris 1997	Ciprofloxacin + olsalazine/steroid therapy vs placebo + olsalazine/steroid therapy	70	14 days. Follow‐up at 7–9 weeks to check how many participants were in remission.	Mild to moderate	Participants passing formed stools without rectal bleeding, with a normal haematological and biochemical profile, a normal sigmoidoscopic appearance and no histological evidence of acute and chronic inflammation on rectal biopsies were considered to be in remission (responders). Participants who either did not achieve remission, or who deteriorated or developed complications during the study were considered as non‐responders.	None
Antibiotics vs placebo (antibiotics + steroids + 5‐ASA vs placebo + steroids + 5‐ASA)	Ohkusa 2010	Amoxicillin + tetracycline + metronidazole vs placebo	210	2 weeks. Follow‐up at 3 months and at 12 months from the beginning of the study to check how many participants were in remission	Mild to severe relapse	Clinical remission was defined as a total Mayo score of ≤ 2 points with no individual subscore exceeding 1 point. Clinical relapse was defined as reappearance of visible blood in stools for 2 consecutive days or recurrence of frequent diarrhoea (≥ 5 bowel movements/day), nocturnal diarrhoea, abdominal cramps, or a combination of these.	None
Antibiotics vs placebo (antibiotics + steroids + 5‐ASA vs placebo + steroids + 5‐ASA)	Turunen 1998	Ciprofloxacin + prednisone therapy vs placebo + prednisone therapy. Certain participants also received mesalamine.	85	6 months. Follow‐up at 12 months from the beginning of the study to check how many participants were in remission	Moderate to severe	Only response was considered. Treatment failure, the primary endpoint, was defined as both symptomatic and endoscopic failure to respond.	Sponsored by industry.
Antibiotics vs placebo (antibiotics + steroids + 5‐ASA vs placebo + steroids + 5‐ASA)	Petersen 2014 (1 arm of the 4‐arm intervention)	Ciprofloxacin + prednisone + mesalamine vs placebo + prednisone + mesalamine	50	Antibiotics and its placebo equivalent were given for 1 week. Assessments at week 0, 1, 4, 8, and 12	Active with CAI ≥ 6	Remission was defined as CAI score ≤ 4	None
Antibiotics + probiotics vs no intervention/placebo (antibiotics + probiotics + 5‐ASA vs steroids + 5‐ASA)	Bataga 2015	Rifaximin + probiotics + mesalazine vs budesonide + mesalazine	48	1 month of rifaximin + 10 days of probiotics and 3 months for budesonide. Follow‐up for both at 6 months to check how many participants were in remission	Minor relapse	Remission defined as UCDAI score 0–2	Not mentioned
Antibiotics + probiotics vs no intervention/placebo (antibiotics + probiotics + steroids + 5‐ASA vs placebo + steroids + 5‐ASA)	Petersen 2014 (1 arm of the 4‐arm intervention)	Ciprofloxacin + probiotic vs placebo. Participants also received prednisolone and some participants mesalamine or rectal steroids, or both.	50	Antibiotics and its placebo equivalent were given for 1 week, probiotics and its placebo equivalent for 7 weeks. Assessments at week 0, 1, 4, 8, and 12.	Active with CAI ≥ 6	Remission defined as CAI score ≤ 4.	None
Antibiotics vs 5‐ASA	Gilat 1987	Metronidazole vs sulfasalazine	46	28 days	Non‐severe	Only improvement of clinical symptoms (blood in stools, number of bowel movements) and sigmoidoscopy grade was considered.	Not mentioned
Antibiotics vs probiotics	Petersen 2014 (1 arm of the 4‐arm intervention)	Ciprofloxacin + placebo vs probiotic + placebo. Participants also received prednisolone and some participants mesalamine or rectal steroids.	50	Antibiotics and its placebo equivalent were given for 1 week, probiotics and its placebo equivalent for 7 weeks. Assessments at week 0, 1, 4, 8, and 12.	Active with CAI ≥ 6	Remission was defined as CAI score ≤ 4	None

5‐ASA: 5‐aminosalicylic acid; CAI: Clinical Activity Index; CG: control group; CRP: C‐reactive protein; IG: intervention group; UCDAI: Ulcerative Colitis Disease Activity Index.

2. Characteristics of included studies for maintenance of remission.

Comparison	Study ID	Intervention	Number of participants	Length of intervention	Disease activity at beginning of study	Definition of remission/relapse	Conflicts of interest
Antibiotics vs 5‐ASA	Gilat 1989	Metronidazole vs sulfasalazine	40	12 months with follow‐ups every 3 months or until recurrence.	In remission	Relapse was defined as the appearance of blood in the stools with an increase in stool frequency, with or without diarrhoea.	Not mentioned

5‐ASA: 5‐aminosalicylic acid.

3. Characteristics of studies on a mixed population of people with active and inactive disease.

Comparison	Study ID	Intervention	Number of participants	Length of intervention	Disease activity at beginning of study	Definition of remission/relapse	Conflicts of interest
Antibiotics vs no treatment (antibiotics + steroids + 5‐ASA vs steroids + 5‐ASA)	Ohkusa 2005	Amoxicillin + tetracycline + metronidazole vs no antibiotics	20	2 weeks. Follow‐up at 3 and 12 months from the beginning of the study to check how many participants were in remission.	Mix of active and inactive based on communication with the author (in the published version it is described as people with mild to moderate relapse). Intervention group: 7/10 participants in remission, 3/10 participants with active disease. Control group: 1/10 participants in remission, 9/10 participants with active disease.	Remission was defined as Lichtiger symptom score < 3 (Lichtiger 1990)	None

5‐ASA: 5‐aminosalicylic acid.

Therapy details for studies on a mixed population of participants with active and inactive disease

One study compared antibiotics plus steroids plus 5‐ASA versus steroids plus 5‐ASA (Ohkusa 2005). More specifically, it compared a combination of oral amoxicillin plus tetracycline plus metronidazole versus no treatment. Any concomitant medication for UC, such as sulfasalazine, 5‐ASA, prednisolone, probiotics, or a combination of these being taken at the start of the study was continued at a stable dose for a set time prior to inclusion in the study.

Funding sources and conflicts of interest

Five studies reported their sources of funding. One received funding from pharmaceutical companies (Turunen 1998); three from universities, foundations, or medical associations (Burke 1990; Ohkusa 2010; Petersen 2014); and one reported internal funding as part of a medical education and research institution (Mishra 2020).

Three studies declared no conflicts of interest (Mishra 2020; Ohkusa 2010; Petersen 2014).

Excluded studies

We excluded 44 studies (46 records) for the reasons presented in the Characteristics of excluded studies table and the summary below.

1. Eleven did not study people with UC (Hillaire 2007; Inoue 2007; Järnerot 1989; Jigaranu 2014; Lee 2018; Leiper 2000; Leiper 2008; Levine 2016; Levine 2019; Selby 2007; Turunen 1995).

2. Ten were not RCTs (Abdallah 1970; Breton 2019; Ishikawa 2017; JPRN‐UMIN000025846; Kordy 2018; Kumar 2010; Lobo 1993; Okahara 2020; Terao 2011; Uehara 2010).

3. Five were letters to the editor (Borok 1996; Guslandi 2006; Kuenstner 2007; Schoonees 1996; Turunen 1999).

4. Five were not outcomes of interest (Koido 2014; Minami 2009; Nomura 2005; Sato 2009; Yukawa 2013).

5. Six had data on induction of remission, but were less than two weeks in length (Casellas 1998; Chapman 1986; Dickinson 1985; Gionchetti 1999; Turner 2019; Wu 2013).

6. Five were not interventions of interest (Aoki 2012; El‐Nachef 2020; Henn 2021; Jena 2020; NCT04898348).

7. Two were not study designs of interest (Gardner 1981; Kato 2014).

Studies awaiting classification

Eight RCTs (10 records) are awaiting classification (Characteristics of studies awaiting classification table).

Ongoing studies

We identified one ongoing study (Characteristics of ongoing studies table).

Risk of bias in included studies

The results of our risk of bias assessments are presented in Figure 2; Figure 3; and the risk of bias tables in the Characteristics of included studies table. Our initial assessment was conducted on the information presented in the published papers. In studies where the risk of bias assessments was unclear, we sought clarification from at least one author or contact person (or both) per study. Where we received responses, we adapted our initial assessment accordingly. We received responses for the following studies: Burke 1990; Mantzaris 1994; Mantzaris 1997; Mantzaris 2001; Mishra 2020; Ohkusa 2005; Ohkusa 2010; and Turunen 1998.

2.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

3.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Allocation

Nine studies clearly described randomisation, corresponding with a low risk of bias (Burke 1990; Mantzaris 1994; Mantzaris 1997; Mantzaris 2001; Mishra 2020; Ohkusa 2005; Ohkusa 2010; Petersen 2014; Turunen 1998). Three studies did not sufficiently describe randomisation, and therefore were at unclear risk of bias (Bataga 2015; Gilat 1987; Gilat 1989).

We rated eight studies at low risk of selection bias, as the method of random allocation of participants to intervention and control groups and allocation concealment was described adequately (Burke 1990; Mantzaris 1994; Mantzaris 1997; Mantzaris 2001; Mishra 2020; Ohkusa 2005; Ohkusa 2010; Turunen 1998). We rated the other four studies at unclear risk of selection bias and allocation concealment, as they did not provide enough information about their selection and allocation concealment processes (Bataga 2015; Gilat 1987; Gilat 1989; Petersen 2014).

Blinding

We rated 10 studies at low risk of performance bias, since they used a control medication or placebo that could not be differentiated from the intervention, and the studies were described as blinded by the authors (Burke 1990; Gilat 1987; Gilat 1989; Mantzaris 1994; Mantzaris 1997; Mantzaris 2001; Mishra 2020; Ohkusa 2010; Petersen 2014; Turunen 1998). We rated Bataga 2015 at unclear risk of performance bias because there was no information about blinding and we received no response from the author. We rated Ohkusa 2005 at high risk of bias as there was no placebo or other control medication to match the intervention medication.

Detection bias was low in eight studies (Burke 1990; Mantzaris 1994; Mantzaris 1997; Mantzaris 2001; Mishra 2020; Ohkusa 2005; Ohkusa 2010; Turunen 1998). In four studies the risk of detection bias was unclear because not enough information was provided (Bataga 2015; Gilat 1987; Gilat 1989; Petersen 2014).

Incomplete outcome data

Attrition bias was low in all 12 studies (Bataga 2015; Burke 1990; Gilat 1987; Gilat 1989; Mantzaris 1994; Mantzaris 1997; Mantzaris 2001; Mishra 2020; Ohkusa 2005; Ohkusa 2010; Petersen 2014; Turunen 1998).

Selective reporting

Reporting bias was low in seven studies, for which all outcomes specified in the methods had data presented within the published reports (Gilat 1989; Mantzaris 1994; Mantzaris 1997; Mantzaris 2001; Mishra 2020; Ohkusa 2010; Turunen 1998). The remaining five studies lacked enough information to judge whether their intended outcomes were consistently reported (Bataga 2015; Burke 1990; Gilat 1987; Ohkusa 2005; Petersen 2014). Therefore, we rated the bias risk as unclear in these five studies, since we did not receive satisfactory clarification from the authors.

Other potential sources of bias

We rated seven studies as low in other potential sources of bias (Burke 1990; Mantzaris 1994; Mantzaris 1997; Mantzaris 2001; Mishra 2020; Ohkusa 2010; Petersen 2014). We rated four studies as having unclear risk of bias due to observed imbalances in baseline participant characteristics for which we received no clarification from the authors (Bataga 2015; Gilat 1987; Gilat 1989; Turunen 1998). We rated one study at high risk of bias due to significant baseline imbalances between the intervention and control group (Ohkusa 2005).

Effects of interventions

See: Table 1; Table 2; Table 3; Table 4; Table 5; Table 6

Summary of findings 1. Antibiotics compared to placebo for the induction of remission in ulcerative colitis.

Antibiotics compared to placebo for the induction of remission in ulcerative colitis
Patient or population: people with active UC Setting: UK and Greece Intervention: antibiotics Comparison: placebo
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with placebo	Risk with antibiotics
Failure to induce clinical remission at end of study	Study population		RR 0.88 (0.74 to 1.06)	608 (7 studies)	⊕⊕⊕⊕ High	—
	576 per 1000a	507 per 1000 (426 to 611)
Failure to induce clinical response at end of study	Study population		RR 0.75 (0.47 to 1.22)	343 (3 study)	⊕⊕⊝⊝ Lowb	—
	691 per 1000	519 per 1000 (325 to 843)
Serious adverse events	Study population		RR 1.19 (0.38 to 3.71)	691 (8 studies)	⊕⊕⊝⊝ Lowc	—
	14 per 1000	17 per 1000 (5 to 52)
Withdrawals due to adverse events	Study population		RR 2.06 (0.27 to 15.72)	691 (8 studies)	⊕⊕⊝⊝ Lowc	—
	3 per 1000	6 per 1000 (1 to 47)
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; UC: ulcerative colitis.
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aWe used the total number of events/total number of participants to estimate the control group risk, unless there were zero events in which case, we entered a value of 0.5 events.
^bDowngraded one level due to imprecision from low event numbers and one level due to inconsistency (I² = 65%).
^cDowngraded two levels due to imprecision from very low event numbers.

Summary of findings 2. Antibiotics plus probiotics compared to placebo or no intervention for the induction of remission in ulcerative colitis.

Antibiotics plus probiotics compared to placebo or no intervention for the induction of remission in ulcerative colitis
Patient or population: people with active UC Setting: Greece, Finland, Denmark Intervention: antibiotics plus probiotics Comparison: placebo or no intervention
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with placebo or no intervention	Risk with antibiotics and probiotics
Failure to induce clinical remission at end of study	Study population		RR 0.68 (0.39 to 1.19)	98 (2 studies)	⊕⊝⊝⊝ Very lowb	—
	327 per 1000a	222 per 1000 (128 to 390)
Failure to induce clinical response at end of study	—		—	—	—	—
Serious adverse events	Study population		RR 1.00 (0.07 to 15.08)	98 (2 studies)	⊕⊝⊝⊝ Very lowb	—
	20 per 1000	20 per 1000 (1 to 302)
Withdrawals due to adverse events	Study population		RR 1.00 (0.07 to 15.08)	98 (2 studies)	⊕⊝⊝⊝ Very lowb	—
	20 per 1000	20 per 1000 (1 to 302)
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; UC: ulcerative colitis.
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aWe used the total number of events/total number of participants to estimate the control group risk, unless there were zero events in which case, we entered a value of 0.5 events.
^bDowngraded two levels due high imprecision from very low event numbers and one level due to unclear risk of bias for multiple risk of bias fields.

Summary of findings 3. Antibiotics compared to 5‐ASA for the induction of remission in ulcerative colitis.

Antibiotics compared to 5‐ASA for the induction of remission in ulcerative colitis
Patient or population: people with active UC Setting: Israel Intervention: antibiotics Comparison: 5‐ASA
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with 5‐ASA	Risk with antibiotics
Failure to induce clinical remission at end of study	—		—	—	—	—
Failure to induce clinical response at end of study	Study population		RR 2.20 (1.17 to 4.14)	46 (1 study)	⊕⊝⊝⊝ Very lowb	—
	350 per 1000a	770 per 1000 (410 to 1000)
Serious adverse events	Study population		Not estimable	46 (1 study)	—	—
	0 per 1000	0 per 1000 (0 to 0)			—	—
Withdrawals due to adverse events	Study population		Not estimable	46 (1 study)	—	—
	0 per 1000	0 per 1000 (0 to 0)			—	—
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). 5‐ASA: 5‐aminosalicylic acid; CI: confidence interval; RR: risk ratio; UC: ulcerative colitis.
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aWe used the total number of events/total number of participants to estimate the control group risk, unless there were zero events in which case, we entered a value of 0.5 events.
^bDowngraded two levels due to high imprecision from very low event numbers and one level due to unclear risk of bias in multiple risk of bias fields.

Summary of findings 4. Antibiotics compared to probiotics for the induction of remission in ulcerative colitis.

Antibiotics compared to probiotics for the induction of remission in ulcerative colitis
Patient or population: people with active UC Setting: Denmark Intervention: antibiotics Comparison: probiotics
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with probiotics	Risk with antibiotics
Failure to induce clinical remission at end of study	Study population		RR 0.47 (0.23 to 0.94)	50 (1 study)	⊕⊝⊝⊝ Very lowb	—
	600 per 1000a	282 per 1000 (138 to 564)
Failure to induce clinical response at end of study	—	—	—	—	—	—
Serious adverse events	Study population		Not estimable	50 (1 study)	—	—
	0 per 1000	0 per 1000 (0 to 0)			—	—
Withdrawals due to adverse events	Study population		Not estimable	50 (1 study)	—	—
	0 per 1000	0 per 1000 (0 to 0)			—	—
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; UC: ulcerative colitis.
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aWe used the total number of events/total number of participants to estimate the control group risk, unless there were zero events in which case, we entered a value of 0.5 events.
^bDowngraded two levels due to high imprecision from very low event numbers and one level due to unclear risk of bias in multiple risk of bias fields.

Summary of findings 5. Antibiotics compared to 5‐ASA for the maintenance of remission in ulcerative colitis.

Antibiotics compared to 5‐ASA for the maintenance of remission in ulcerative colitis
Patient or population: people with UC in remission Setting: Israel Intervention: antibiotics Comparison: 5‐ASA
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with 5‐ASA	Risk with antibiotics
Failure to maintain clinical remission at end of study	Study population		RR 0.71 (0.47 to 1.06)	40 (1 study)	⊕⊝⊝⊝ Very lowb	—
	833 per 1000a	592 per 1000 (392 to 883)
Failure to maintain clinical response at end of study	—	—	—	—	—	—
Serious adverse events	Study population		Not estimable	40 (1 study)	—	—
	0 per 1000	0 per 1000 (0 to 0)			—	—
Withdrawals due to adverse events	Study population		Not estimable	40 (1 study)	—	—
	0 per 1000	0 per 1000 (0 to 0)			—	—
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). 5‐ASA: 5‐aminosalicylic acid; CI: confidence interval; RR: risk ratio; UC: ulcerative colitis.
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aWe used the total number of events/total number of participants to estimate the control group risk, unless there were zero events in which case, we entered a value of 0.5 events.
^bDowngraded two levels due to high imprecision from very low event numbers and one level due to unclear risk of bias in multiple risk of bias fields.

Summary of findings 6. Antibiotics compared to no treatment for a mixed population of active and inactive people with ulcerative colitis.

Antibiotics compared to no treatment for a mixed population of active and inactive people with ulcerative colitis
Patient or population: people with UC with active and inactive disease Setting: Japan Intervention: antibiotics Comparison: no treatment
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with no treatment	Risk with antibiotics
Failure to induce clinical remission at end of study	Study population		RR 0.56 (0.29 to 1.07)	20 (1 study)	⊕⊝⊝⊝ Very lowb	—
	900 per 1000a	504 per 1000 (261 to 963)
Failure to induce clinical response at end of study	—		—	—	—	—
Serious adverse events	Study population		Not estimable	20 (1 study)	—	—
	0 per 1000	0 per 1000 (0 to 0)			—	—
Withdrawals due to adverse events	Study population		Not estimable	20 (1 study)	—	—
	0 per 1000	0 per 1000 (0 to 0)			—	—
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; UC: ulcerative colitis.
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aWe used the total number of events / total number of participants to estimate the control group risk, unless there were zero events in which case, we entered a value of 0.5 events.
^bDowngraded two levels due to high imprecision from very low event numbers and one level due to high and unclear risk of bias in multiple risk of bias fields.

Antibiotics versus placebo for induction of remission

Eight studies compared the effect of antibiotics versus placebo (Burke 1990; Mantzaris 1994; Mantzaris 1997; Mantzaris 2001; Mishra 2020; Ohkusa 2010; Petersen 2014; Turunen 1998).

The length of these interventions ranged from five days (Mishra 2020) up to six months of antibiotic therapy (Turunen 1998). Two studies followed up their participants at 12 months from the beginning of their studies to check how many remained in remission (Ohkusa 2010; Turunen 1998).

Mantzaris 1994, Mantzaris 2001, and Mishra 2020 included people with severe UC. Mantzaris 1997 included people with mild‐to‐moderate disease, Ohkusa 2010 with mild‐to‐severe disease, and Turunen 1998 with moderate‐to‐severe disease. Petersen 2014 included people in active disease with a CAI of 6 or greater. Burke 1990 included people across the spectrum of severity (mild to severe). Bataga 2015 included people with minor relapse. Gilat 1987 included people with non‐severe disease activity.

Burke 1990 subdivided each of its two intervention arms into three subgroups (subgroup 1: inpatient prednisolone 60 mg; subgroup 2: outpatient prednisolone 30 mg; subgroup 3: outpatient hydrocortisone). Unfortunately, due to the time that has passed since the study's publication, the author could not provide us with additional subgroup data.

Primary outcomes

Achieved clinical remission

The studies found no difference in the proportion of participants who failed to achieve clinical remission at end of intervention between antibiotics and placebo (154/304 participants with antibiotics versus 175/304 participants with placebo; RR 0.88, 95% CI 0.74 to 1.06; studies = 8; I² = 22%; Analysis 1.1). Antibiotics compared to placebo made no difference in failure to achieve clinical remission. The certainty of the evidence was high (Table 1).

1.1. Analysis.

Comparison 1: Antibiotics versus placebo for induction of remission, Outcome 1: Failure to induce remission at end of study

In subgroup analyses for additional therapies, there was no difference for antibiotics plus steroids versus placebo plus steroids (RR 0.78, 95% CI 0.51 to 1.20); antibiotics plus steroids plus 5‐ASA versus placebo plus steroids plus 5‐ASA (RR 0.96, 95% CI 0.85 to 1.09); or antibiotics plus probiotics plus steroids plus 5‐ASA versus placebo plus probiotics plus steroids plus 5‐ASA (RR 0.67, 95% CI 0.37 to 1.19) (Analysis 1.1).

One study reported the proportion of participants who failed to achieve clinical remission 12 months after the intervention (Ohkusa 2010). Antibiotics probably lead to a slightly lower proportion of participants who fail to achieve clinical remission compared to placebo at 12 months from the intervention (77/105 participants with antibiotics versus 90/105 with placebo; RR 0.86, 95% CI 0.74 to 0.98; Analysis 1.2). The certainty of the evidence was moderate, downgraded once due to imprecision from low event numbers.

1.2. Analysis.

Comparison 1: Antibiotics versus placebo for induction of remission, Outcome 2: Failure to induce remission at 12 months from study end

Secondary outcomes

Achieved clinical response

Three studies reported the proportion of participants who failed to achieve clinical response at end of intervention (Mishra 2020; Ohkusa 2010; Turunen 1998). There was no clear difference in the proportion of participants who failed to achieve clinical response at end of intervention when antibiotics were compared to placebo (102/168 participants with antibiotics versus 121/175 participants with placebo; RR 0.75, 95% CI 0.47 to 1.22; studies = 3; I² = 65%; Analysis 1.3). The certainty of the evidence was low, downgraded one level due to imprecision from low event numbers and one level due to inconsistency (Table 1).

1.3. Analysis.

Comparison 1: Antibiotics versus placebo for induction of remission, Outcome 3: Failure to induce clinical response at end of study

In subgroup analyses for additional therapies, there were no differences for antibiotics plus steroids versus placebo plus steroids (RR 0.75, 95% CI 0.39 to 1.46), or antibiotics plus steroids plus 5‐ASA versus placebo plus steroids plus 5‐ASA (RR 0.72, 95% CI 0.33 to 1.56) (Analysis 1.3).

Two studies reported the proportion of participants who failed to achieve clinical response 12 months after the intervention (Ohkusa 2010; Turunen 1998). Antibiotics probably lead to a lower proportion of participants who fail to achieve clinical response compared to placebo (70/143 participants with antibiotics versus 110/150 participants with placebo; RR 0.66, 95% CI 0.54 to 0.80; Analysis 1.4). The certainty of the results was moderate, downgraded once due to imprecision from low event numbers.

1.4. Analysis.

Comparison 1: Antibiotics versus placebo for induction of remission, Outcome 4: Failure to induce clinical response at 12 months from study end

Clinical response in Petersen 2014 was defined as improvement in CAI scores. At the end of the study, the mean CAI score at week 12 for the antibiotics plus steroids plus 5‐ASA group was 4.14 compared to 4.9 for the placebo plus steroids plus 5‐ASA group. The other studies did not report on clinical response.

Achieved endoscopic remission

Four studies reported data on endoscopic remission/response (Burke 1990; Mantzaris 2001; Ohkusa 2010; Turunen 1998); however, the provided data were insufficient for meta‐analysis. Burke 1990 reported the mean (SD) scores as the end of study as measured on a 0 to 4 scale (0 = normal; 1 = loss of vascular pattern but no granularity; 2 = granularity; 3 = granularity plus contact bleeding; 4 = spontaneous bleeding, friability, and frank ulceration), which for the antibiotics plus steroids group was 1.3 (SD 1.5) and for the placebo plus steroids group was 1.9 (SD 1.5). Mantzaris 2001 provided a narrative description of the endoscopic results for the total study non‐responders, mentioning that 9/12 non‐responders underwent sigmoidoscopy, which revealed a mildly inflamed rectum but [no] severe inflammation proximal to the sigmoid colon with numerous pseudopolyps, deep ulcers, and extensive areas of denuded mucosa. Turunen 1998 reported the proportion of participants who failed to achieve endoscopic remission when receiving antibiotics plus steroids plus 5‐ASA compared to placebo plus steroids plus 5‐ASA based on a 0 to 4 scoring scale (7/38 with antibiotics plus steroids plus 5‐ASA versus 23/45 with placebo plus steroids plus 5‐ASA). Ohkusa 2010 evaluated endoscopic remission/response according to the Mayo system sigmoidoscopy subscore, with scores of 0 to 3 and reported that the antibiotics group had a mean score of 1.33 at three months and 1.29 at 12 months, while the placebo group had a mean score of 1.59 at three months and 1.56 at 12 months.

Achieved histological remission

Two studies reported data on histological remission/response (Burke 1990; Turunen 1998); however, the provided data were insufficient for meta‐analysis. Burke 1990 measured histological scoring using a scale of 0 to 3 for mucous depletion, surface degeneration, and polymorph infiltration. The total active inflammatory score ranged from 0 to 9, which for the antibiotics group was 2.8 (SD 2.5) and for the placebo group was 4.0 (SD 2.6). They also mentioned that 35/42 participants in the antibiotic plus steroids groups and 42/42 participants in the placebo plus steroids group failed to achieve a histological score of 0. Turunen 1998 used a 0 to 3 scale for histological response and mentioned that 4/38 participants in the antibiotics plus steroids plus 5‐ASA group and 23/25 participants in the placebo plus steroids plus 5‐ASA groups failed to achieve histological response at three months.

Clinical relapse

None of the studies reported clinical relapse.

Endoscopic relapse

None of the studies reported endoscopic relapse.

Health‐related quality of life

None of the studies reported health‐related quality of life.

Any adverse event

Four studies reported total AEs. In Burke 1990, two participants required surgery (one in each group); Mantzaris 1994 reported five participants with "a few mild, transient, steroid‐related side effects"; Mantzaris 2001 reported four participants with "a few mild, transient, steroid‐related side effects"; and Mishra 2020 documented that one participant in the control group had evidence of an entamoeba cyst. Mantzaris 1997 reported four total AEs; Ohkusa 2010 71 (participants experienced one or more of nausea, fever, watery diarrhoea, urticaria, and fever); Petersen 2014 six (but for only this arm of the intervention) treated with antibiotics experienced itching, nausea or bloating, and abdominal pain and two with placebo experienced either nausea or bloating and abdominal pain) and Turunen 1998 three (two cases of deep venous thrombosis were recorded in the antibiotics group and one in the control group).

Serious adverse events

The studies found no clear difference in SAEs when antibiotics were compared to placebo (6/342 participants with antibiotics versus 5/349 participants with placebo; RR 1.19, 95% CI 0.38 to 3.71; Analysis 1.5). The certainty of the evidence was low, downgraded two levels due to imprecision from very low event numbers (Table 1).

1.5. Analysis.

Comparison 1: Antibiotics versus placebo for induction of remission, Outcome 5: Serious adverse events

In a subgroup analysis for additional therapies, there was no difference for antibiotics plus steroids versus placebo plus steroids (RR 1.19, 95% CI 0.38 to 3.71; Analysis 1.5).

Withdrawals from the study due to adverse events

The studies found no clear difference in withdrawals due to AEs when antibiotics were compared to probiotics (3/342 participants with antibiotics versus 1/349 with placebo; RR 2.06, 95% CI 0.27 to 15.72; Analysis 1.6). The certainty of the evidence was low, downgraded two levels due to imprecision from very low event numbers (Table 1).

1.6. Analysis.

Comparison 1: Antibiotics versus placebo for induction of remission, Outcome 6: Withdrawals due to adverse events

In subgroup analyses for additional therapies, there was no difference for antibiotics plus steroids versus placebo plus steroids (RR 1.00, 95% CI 0.06 to 15.47), or antibiotics plus steroids plus 5‐ASA versus placebo plus steroids plus 5‐ASA (RR 5.00, 95% CI 0.24 to 102.91) (Analysis 1.6).

Antibiotics plus probiotics versus no intervention/placebo for induction of remission

Bataga 2015 and Petersen 2014 compared antibiotics plus probiotics versus no intervention/placebo.

Bataga 2015 lasted six months, and Petersen 2014 lasted for 12 weeks. Bataga 2015 included people with minor relapse, and Petersen 2014 included people with active UC. In Bataga 2015, antibiotics were given for one month and probiotics for 10 days. In Petersen 2014, antibiotics and their placebo equivalent were given for one week, and the probiotics and their placebo equivalent for seven weeks.

Primary outcomes

Achieved clinical remission

The studies found no clear difference in the proportion of participants who failed to achieve remission at end of intervention when antibiotics were compared to probiotics (11/49 participants with antibiotics versus 16/49 participants with probiotics; RR 0.68, 95% CI 0.39 to 1.19; studies = 2; I² = 0%; Analysis 2.1). The certainty of the evidence was very low, downgraded two levels due to high imprecision from very low event numbers and one level due to unclear risk of bias for multiple risk of bias domains (Table 2).

2.1. Analysis.

Comparison 2: Antibiotics and probiotics versus no intervention/placebo for induction of remission, Outcome 1: Failure to induce remission at end of study

In subgroup analyses for additional therapies, there were no differences for antibiotics plus probiotics plus 5‐ASA versus steroids plus 5‐ASA (RR 1.00, 95% CI 0.07 to 15.08), or antibiotics plus probiotics plus steroids plus 5‐ASA versus placebo plus steroids plus 5‐ASA (RR 0.67, 95% CI 0.37 to 1.19) (Analysis 2.1).

Secondary outcomes

Achieved clinical response

Clinical response was defined as improvement in CAI scores. At the end of the study, the mean CAI score at week 12 for the antibiotics plus probiotics group was 4.1 compared to 4.9 for the placebo group. As no SDs were provided, we could not conduct an analysis.

Achieved endoscopic remission

None of the studies reported endoscopic remission.

Achieved histological remission

None of the studies reported histological remission.

Clinical relapse

None of the studies reported clinical relapse.

Endoscopic relapse

None of the studies reported endoscopic relapse.

Health‐related quality of life

None of the studies reported health‐related quality of life.

Any adverse event

The total AEs reported in Petersen 2014 were: "four treated with ciprofloxacin experienced either itching, nausea or bloating and abdominal pain and two with placebo experienced either nausea or bloating and abdominal pain. During the seven weeks of probiotics/placebo treatment, 18 patients reported side effects, nine treated with probiotics (itching, nausea, bloating, abdominal pain) and nine treated with placebo (nausea, bloating, abdominal pain, sensitivity towards smells)".

Serious adverse events

The studies found no clear difference in SAEs when antibiotics were compared to probiotics (1/49 participants with antibiotics versus 1/49 participants with probiotics; RR 1.00, 95% CI 0.07 to 15.08; Analysis 2.2). The certainty of the evidence was very low, downgraded two levels due to high imprecision from very low event numbers and one level due to unclear risk of bias for multiple risk of bias fields (Table 2).

2.2. Analysis.

Comparison 2: Antibiotics and probiotics versus no intervention/placebo for induction of remission, Outcome 2: Serious adverse events

Withdrawals from the study due to adverse events

The studies found no clear difference in withdrawals due to AEs when antibiotics were compared to probiotics (1/49 participants with antibiotics versus 1/49 participants with probiotics; RR 1.00, 95% CI 0.07 to 15.08; Analysis 2.3). The certainty of the evidence was very low, downgraded two levels due to high imprecision from very low event numbers and one level due to unclear risk of bias for multiple risk of bias fields (Table 2).

2.3. Analysis.

Comparison 2: Antibiotics and probiotics versus no intervention/placebo for induction of remission, Outcome 3: Withdrawals due to adverse events

Antibiotics versus 5‐aminosalicylic acid for induction of remission

One study compared the effect of antibiotics to 5‐ASA (Gilat 1987). The length of the intervention was 28 days, and the participants had non‐severe UC.

Primary outcomes

Achieved clinical remission

The study did not report the proportion of participants who failed to achieve remission.

Secondary outcomes

Achieved clinical response

The study reported the proportion of participants who failed to achieve clinical response. The study assessed 46 participants at 28 days from the beginning of the intervention and found that antibiotics might lead to fewer participants failing to achieve clinical response compared to 5‐ASA (20/26 participants with antibiotics versus 7/20 participants with 5‐ASA; RR 2.20, 95% CI 1.17 to 4.14; Analysis 3.1). The certainty of the evidence was very low. It was downgraded two levels due to high imprecision from very low event numbers and one level due to unclear risk of bias in multiple risk of bias domains (Table 3).

3.1. Analysis.

Comparison 3: Antibiotics versus 5‐ASA for induction of remission, Outcome 1: Failure to induce clinical response at end of study

Achieved endoscopic remission

Data on endoscopic remission was not provided, but data on endoscopic response were. Sigmoidoscopies were graded on a 0 to 2 scale, and it was reported that 17/26 participants failed to achieve endoscopic response in the antibiotics group compared to 7/20 participants in the 5‐ASA group.

Achieved histological remission

The study did not report histological remission.

Clinical relapse

The study did not report clinical relapse.

Endoscopic relapse

The study did not report endoscopic relapse.

Health‐related quality of life

The study did not report health‐related quality of life.

Any adverse event

The study reported "a few" total AEs, "such as bitter taste in the mouth or nausea".

Serious adverse events

The study reported zero SAEs. Because of this, it was not possible to estimate the effect of antibiotics compared to 5‐ASA on SAEs.

Withdrawals from the study due to adverse events

The study reported zero withdrawals due to AEs. Because of this, it was not possible to estimate the effect of antibiotics compared to 5‐ASA on withdrawals due to AEs.

Antibiotics versus probiotics for induction of remission

Petersen 2014 compared antibiotics versus probiotics in one arm of their four‐arm intervention.

Primary outcomes

Achieved clinical remission

The study found no clear difference in the proportion of participants who failed to achieve remission at 12 weeks when antibiotics were compared to probiotics (7/25 participants with antibiotics versus 15/25 participants with probiotics; RR 0.47, 95% CI 0.23 to 0.94; studies = 1; Analysis 4.1). The certainty of the evidence was very low, downgraded two levels due to high imprecision from very low event numbers and one level due to unclear risk of bias in multiple risk of bias fields (Table 4).

4.1. Analysis.

Comparison 4: Antibiotics versus probiotics for induction of remission, Outcome 1: Failure to induce remission at end of study

Secondary outcomes

Achieved clinical response

Clinical response was defined as improvement in CAI scores. At the end of the study, the mean CAI score for the antibiotics group was 4.14 compared to 5.29 for the probiotics group. As no SD values were provided, we could not conduct an analysis.

Achieved endoscopic remission

The study did not report endoscopic remission.

Achieved histological remission

The study did not report histological remission.

Clinical relapse

The study did not report clinical relapse.

Endoscopic relapse

The study did not report endoscopic relapse.

Health‐related quality of life

The study did not report health‐related quality of life.

Any adverse event

The total adverse reported for the full study were: "four treated with ciprofloxacin experienced either itching, nausea or bloating and abdominal pain and two with placebo experienced either nausea or bloating and abdominal pain. During the seven weeks of probiotics/placebo treatment, 18 patients reported side effects, nine treated with probiotics (itching, nausea, bloating, abdominal pain) and nine treated with placebo (nausea, bloating, abdominal pain, sensitivity towards smells)" (Gilat 1987).

Serious adverse events

SAEs were zero in both groups. For this reason, we could not estimate the effect of the intervention on SAEs.

Withdrawals from the study due to adverse events

Withdrawals due to AEs were zero in both groups. For this reason, we could not estimate the effect of the intervention on withdrawals due to AEs.

We contacted the authors for more information but received no response.

Antibiotics versus 5‐aminosalicylic acid for maintenance of remission

Gilat 1989 compared antibiotics versus 5‐ASA for maintenance of remission.

The length of the intervention was 12 months, with follow‐up every three months or until disease recurrence. The study included people with UC in remission.

Primary outcomes

Relapses

The study assessed 40 participants and found no clear difference in the proportion who failed to maintain remission at 12 months from the beginning of the intervention when antibiotics were compared to 5‐ASA (13/22 participants with antibiotics versus 15/18 with 5‐ASA; RR 0.71, 95% CI 0.47 to 1.06; Analysis 5.1). The certainty of the evidence was very low, downgraded two levels due to high imprecision from very low event numbers and one level due to unclear risk of bias in multiple risk of bias fields (Table 5).

5.1. Analysis.

Comparison 5: Antibiotics versus 5‐ASA for maintenance of remission, Outcome 1: Failure to maintain remission at end of study

Secondary outcomes

Achieved clinical response

The study did not report clinical response.

Achieved endoscopic remission

Data on endoscopic remission were not provided, but endoscopic response were. Endoscopic response was defined as no worsening of the participants' sigmoidoscopy findings or improvement (13/22 participants in the antibiotics group and 15/18 participants in the 5‐ASA groups).

Achieved histological remission

The study did not report histological remission.

Clinical relapse

The study did not report clinical relapse.

Endoscopic relapse

The study did not report endoscopic relapse.

Health‐related quality of life

The study did not report health‐related quality of life.

Any adverse event

The study reported "a few" total AEs, "such as bitter taste in the mouth or nausea".

Serious adverse events

SAEs were zero in both groups. For this reason, we could not estimate the effect of the intervention on SAEs.

Withdrawals from the study due to adverse events

Withdrawals due to AEs were zero in both groups. For this reason, we could not estimate the effect of the intervention on withdrawals due to AEs.

Antibiotics versus no treatment in a mixed population with active and inactive disease

One study examined the effect of an antibiotic combination therapy plus steroids plus 5‐ASA versus steroids plus 5‐ASA (Ohkusa 2005). The intervention length was two weeks.

In the published paper, the authors mentioned that they included people with active, mild‐to‐moderate UC. However, in our correspondence, they mentioned that based on their definition of remission, which was a Lichtiger symptom score less than three (Lichtiger 1990), 7/10 participants in the intervention group were in remission and 1/10 participants in the control group were in remission, meaning that this study used a mix of participants in remission and with active disease. Our team debated whether we should consider this study an induction or maintenance study or exclude it from our review. We decided that we could not exclude it, as it was not explicitly violating our inclusion criteria; however, we could neither consider it as a study on the induction of remission nor the maintenance of remission, as the authors could not provide us with separate data for the study participants in remission and with active disease. Therefore, we are presenting this study in a separate category of studies on a mix of participants with active and inactive disease.

Primary outcomes

Achieved clinical remission

The study assessed 20 participants and found no clear difference at three months in the proportion who failed to achieve remission when the antibiotics plus steroids plus 5‐ASA were compared to steroids plus 5‐ASA (5/10 participants with antibiotics plus steroids plus 5‐ASA versus 9/10 participants with steroids plus 5‐ASA; RR 0.56, 95% CI 0.29 to 1.07; Analysis 6.1). The certainty of the evidence was very low, downgraded two levels due to high imprecision from very low event numbers and one level due to high and unclear risk of bias in multiple risk of bias fields (Table 6).

6.1. Analysis.

Comparison 6: Antibiotics versus no treatment in mixed population with active and inactive disease, Outcome 1: Failure to induce remission at 3 months

The study found that antibiotics plus steroids plus 5‐ASA may lead to fewer participants failing to achieve remission at 12 months compared to steroids plus 5‐ASA (2/10 participants with antibiotics plus steroids plus 5‐ASA versus 9/10 participants with steroids plus 5‐ASA; RR 0.22, 95% CI 0.06 to 0.78; I² = 0%; Analysis 6.2). The certainty of the evidence was very low. It was downgraded two levels due to high imprecision from very low event numbers and one level due to high and unclear risk of bias in multiple risk of bias fields (Table 6).

6.2. Analysis.

Comparison 6: Antibiotics versus no treatment in mixed population with active and inactive disease, Outcome 2: Failure to maintain remission at 12 months

Secondary outcomes

Achieved clinical response

Clinical response was scored according to the Lichtiger symptom score (Lichtiger 1990) with a mean of 3 (range 0 to 3) at three months and of 2 (range 1 to 5) at 12 months for participants in the antibiotics plus steroids plus 5‐ASA group and of 4 (range 2 to 8) at three months and of 4 (range 3 to 8) at 12 months for participants in the steroids plus 5‐ASA group. As no SD values were provided, we could not conduct an analysis.

Achieved endoscopic remission

Endoscopic response was scored according to the Matts grading score (Matts 1961) with a mean of 1 (range 1 to 2) at three months and of 1 (range 1 to 3) at 12 months for participants in the antibiotics plus steroids plus 5‐ASA group and of 3 (range 1 to 4) at three months and of 3 (range 1 to 4) at 12 months for participants in the steroids plus 5‐ASA group. As no SD values were provided, we could not conduct an analysis.

Achieved histological remission

Histological response was scored according to the Matts grading score (Matts 1961) with a mean of 14 (range 11 to 18) at three months and of 13 (range 9 to 21) at 12 months for participants in the antibiotics plus steroids plus 5‐ASA group and of 17 (range 11 to 20) at three months and of 16 (range 9 to 20) at 12 months for participants in the steroids plus 5‐ASA group. As no SD values were provided, we could not conduct an analysis.

Clinical relapse

The study did not report clinical relapse.

Endoscopic relapse

Endoscopic relapse was defined as the reappearance of erosion or an ulcer; however, it was unclear the number of participants who had experienced an endoscopic relapse in the antibiotics plus steroids plus 5‐ASA group, and at which stage. It was mentioned that two participants relapsed at three to five months to a score of 4 and one more at 12 to 14 months in the steroids plus 5‐ASA group.

Health‐related quality of life

The study reported that 8/10 participants had improved well‐being in the antibiotics plus steroids plus 5‐ASA group; however, it was not mentioned how it was measured or what the well‐being outcome was in the steroids plus 5‐ASA group.

Any adverse event

The total number of AEs throughout the whole study was seven and they all occurred in the antibiotics group. Three participants experienced metallic taste, two had malaise, one had watery diarrhoea without blood, and one had urticaria and fever.

Serious adverse events

SAEs were zero in both groups. For this reason, we could not estimate the effect of the intervention on SAEs.

Withdrawals from the study due to adverse events

Withdrawals due to AEs zero in both groups. For this reason, we could not estimate the effect of the intervention on withdrawals due to AEs.

Discussion

Summary of main results

This review included 12 RCTs with 847 participants, which examined the efficacy and safety of antibiotics for inducing and maintaining remission in people with UC.

There is high certainty evidence that there is no difference between antibiotics and placebo in the proportion of participants who failed to achieve clinical remission at end of intervention. Subgroup analysis considering various concurrent therapies, such as steroids, 5‐ASA, or probiotics, also found no difference in failure to achieve clinical remission. However, these were low certainty due to very serious imprecision resulting from low event and sample size numbers. One study found the proportion of participants who failed to achieve clinical remission 12 months after the intervention was probably lower with antibiotics than placebo, but this evidence was downgraded due to imprecision from low event numbers.

Three studies that reported the proportion of participants who failed to achieve clinical response at end of intervention found there may be no clear difference between antibiotics and placebo, with certainty of the evidence downgraded due to imprecision and inconsistency. Two studies found that antibiotics probably lead to a lower proportion of participants who fail to achieve clinical response than placebo at 12 months from the intervention, with certainty downgraded due to imprecision.

There was no clear difference in SAEs or withdrawals due to AEs between antibiotics and placebo, but certainty was low due to serious imprecision. Meta‐analysis of other outcomes was not possible due to heterogeneous outcomes or lack of data.

Two studies that compared the combination of antibiotics plus probiotics to placebo found no clear difference in failure to achieve clinical remission at the end of the intervention, SAEs, or withdrawals due to AEs. Certainty was downgraded to low due to very serious imprecision and risk of bias.

No other comparisons were reported by more than one study, so no further meta‐analysis was possible.

Overall completeness and applicability of evidence

There are several issues with the overall evidence of the included studies, which limits our ability to draw convergent conclusions to inform clinical practice and decision‐making.

First, the specific regimen of antibiotic therapy varied widely in not only the antimicrobial of interest but also the dosing regimen, course length, and concurrent therapy. The case definitions relating to the outcomes of clinical/endoscopic remission and clinical/endoscopic response varied widely, with a clear lack of uniformity across the included studies. Furthermore, it was common for the terms remission and response to be used interchangeably.

Second, the reporting of AEs was inconsistent across the included studies, with a lack of uniformity on what constitutes a minor AE/SAE. While the outcome of withdrawal is a much more objective measure in this context, it can be argued that this is of less interest to patients and therefore does limit the application of these data.

Third, the reporting of other outcomes considered important to patients and clinicians were also poorly reported, including quality of life indices and symptom severity scoring. As antibiotics were rarely used as a sole therapy, and therefore replacement for first‐ or second‐line medications, but rather to augment their efficacy, such outcomes are of equal or perhaps greater interest in this context. Additionally, the monitoring of these outcomes varied with just two studies measuring achievement of clinical remission at 12 months and therefore, evidence at this time point should be considered incomplete.

Fourth, the situation in which multiple antibiotics are compared to placebo makes discerning the individual contribution of each antibiotic towards the treatment effect extremely challenging. Cumulatively, these factors limit how the findings of this review can be applied to individual patients with confidence.

Fifth, all but one study allowed concurrent treatment, such as mesalazine, sulfasalazine, steroids, 5‐ASA, and probiotics or allowed patients to continue receiving concurrent medications. This causes uncertainty in terms of attributing treatment effects or AEs to the concomitant treatment or the study intervention itself.

The complex clinical and methodological heterogeneity issues limit the scope for meaningful subgroup analysis of key factors, such as gender, age, dosage and type of antimicrobial, and extent of disease, ultimately limiting the completeness of the evidence. There is insufficient information available to fully describe the relevance of the evidence due to poor reporting across the included studies. While no study was restrictive in terms of gender, there were some restrictions regarding the age of the participants, and the severity and pattern of the disease. For instance, several studies explicitly excluded paediatric populations, whereas others integrated participants of all ages within their analyses. Formal subgroup analysis with age as an effect modifier was not possible due to the lack of studies reporting outcomes separately for paediatric and adult populations. Similarly, due to inadequate reporting within the included studies, subgroup analysis based on disease severity or pattern of disease was not possible.

Quality of the evidence

The certainty of the evidence for the primary outcome for antibiotics versus placebo was high, but the evidence was of mostly low or very low certainty for other outcomes, primarily due to imprecision resulting from very sparse availability of data and issues with risk of bias.

Risk of bias was uncertain or high in several studies in a number of domains, with just two studies rated as uncertain or high in a single domain (Burke 1990; Turunen 1998), which is perhaps an indication of the overall quality of the included studies. There was selective reporting in five of the 12 included studies, while there were also concerns about baseline characteristic imbalances in five studies. There were no concerns about the risk of bias in five studies.

The study samples ranged from 20 to 210 participants, with only two studies including 100 participants or more within their trial. The statistical analysis showed inconsistency to be low in all but one outcome. As shown in the summary of findings tables, all but one of our outcomes were downgraded one or two levels due to imprecision from low sample sizes.

We have previously published work highlighting how common this type of difficulty in reporting sample size estimation is (Iheozor‐Ejiofor 2020b). It is an issue we have encountered in two other recent reviews we have published within the context of UC (Iheozor‐Ejiofor 2020a; Kaur 2020). This certainly is an issue that requires attention from the scientific community within this field to improve the quality of future reviews.

The follow‐up time (time point from the first observation to the last) was very heterogeneous. This certainly contributed to the inconsistency seen and further downgrades the overall quality of the analysed data.

Potential biases in the review process

For the main comparisons within this review, where different antibiotics were combined with other therapies versus placebo to analyse their effect on the induction of remission in UC, statistical heterogeneity was low for all but one comparison. However, we must highlight the clinical heterogeneity present, resulting from variation in the antibiotics administered and the therapies they were added to. The author team strongly considered this issue; however, we decided not to downgrade the certainty of the evidence for inconsistency further because different antibiotics have very similar modes of action.

The issue with one of the studies, in which both participants with active and inactive disease were considered in a single population, was not foreseen in this context. Therefore, we had to make a post‐hoc decision about how to deal with this study (see 'Induction/maintenance of remission' under Included studies). While explained transparently in this review, this potentially could introduce some bias.

Agreements and disagreements with other studies or reviews

This review is the first Cochrane Review on the topic of antibiotics for the induction and maintenance of remission in people with UC.

The most recent iteration of The ECCO guidance on the management of UC recommends the use of antibiotic therapy in instances where infection is considered or immediately prior to surgery. This approach is rationalised by stating that there is no consistent benefit to antibiotic therapy in addition to conventional therapy. The European Society of Paediatric Gastroenterology, Hepatology, and Nutrition (ESPGHAN) guidance does not routinely recommend antibiotic therapy in children unless C diff infection is suspected and subsequently confirmed (Turner 2018). The findings of this review support these recommendations with no evidence of reasonable certainty currently available to demonstrate clinical efficacy and support a reconsideration by guideline committees.

In 2011, a systematic review and meta‐analysis studying the use of antibiotic therapy in people with IBD was published (Khan 2011). The review found a statistically significant benefit for antibiotics in reducing remission in people with active UC. However, the diverse number of antibiotics tested and the array of combinations used made the data difficult to interpret. The inclusion criteria of this review differed from ours in that Khan 2011 limited inclusion to studies that compared antibiotic therapy to placebo or no treatment. In contrast, we accepted active comparators in addition to placebo or no therapy. Furthermore, the reviews differed in the study duration for induction of remission, where we stipulated a minimum of two weeks. The review by Khan 2011 did not specify a minimum time point to assess remission, but rather a specific minimum duration of therapy and maximum time point for the assessment of remission.

Authors' conclusions

Implications for practice.

There is high certainty evidence that there is no difference between antibiotics and placebo at the end of the delivery of an intervention. However, there is evidence that there may be a greater proportion of participants who achieve clinical remission and probably a greater proportion who achieved clinical response with antibiotics, when compared with placebo at 12 months. There is no clear difference in serious adverse events or withdrawals due to adverse events between antibiotics and placebo.

No clear conclusions can be drawn for any other comparisons.

Implications for research.

In accordance with the findings of other recently published reviews within inflammatory bowel disease, it is clear a large proportion of research within this field is plagued by disparate works produced in isolation of the wider field, thus leading to a compendium of research which fails to offer clarity on the important questions decision‐makers and patients need answering.

A clear direction for future research appears to comparisons of antibiotics and placebo (in addition to standard therapies) with longer term measurement of outcomes. Only two studies considered response or remission at 12 months and both these studies found evidence that antibiotics may be or probably are superior to placebo. Conversely, given the high certainty evidence of studies measuring outcomes at three months or less, no further studies in this area appear warranted. Additionally, as there were single studies of other head‐to‐head comparisons, there may be scope for future studies in this area.

It is also clear, as has already been stated, that appropriate completion and reporting of sample size estimation remain a weakness that should be addressed in future works, as well as clear reporting in line with the CONSORT statement to reduce risk of bias and enhance the certainty of the evidence base as a whole, regardless of findings.

History

Protocol first published: Issue 9, 2020

Appendices

Appendix 1. Cochrane CENTRAL and CDSR search strategy (Ovid)

1. exp ulcerative colitis/

2. Inflammatory Bowel Diseases/

3. (colitis or proctocolitis or proctosigmoiditis or proctitis or rectosigmoiditis or rectocolitis or colorectitis or coloproctitis or UC).tw,kw.

4. inflammatory bowel disease*.tw,kw.

5. IBD.tw,kw.

6. or/1‐5

7. exp Anti‐Bacterial Agents/

8. exp *Anti‐Infective Agents/

9. (antibiotic* or anti‐biotic* or antibacterial* or anti‐bacterial*).tw,kw.

10. (antimicrobial* or anti‐microbial* or antiseptic* or anti‐septic*).tw,kw.

11. (bactericid* or bacteriocid*).tw,kw.

12. Bacteriostatic.tw,kw.

13. (ciprofloxacin or metronidazole or levamisole or ornidazole or fusidin or rifaximin or vancomycin or fusidic acid or nitazoxanide or teicoplanin or rifampicin or bacitracin or fidaxomicin or amoxicillin or azithromycin or cephalosporin* or cephalexin or clarithromycin or clindamycin or doxycycline or erythromycin or flouroquinolone* or levofloxacin or macrolide* or nitrofurantoin or penicillin or tetracycline or trimethoprim).mp.

14. or/7‐13

15. 6 and 14

Appendix 2. MEDLINE search strategy (Ovid)

1. exp ulcerative colitis/

2. Inflammatory Bowel Diseases/

3. (colitis or proctocolitis or proctosigmoiditis or proctitis or rectosigmoiditis or rectocolitis or colorectitis or coloproctitis or UC).tw,kw.

4. inflammatory bowel disease*.tw,kw.

5. IBD.tw,kw.

6. or/1‐5

7. exp Anti‐Bacterial Agents/

8. exp *Anti‐Infective Agents/

9. (antibiotic* or anti‐biotic* or antibacterial* or anti‐bacterial*).tw,kw.

10. (antimicrobial* or anti‐microbial* or antiseptic* or anti‐septic*).tw,kw.

11. (bactericid* or bacteriocid*).tw,kw.

12. Bacteriostatic.tw,kw.

13. (ciprofloxacin or metronidazole or levamisole or ornidazole or fusidin or rifaximin or vancomycin or fusidic acid or nitazoxanide or teicoplanin or rifampicin or bacitracin or fidaxomicin or amoxicillin or azithromycin or cephalosporin* or cephalexin or clarithromycin or clindamycin or doxycycline or erythromycin or flouroquinolone* or levofloxacin or macrolide* or nitrofurantoin or penicillin or tetracycline or trimethoprim).mp.

14. or/7‐13

15. 6 and 14

16. randomized controlled trial.pt.

17. controlled clinical trial.pt.

18. random*.mp.

19. placebo.ab.

20. trial.ab.

21. groups.ab.

22. drug therapy.fs.

23. or/16‐22

24. exp animals/ not humans.sh.

25. 23 not 24

26. 15 and 25

Note: Lines 16‐25. RCT filter, used the “Cochrane Highly Sensitive Search Strategy for identifying randomized trials in MEDLINE: sensitivity‐maximizing version (2008 revision); Ovid format”. We made the following minor revision: we used “random*” instead of “randomized.ab” or “randomly.ab.” to capture word variations such as “randomised, randomization, random”.

Appendix 3. Embase search strategy (Ovid)

1. exp ulcerative colitis/

2. inflammatory bowel disease/

3. (colitis or proctocolitis or proctosigmoiditis or proctitis or rectosigmoiditis or rectocolitis or colorectitis or coloproctitis or UC).tw,kw.

4. inflammatory bowel disease*.tw,kw.

5. IBD.tw,kw.

6. or/1‐5

7. exp antibiotic agent/

8. exp *antiinfective agent/

9. (antibiotic* or anti‐biotic* or antibacterial* or anti‐bacterial*).tw,kw.

10. (antimicrobial* or anti‐microbial* or antiseptic* or anti‐septic*).tw,kw.

11. (bactericid* or bacteriocid*).tw,kw.

12. Bacteriostatic.tw,kw.

13. (ciprofloxacin or metronidazole or levamisole or ornidazole or fusidin or rifaximin or vancomycin or fusidic acid or nitazoxanide or teicoplanin or rifampicin or bacitracin or fidaxomicin or amoxicillin or azithromycin or cephalosporin* or cephalexin or clarithromycin or clindamycin or doxycycline or erythromycin or flouroquinolone* or levofloxacin or macrolide* or nitrofurantoin or penicillin or tetracycline or trimethoprim).mp.

14. or/7‐13

15. 6 and 14

16. random:.tw.

17. placebo:.mp.

18. double‐blind:.tw.

19. or/16‐18

20. exp animal/ not human/

21. 19 not 20

22. 15 and 21

Note: Lines #16‐19, RCT filter, used the Hedge Best balance of sensitivity and specificity filter. https://hiru.mcmaster.ca/hiru/HIRU_Hedges_EMBASE_Strategies.aspx

Data and analyses

Comparison 1. Antibiotics versus placebo for induction of remission.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Failure to induce remission at end of study	7	608	Risk Ratio (M‐H, Random, 95% CI)	0.88 [0.74, 1.06]
1.1.1 Antibiotics + steroids vs placebo + steroids	4	228	Risk Ratio (M‐H, Random, 95% CI)	0.78 [0.51, 1.20]
1.1.2 Antibiotics + steroids + 5‐ASA vs placebo + steroids + 5‐ASA	3	330	Risk Ratio (M‐H, Random, 95% CI)	0.96 [0.85, 1.09]
1.1.3 Antibiotics + probiotics + steroids + 5‐ASA vs placebo + probiotics + steroids + 5‐ASA	1	50	Risk Ratio (M‐H, Random, 95% CI)	0.67 [0.37, 1.19]
1.2 Failure to induce remission at 12 months from study end	1	210	Risk Ratio (M‐H, Random, 95% CI)	0.86 [0.74, 0.98]
1.2.1 Antibiotics + steroids + 5‐ASA vs placebo + steroids + 5‐ASA	1	210	Risk Ratio (M‐H, Random, 95% CI)	0.86 [0.74, 0.98]
1.3 Failure to induce clinical response at end of study	3	343	Risk Ratio (M‐H, Random, 95% CI)	0.75 [0.47, 1.22]
1.3.1 Antibiotics + steroids vs placebo + steroids	1	50	Risk Ratio (M‐H, Random, 95% CI)	0.75 [0.39, 1.46]
1.3.2 Antibiotics + steroids + 5‐ASA vs placebo + steroids + 5‐ASA	2	293	Risk Ratio (M‐H, Random, 95% CI)	0.72 [0.33, 1.56]
1.4 Failure to induce clinical response at 12 months from study end	2	293	Risk Ratio (M‐H, Random, 95% CI)	0.66 [0.54, 0.80]
1.4.1 Antibiotics + steroids + 5‐ASA vs placebo + steroids + 5‐ASA	2	293	Risk Ratio (M‐H, Random, 95% CI)	0.66 [0.54, 0.80]
1.5 Serious adverse events	8	691	Risk Ratio (M‐H, Random, 95% CI)	1.19 [0.38, 3.71]
1.5.1 Antibiotics + steroids vs placebo + steroids	4	228	Risk Ratio (M‐H, Random, 95% CI)	1.19 [0.38, 3.71]
1.5.2 Antibiotics + steroids + 5‐ASA vs placebo + steroids + 5‐ASA	4	413	Risk Ratio (M‐H, Random, 95% CI)	Not estimable
1.5.3 Antibiotics + probiotics + steroids + 5‐ASA vs placebo + probiotics + steroids + 5‐ASA	1	50	Risk Ratio (M‐H, Random, 95% CI)	Not estimable
1.6 Withdrawals due to adverse events	8	691	Risk Ratio (M‐H, Random, 95% CI)	2.06 [0.27, 15.72]
1.6.1 Antibiotics + steroids vs placebo + steroids	4	228	Risk Ratio (M‐H, Random, 95% CI)	1.00 [0.06, 15.47]
1.6.2 Antibiotics + steroids + 5‐ASA vs placebo + steroids + 5‐ASA	4	413	Risk Ratio (M‐H, Random, 95% CI)	5.00 [0.24, 102.91]
1.6.3 Antibiotics + probiotics + steroids + 5‐ASA vs placebo + probiotics + steroids + 5‐ASA	1	50	Risk Ratio (M‐H, Random, 95% CI)	Not estimable

Comparison 2. Antibiotics and probiotics versus no intervention/placebo for induction of remission.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Failure to induce remission at end of study	2	98	Risk Ratio (M‐H, Random, 95% CI)	0.68 [0.39, 1.19]
2.1.1 Antibiotics + probiotics + 5‐ASA vs steroids + 5‐ASA	1	48	Risk Ratio (M‐H, Random, 95% CI)	1.00 [0.07, 15.08]
2.1.2 Antibiotics + probiotics + steroids + 5‐ASA vs placebo + steroids + 5‐ASA	1	50	Risk Ratio (M‐H, Random, 95% CI)	0.67 [0.37, 1.19]
2.2 Serious adverse events	2	98	Risk Ratio (M‐H, Random, 95% CI)	1.00 [0.07, 15.08]
2.2.1 Antibiotics + probiotics + 5‐ASA vs steroids + 5‐ASA	1	48	Risk Ratio (M‐H, Random, 95% CI)	1.00 [0.07, 15.08]
2.2.2 Antibiotics + probiotics + steroids + 5‐ASA vs placebo + steroids + 5‐ASA	1	50	Risk Ratio (M‐H, Random, 95% CI)	Not estimable
2.3 Withdrawals due to adverse events	2	98	Risk Ratio (M‐H, Random, 95% CI)	1.00 [0.07, 15.08]
2.3.1 Antibiotics + probiotics + 5‐ASA vs steroids + 5‐ASA	1	48	Risk Ratio (M‐H, Random, 95% CI)	1.00 [0.07, 15.08]
2.3.2 Antibiotics + probiotics + steroids + 5‐ASA vs placebo + steroids + 5‐ASA	1	50	Risk Ratio (M‐H, Random, 95% CI)	Not estimable

Comparison 3. Antibiotics versus 5‐ASA for induction of remission.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Failure to induce clinical response at end of study	1	46	Risk Ratio (M‐H, Random, 95% CI)	2.20 [1.17, 4.14]
3.2 Serious adverse events	1	46	Risk Ratio (M‐H, Random, 95% CI)	Not estimable
3.3 Withdrawals due to adverse events	1	46	Risk Ratio (M‐H, Random, 95% CI)	Not estimable

3.2. Analysis.

Comparison 3: Antibiotics versus 5‐ASA for induction of remission, Outcome 2: Serious adverse events

3.3. Analysis.

Comparison 3: Antibiotics versus 5‐ASA for induction of remission, Outcome 3: Withdrawals due to adverse events

Comparison 4. Antibiotics versus probiotics for induction of remission.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
4.1 Failure to induce remission at end of study	1	50	Risk Ratio (M‐H, Random, 95% CI)	0.47 [0.23, 0.94]
4.2 Serious adverse events	1	50	Risk Ratio (M‐H, Random, 95% CI)	Not estimable
4.3 Withdrawals due to adverse events	1	50	Risk Ratio (M‐H, Random, 95% CI)	Not estimable

4.2. Analysis.

Comparison 4: Antibiotics versus probiotics for induction of remission, Outcome 2: Serious adverse events

4.3. Analysis.

Comparison 4: Antibiotics versus probiotics for induction of remission, Outcome 3: Withdrawals due to adverse events

Comparison 5. Antibiotics versus 5‐ASA for maintenance of remission.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
5.1 Failure to maintain remission at end of study	1	40	Risk Ratio (M‐H, Random, 95% CI)	0.71 [0.47, 1.06]
5.2 Serious adverse events	1	40	Risk Ratio (M‐H, Random, 95% CI)	Not estimable
5.3 Withdrawals due to adverse events	1	40	Risk Ratio (M‐H, Random, 95% CI)	Not estimable

5.2. Analysis.

Comparison 5: Antibiotics versus 5‐ASA for maintenance of remission, Outcome 2: Serious adverse events

5.3. Analysis.

Comparison 5: Antibiotics versus 5‐ASA for maintenance of remission, Outcome 3: Withdrawals due to adverse events

Comparison 6. Antibiotics versus no treatment in mixed population with active and inactive disease.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
6.1 Failure to induce remission at 3 months	1	20	Risk Ratio (M‐H, Random, 95% CI)	0.56 [0.29, 1.07]
6.2 Failure to maintain remission at 12 months	1	20	Risk Ratio (M‐H, Random, 95% CI)	0.22 [0.06, 0.78]
6.3 Serious adverse events	1	20	Risk Ratio (M‐H, Random, 95% CI)	Not estimable
6.4 Withdrawals due to adverse events	1	20	Risk Ratio (M‐H, Random, 95% CI)	Not estimable

6.3. Analysis.

Comparison 6: Antibiotics versus no treatment in mixed population with active and inactive disease, Outcome 3: Serious adverse events

6.4. Analysis.

Comparison 6: Antibiotics versus no treatment in mixed population with active and inactive disease, Outcome 4: Withdrawals due to adverse events

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Bataga 2015.

Study characteristics
Methods	Study design: RCT Setting: Romania, single centre Study period: 2011–2013
Participants	Inclusion criteria: remission of disease at the time of study inclusion; mild to moderate disease severity at initial diagnosis. Initial severity of the disease was evaluated according to UCDAI score. Remission defined as UCDAI score 0–2. All participants included were receiving treatment with mesalazine 3 g/day. Participants were followed up during this period until they had a relapse with minor symptoms (specified as diarrhoea and bloody stools) Exclusion criteria: severe forms of the disease, aged < 18 years, pregnancy, lack of compliance to treatment, presence of infectious colitis at the time of inclusion, presence of dysplasia at the time of inclusion Disease activity: relapsing with minor symptoms, mainly diarrhoea and bloody stools (mean UCDAI score at the time of inclusion = 1, followed up until relapse, which represents the time at which randomisation commenced) Disease duration IG: n/a Disease duration CG: n/a Extent of disease IG: left sided Extent of disease CG: left sided Concurrent therapy IG: mesalazine 3 g/day Concurrent therapy CG: mesalazine 3 g/day Number randomised to IG: 24 Number randomised to CG: 24 Number assessed in IG: 23 Number assessed in CG: 23 Age at randomisation IG: mean age for the whole cohort 53.2 years Age at randomisation CG: mean age for the whole cohort 53.2 years Sex (M/F) IG: 27/21 for the whole cohort Sex (M/F) CG: 27/21 for the whole cohort
Interventions	Length of intervention: 40 days for IG and 3 months for CG. Evaluation at 6 months from beginning of intervention. Participants were followed up during 2011–2013 until they had a relapse with minor symptoms, at which point their treatment with rifaximin (IG) or budesonide (CG) commenced. IG: oral rifaximin 800 mg/day for 1 month + probiotics for 10 days. Rifaximin was administrated for 1 month, followed by a probiotic for 10 days. The probiotic included Lactobacillus acidophilus, Bifidobacterium infantis, and Enterococcus faecium CG: oral budesonide 9 mg/day for 3 months with progressively decreased dose/month All participants were treatment with mesalazine 3 g/day
Outcomes	Primary outcomes Definition of treatment success as defined by authors: remission at 6‐month evaluation from beginning of the intervention. Remission defined as UCDAI score 0–2 Number failing to achieve/maintain remission IG: 1 (infection with C diff) Number failing to achieve/maintain remission CG: 1 (infection with C diff) Secondary outcomes Serious adverse events IG: 1 (infection with C diff) Serious adverse events CG: 1 (infection with C diff) Total adverse events IG: 21 mild abdominal pain, 1 infection with C diff Total adverse events CG: 20 mild abdominal pain, 1 infection with C diff Withdrawal from the study due to adverse events IG: 1 (infection with C diff) Withdrawal from the study due to adverse events CG: 1 (infection with C diff) Clinical response definition by the authors: total UCDAI scores < 2, equivalent to clinical remission of the disease at 6 months' follow‐up Number failing to achieve clinical response IG: 1 Number failing to achieve clinical response CG: 1 Endoscopic response definition by the authors: unclear Number failing to achieve endoscopic response IG: n/a Number failing to achieve endoscopic response CG: n/a Histological response definition by the authors: unclear Number failing to achieve histological response IG: n/a Number failing to achieve histological response CG: n/a Clinical relapse definition by the authors: n/a Endoscopic relapse definition by the authors: n/a Number who had an endoscopic relapse IG: n/a Number who had an endoscopic relapse CG: n/a HRQoL measurement tool: n/a HRQoL IG: n/a HRQoL CG: n/a
Notes	Funding source: not mentioned Conflict of interest: not mentioned Author contact: simonabataga@yahoo.com
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not mentioned by authors. No information given after contacting authors.
Allocation concealment (selection bias)	Unclear risk	Not mentioned by authors. No information given after contacting authors.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not mentioned by authors. No information given after contacting authors.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not mentioned by authors. No information given after contacting authors.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Data reported for all randomised participants.
Selective reporting (reporting bias)	Unclear risk	In their methods, authors mentioned histological and endoscopic evaluations. No such results were presented. No information given after contacting authors.
Other bias	Unclear risk	No baseline comparisons between groups presented. 1 group received the intervention for 1 month, while the other for 3 months. No information given after contacting authors.

Burke 1990.

Study characteristics
Methods	Study design: RCT Setting: Leeds General Infirmary, UK, single centre Study period: 1987–1989
Participants	Inclusion criteria: people with UC relapsing with an acute attack of colitis Exclusion criteria: known Crohn's disease, those who had received antibiotics within the preceding 2 months (other than sulfasalazine), people with fulminant disease, pregnant or lactating women, people with renal impairment and people with evidence of an infective diarrhoea Disease activity: mild to severe IG mild: 7, moderate: 32, severe: 3 CG: mild: 9, moderate: 32, severe: 1 Disease duration IG: 5.6 (SD 5.8) years Disease duration CG: 7.5 (SD 8.8) years Extent of disease IG: proctitis: 12; to splenic flexure: 11; extensive: 19 Extent of disease CG: proctitis: 10; to splenic flexure: 16; extensive: 16 Concurrent therapy IG: participants who were already receiving sulfasalazine were allowed to continue on this drug. The dose and mode of delivery of the steroid was decided by the supervising clinician who allocated them to receive either hydrocortisone enemas (Colifoam, Stafford‐Miller Ltd) or prednisolone 30 mg daily for 1 month as an outpatient, or prednisolone 60 mg (or intravenous equivalent of hydrocortisone) daily as an inpatient depending on the severity and extent of their disease. The inpatient steroid dosage was reduced according to a predetermined protocol in keeping with study authors' current practice. The initial dose of prednisolone for inpatients was reduced at 72 hour intervals by 10 mg until a dose of 30 mg was reached and then by 5 mg decrements to 15 mg, provided that bowel frequency was ≤ 3 per 24 hours, there was absent to mild pain and malaise, no weight loss of > 1 kg, temperature < 37.5 °C, and pulse rate remained < 90 per minute Inpatient prednisolone (60 mg): 17 Outpatient prednisolone (30 mg): 14 Outpatient hydrocortisone: 11 Concurrent therapy CG: same as IG Inpatient prednisolone (60 mg): 16 Outpatient prednisolone (30 mg): 16 Outpatient hydrocortisone: 10 Number randomised to IG: 42 (group 1: 17, group 2: 14, group 3: 11) Number randomised to CG: 42 (group 1: 16, group 2: 16, group 3: 10) Number assessed in IG: 40 Number assessed in CG: 40 Age at randomisation IG: 43.5 (SD 15.8) years Age at randomisation CG: 43.3 (SD 15.7) years Sex (M/F) IG: 21/21 Sex (M/F) CG: 24/18
Interventions	Length of intervention: 21 days for participants in subgroup 1 (inpatient) and 28 days for participants in subgroups 2 and 3 (outpatient) IG (3 subgroups): oral tobramycin 120 mg twice daily for 7 days as an adjunct to steroid therapy (subgroup 1: inpatients prednisolone 60 mg, subgroup 2: outpatients prednisolone 30 mg, subgroup 3: outpatients hydrocortisone) CG (3 subgroups): placebo for 7 days as an adjunct to steroid therapy (subgroup 1: inpatients prednisolone 60 mg, subgroup 2: outpatients prednisolone 30 mg, subgroup 3: outpatients hydrocortisone) Participants who were already receiving sulfasalazine were allowed to continue. The dose and delivery of steroid was decided but the supervising clinician who allocated them to receive either coliform enemata 1 twice daily, prednisolone 30 mg daily for 1 month as an outpatient or prednisolone 60 mg daily (or intravenous equivalent of hydrocortisone) as an inpatient depending on the severity and extent of their disease
Outcomes	Primary outcomes Definition of treatment success as defined by authors: at endpoint participants were considered to be in symptomatic remission if their bowel actions were < 3/day without blood, there was no malaise, abdominal pain, anorexia, vomiting, abdominal tenderness, fever, or any complication or active extraintestinal manifestation of ulcerative colitis, based on the Truelove and Witts criteria (Truelove 1955) Number failing to achieve/maintain remission IG: 11 (9/40 + 2 dropouts) Number failing to achieve/maintain remission CG: 24 (22/40 + 2 dropouts) Secondary outcomes Serious adverse events IG: 1 (required surgery during trial, unrelated to the intervention according to authors) Serious adverse events CG: 1 (required surgery during trial, unrelated to the intervention according to authors) Total adverse events IG: 1 (required surgery during trial, unrelated to the intervention according to authors) Total adverse events CG: 1 (required surgery during trial, unrelated to the intervention according to authors) Withdrawal from the study due to adverse events IG: 1 (required surgery during trial, unrelated to the intervention according to authors) Withdrawal from the study due to adverse events CG: 1 (required surgery during trial, unrelated to the intervention according to authors) Clinical response definition by the authors: n/a Number failing to achieve clinical response IG: n/a Number failing to achieve clinical response CG: n/a Endoscopic response definition by the authors: sigmoidoscopic appearances were scored: 0 = normal; 1 = loss of vascular pattern but no granularity; 2 = granularity; 3 = granularity plus contact bleeding; 4 = spontaneous bleeding, friability, and frank ulceration Number failing to achieve endoscopic response IG: n/a, only scores presented: IG day 0: 3.5 (SD 0.7), IG day 7: 2.2 (SD 1.2), IG endpoint: 1.3 (SD 1.5) Number failing to achieve endoscopic response CG: n/a, only scores presented CG day 0: 3.4 (SD 0.7), CG day 7: 2.7 (SD 1.0), CG endpoint: 1.9 (SD 1.5) Histological response definition by the authors: histological scoring was using a scale of 0–3 for mucous depletion, surface degeneration, and polymorph infiltration. Thus, the total active inflammatory score was 0–9, and was determined by MFD without knowledge of the clinical severity, treatment, or the source and timing of the biopsy Number failing to achieve histological response IG: 35 (7 achieved a histological score of 0) IG day 0: 5.5 (SD 2.7), IG day 7: 4.4 (SD 2.7), IG endpoint: 2.8 (SD 2.5) Number failing to achieve histological response CG: 42 (0 achieved a histological score of 0) CG day 0: 6.4 (SD 2.1), CG day 7: 5.5 (SD 2.7), CG endpoint: 4.0 (SD 2.6) Clinical relapse definition by the authors: n/a Endoscopic relapse definition by the authors: n/a Number who had an endoscopic relapse IG: n/a Number who had an endoscopic relapse CG: n/a HRQoL measurement tool: n/a HRQoL IG: n/a HRQoL CG: n/a
Notes	Funding source: special trustees of the Leeds General Infirmary, not by Eli Lilly, which is mentioned in the text, who only provided funding for the double blind trial drugs and consumables (author's response) Conflict of interest: none (author's response) Author contact: Denis.Burke@ncic.nhs.uk
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was stratified in blocks for each of the 3 treatment groups, that is inpatient, outpatient with systemic steroid, or outpatient with steroid enema. They were further stratified for first attack or relapse.
Allocation concealment (selection bias)	Low risk	Not mentioned in the paper, but author responded and said their pharmacy independently drew up the randomisation and provided the trial medication or control preparation according to this with clinicians unaware of this. The code was not broken until after assessments and data collection.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Described as double blind.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Not mentioned in the paper, but author responded to say that the assessors were blind.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Results presented for all randomised participants except 2 dropouts.
Selective reporting (reporting bias)	Unclear risk	Results were not clearly presented for the subgroups. Author responded that, unfortunately, following several moves, retirement and lack of electronic records in 1990, he was unable to locate that data.
Other bias	Low risk	None.

Gilat 1987.

Study characteristics
Methods	Study design: RCT, cross‐over Setting: 6 medical centres in Israel Study period: n/a
Participants	Inclusion criteria: people diagnosed with UC in relapse (> 7 days) or new patients with symptoms for > 1 month, in whom specific colitis had been excluded (Salmonella, Shigella, Entamoeba histolytica, and Campylobacter) Exclusion criteria: known drug allergy to metronidazole and sulfasalazine, people with glucose‐6‐phosphate dehydrogenase deficiency without prior uneventful and safe exposure to sulphonamides, those with uncertain diagnosis, those who were unco‐operative and unreliable, those with a severe attack of UC (> 9 bowel movements daily for > 2 days, fever > 38 °C for > 2 days, pulse > 100/minute (consistently), white blood count > 20,000/mm3, toxic megacolon, severe abdominal pain and tenderness for > 2 days), aged < 16 years and pregnant women Disease activity: active, acute, non‐severe Disease duration IG: n/a Disease duration CG: n/a Extent of disease IG: proctitis: 11, 15 more extensive colitis Extent of disease CG: proctitis: 12, 7 more extensive colitis Concurrent therapy IG: n/a Concurrent therapy CG: n/a Number randomised to IG: 26 Number randomised to CG: 20 Number assessed in IG: 23 Number assessed in CG: 19 Age at randomisation IG: n/a Age at randomisation CG: n/a Sex (M/F) IG: 12/11 (without dropouts) Sex (M/F) CG: 7/12 (without dropouts)
Interventions	Length of intervention: 28 days IG: metronidazole 1.35 g/day CG: sulfasalazine 4.5 g/day Participants received 9 tablets/day. Each tablet contained either metronidazole 0.15 g or sulfasalazine 0.5 g. The coated tablets had an identical appearance and taste. Therapy was started with 4 tablets the first day, 6 tablets the second, 8 tablets the third, and 9 tablets from the fourth day on. Therapy continued with 9 tablets/day but reduction in dosage because of GI intolerance was allowed to 7 tablets/day but no fewer. Participants unable to tolerate 7 tablets were excluded. Participants had not received specific therapy, steroids, or sulfasalazine for ≥ 7 days prior to trial. Symptomatic therapy (antidiarrhoeal or antiemetic) was avoided if possible Cross‐over (4 participants): the acute study was divided into 2 steps according to the therapy given. Once a participant was in remission for ≥ 1 month a cross‐over therapy was administered during the next exacerbation using the same method. The co‐ordinator, who was unblinded, supplied the physicians with the appropriate medication, to keep the study double blinded. The cross‐over trial was discontinued when results of the main trial became apparent.
Outcomes	Primary outcomes Definition of treatment success as defined by authors: clinical response (see below) Number failing to achieve/maintain remission IG: n/a Number failing to achieve/maintain remission CG: n/a Secondary outcomes Serious adverse events IG: 0 Serious adverse events CG: 0 Total adverse events IG: "A few patients had minor GI side‐effects such as bitter taste in the mouth or nausea" Total adverse events CG: "A few patients had minor GI side‐effects such as bitter taste in the mouth or nausea" Withdrawal from the study due to adverse events IG: 0 Withdrawal from the study due to adverse events CG: 0 Clinical response definition by the authors: improvement of clinical symptoms (blood in stools, number of bowel movements) and sigmoidoscopy grade Number failing to achieve clinical response IG: 20 (17 + 3 dropouts) Number failing to achieve clinical response CG: 7 (6 + 1 dropout) Endoscopic response definition by the authors: sigmoidoscopies were graded as follows: grade 0 = normal vascular pattern and a shiny mucosa, grade 1 = pinkish mucosa without vascular pattern and with contact bleeding, grade 2 = red inflamed mucosa with spontaneous bleeding and exudation. Patients with grade 0 were not included in the trial. Participants were considered improved at the end of the trial if blood disappeared, stools improved in quantity and quality, and sigmoidoscopic appearance improved by at least on grade. Number failing to achieve endoscopic response IG: 17 Number failing to achieve endoscopic response CG: 7 Histological response definition by the authors: n/a Number failing to achieve histological response IG: n/a Number failing to achieve histological response CG: n/a Clinical relapse definition by the authors: n/a Endoscopic relapse definition by the authors: n/a Number who had an endoscopic relapse IG: n/a Number who had an endoscopic relapse CG: n/a HRQoL measurement tool: n/a HRQoL IG: n/a HRQoL CG: n/a
Notes	Funding source: the authors appreciated the help of the Teva Pharmaceutical Co, Tel Aviv, Israel, in preparing and supplying the coded medicines. Conflict of interest: not mentioned Author contact: Dr T Gilat fireman@hy.health.gov.il
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Stratified randomisation, but the process was not explained. No information given after contacting authors.
Allocation concealment (selection bias)	Unclear risk	No information provided and no information given after contacting authors.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	The authors considered it a double blind study and mentioned the 2 tablets were identical. No more information was given after contacting authors.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not mentioned and no information given after contacting authors.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Results were presented for all participants who completed the study.
Selective reporting (reporting bias)	Unclear risk	The intended outcomes were not clear in the methods; the presented remission, clinical response, and endoscopic results were mixed together; and the cross‐over design was abandoned with unclear results, when authors had the results of the first phase.
Other bias	Unclear risk	Randomisation numbers did not seem balanced. No information given after contacting authors.

Gilat 1989.

Study characteristics
Methods	Study design: RCT, cross‐over Setting: 5 medical centres in Israel Study period: n/a
Participants	Inclusion criteria: people with confirmed UC in remission for 1–11 months, who had been treated with sulfasalazine only or untreated. Patients had to have been off steroids for ≥ 1 month. The presence of remission was confirmed both clinically and sigmoidoscopically Exclusion criteria: known drug allergy to metronidazole or sulfasalazine, glucose‐6‐phosphate dehydrogenase deficiency without prior uneventful and safe exposure to sulphonamides, uncertain diagnosis, unco‐operative and unreliable patients or those refusing to participate, aged < 16 years and pregnant women Disease activity: remission Disease duration IG: n/a Disease duration CG: n/a Extent of disease IG: 9 proctitis, 11 more extensive colitis (after dropouts) Extent of disease CG: 7 proctitis, 11 more extensive colitis (after dropouts) Concurrent therapy IG: n/a Concurrent therapy CG: n/a Number randomised to IG: 22 Number randomised to CG: 18 Number assessed in IG: 20 Number assessed in CG: 15 Age at randomisation IG: mean 49.2 (SD 14.41) years after dropouts Age at randomisation CG: mean 47.12 (SD 16.0) years after dropouts Sex (M/F) IG: 10/10 after dropouts Sex (M/F) CG: 7/8 after dropouts
Interventions	Length of intervention: 12 months or until recurrence. Follow‐up every 3 months until end of study or recurrence IG: metronidazole 0.6 g/day CG: sulfasalazine 2 g/day Cross‐over: participants developing an acute attack of UC were withdrawn from the trial and treated openly as necessary. After remission was induced and maintained for over 1 month with sulfasalazine 2–3 g only, they were offered a cross‐over trial. The alternative (coded) medicine was provided by a non‐blinded observer. The cross‐over trial followed an identical design. 6 participants
Outcomes	Primary outcomes Definition of treatment success as defined by authors: relapse was defined as the appearance of blood in the stools with an increase in stool frequency, with or without diarrhoea. The presence of active colitis was confirmed sigmoidoscopically in each case. Stool cultures were performed, excluding infectious colitis. Treatment success would be maintenance of remission Number failing to achieve/maintain remission IG: 13 (11 + 2 dropouts) Number failing to achieve/maintain remission CG: 15 (12 + 3 dropouts) Secondary outcomes Serious adverse events IG: 0 Serious adverse events CG: 0 Total adverse events IG: "A few patients had minor GI side‐effects such as bitter taste in the mouth or nausea" Total adverse events CG: "A few patients had minor GI side‐effects such as bitter taste in the mouth or nausea" Withdrawal from the study due to adverse events IG: 0 Withdrawal from the study due to adverse events CG: 0 Clinical response definition by the authors: n/a Number failing to achieve clinical response IG: n/a Number failing to achieve clinical response CG: n/a Endoscopic response definition by the authors: no worsening of their sigmoidoscopy finding or improvement Number failing to achieve endoscopic response IG: 13 (11 + 2 dropouts) Number failing to achieve endoscopic response CG: 15 (12 + 3 dropouts) Histological response definition by the authors: n/a Number failing to achieve histological response IG: n/a Number failing to achieve histological response CG: n/a Clinical relapse definition by the authors: relapse was defined as the appearance of blood in the stools with an increase in stool frequency, with or without diarrhoea Endoscopic relapse definition by the authors: n/a Number who had an endoscopic relapse IG: n/a Number who had an endoscopic relapse CG: n/a HRQoL measurement tool: n/a HRQoL IG: n/a HRQoL CG: n/a
Notes	Funding source: the authors appreciated the help of the Teva Pharmaceutical Co, Tel Aviv, Israel, in preparing and supplying the coded medicines. Conflict of interest: not mentioned Author contact: Dr T Gilat; fireman@hy.health.gov.il; barmeir@netvision.net
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Stratified randomisation, but the process was not explained. No information given after contacting authors.
Allocation concealment (selection bias)	Unclear risk	No information provided and no information given after contacting authors.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	The authors considered it a double blind study and mention the 2 tablets were identical. No more information was given after contacting authors.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not mentioned and no information given after contacting authors.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Results are presented for all participants who completed the study.
Selective reporting (reporting bias)	Low risk	The remission and endoscopic response results are presented.
Other bias	Unclear risk	Randomisation numbers do not seem balanced. No information given after contacting authors.

Mantzaris 1994.

Study characteristics
Methods	Study design: RCT Setting: First Department of Gastroenterology, Evangelismos Hospital, Athens, Greece, single‐centre Study period: n/a
Participants	Inclusion criteria: all participants had an acute, severe attack of UC, according to the criteria of Truelove and Witts (Truelove 1955). In particular, they were passing > 8 bowel movements daily and presented with fever (> 37.5 °C) tachycardia (pulse rate > 100/minute), raised erythrocyte sedimentation rate (> 30 mm/hour) and either a low serum albumin (< 3.5 g/L) or anaemia (haemoglobin < 10 g/dL) or both. UC had previously been diagnosed by colonoscopy and colonic histology in participants presenting with a relapse of disease Exclusion criteria: n/a Disease activity: acute, severe Disease duration IG: median 2 (range 0.5–5) years Disease duration CG: median 2.2 (range 0.6–6) years Extent of disease IG: total colitis 10, left sided colitis 7, proctosigmoiditis 2 Extent of disease CG: total colitis 11, left sided colitis 8, proctosigmoiditis 1 Concurrent therapy IG: n/a Concurrent therapy CG: n/a Number randomised to IG: 19 Number randomised to CG: 20 Number assessed in IG: 19 Number assessed in CG: 20 Age at randomisation IG: median 42 (range 16–65) years Age at randomisation CG: median 41 (range 15–66) years Sex (M/F) IG: 9/10 Sex (M/F) CG: 9/11
Interventions	Length of intervention: 10 days (follow‐up at 8 months) IG: metronidazole 0.5 g 3 times daily + tobramycin 4 mg/kg in divided doses every 8 hours CG: identical placebo drugs All participants received a standard therapeutic regimen for 10 days consisting of intravenous hydrocortisone (100 mg 4 times daily) and hydrocortisone 100 mg in rectal drips twice daily. Participants also received parenteral nutrition (2500 kcal and nitrogen 18 g daily) and were not allowed to eat or drink anything except water. Electrolytes losses were substituted and blood was transfused to maintain a haemoglobin level > 10 g/dL
Outcomes	Primary outcomes Definition of treatment success as defined by authors: participants were classified as responders or non‐responders by the method described by Chapman 1986, in brief, participants passing ≤ 3 formed stools or forming stool daily without rectal bleeding or systemic signs of severe colitis were considered to be improved (responders) and were given oral therapy. Participants who either did not improve or who deteriorated or developed complications (non‐responders), proceeded to emergency colectomy Number failing to achieve/maintain remission IG: 7 (the remainder of the participants did not relapse within 8 months) Number failing to achieve/maintain remission CG: 7 (the remainder of the participants did not relapse within 8 months) Secondary outcomes Serious adverse events IG: 3 "none had to be withdrawn from the trial except patients who developed toxic megacolon" Serious adverse events CG: 2 "none had to be withdrawn from the trial except patients who developed toxic megacolon" Total adverse events IG: "mild, transient, steroid‐related side effects" Total adverse events CG: "mild, transient, steroid‐related side effects" Withdrawal from the study due to adverse events IG: 0 "none had to be withdrawn from the trial except patients who developed toxic megacolon" 3, but they are included in the analysis Withdrawal from the study due to adverse events CG: 0 "none had to be withdrawn from the trial except patients who developed toxic megacolon" 2, but they are included in the analysis Clinical response definition by the authors: same as primary outcome Number failing to achieve clinical response IG: 7 Number failing to achieve clinical response CG: 7 Endoscopic response definition by the authors: n/a Number failing to achieve endoscopic response IG: n/a Number failing to achieve endoscopic response CG: n/a Histological response definition by the authors: n/a Number failing to achieve histological response IG: n/a Number failing to achieve histological response CG: n/a Clinical relapse definition by the authors: n/a Endoscopic relapse definition by the authors: n/a Number who had an endoscopic relapse IG: n/a Number who had an endoscopic relapse CG: n/a HRQoL measurement tool: n/a HRQoL IG: n/a HRQoL CG: n/a
Notes	Funding source: none (author's response) Conflict of interest: none (author's response) Author contact: Gerassimos Mantzaris; gjmantzaris@gmail.com
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The author provided the information that the trial used envelopes to draw treatment or no treatment.
Allocation concealment (selection bias)	Low risk	The author provided the information that sealed envelopes were used that were handled by a senior member of staff at the hospital who was unrelated to the study and distributed by the nursing staff who were also unrelated to the study.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Identical placebo drugs. Author confirmed that both participants and personnel were blinded.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	The author provided the information that the outcome assessor was blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Data presented for all randomised participants.
Selective reporting (reporting bias)	Low risk	All outcomes presented in the results.
Other bias	Low risk	None.

Mantzaris 1997.

Study characteristics
Methods	Study design: RCT Setting: First Department of Gastroenterology, Evangelismos Hospital, Athens, Greece, single‐centre Study period: n/a
Participants	Inclusion criteria: all participants presented with an exacerbation of UC that was graded as mild in 37 participants or moderate in 33 participants according to the criteria of Truelove and Witts (Truelove 1955) Exclusion criteria: n/a Disease activity: mild (IG 18, CG 16) to moderate (IG 20, CG 16) Disease duration IG: median 2 (range 0.5–5) years Disease duration CG: median 2.2 (range 0.7–6) years Extent of disease IG: total colitis 10, left sided colitis 11, proctosigmoiditis 13 Extent of disease CG: total colitis 11, left sided colitis 12, proctosigmoiditis 13 Concurrent therapy IG: n/a Concurrent therapy CG: n/a Number randomised to IG: 34 Number randomised to CG: 36 Number assessed in IG: 34 Number assessed in CG: 36 Age at randomisation IG: median 41 (range 16–65) years Age at randomisation CG: median 42 (range 15–66) years Sex (M/F) IG: 16/18 Sex (M/F) CG: 17/19
Interventions	Length of intervention: 14 days and follow‐up at week 7 or 9 IG: oral ciprofloxacin 250 mg twice daily, 32 participants (as reported in publication; should be 34) CG: identical placebo drugs (36 participants) All participants received a standard therapeutic regimen consisting of oral olsalazine 500 mg twice daily with meals, and oral and rectal steroids for 7–9 weeks, depending on the activity of disease at study entry. Participants with a mild exacerbation of UC received initially oral prednisolone 20 mg once daily for 4 weeks; prednisolone was progressively tailed off in the next 3 weeks by reducing the daily dose by 5 mg every week. Participants with moderate attacks of disease received initially prednisolone 40 mg orally once daily for 1 week and then 30 mg for a second week. Prednisolone was then continued as in cases of mild flare up of UC. All participants received betamethasone enemas (2 g at night) for 7–9 weeks.
Outcomes	Primary outcomes Definition of treatment success as defined by authors: participants were assessed using a blinded method by a single investigator on study entry, at the end of the trial (day 14) and at the end of the steroid treatment (week 7 or 9). On day 14, participants passing < 3 formed stools without rectal bleeding were considered to be improved and were given only steroids for the remaining period according to the study protocol; on the contrary, participants who either did not improve, or who deteriorated or developed complications within the first 2 weeks of therapy were either given higher doses of oral steroids or were admitted to the hospital for an intravenous steroid regimen. On completion of the steroid therapy, participants were reassessed using a blind method. Participants passing formed stools without rectal bleeding, with a normal haematological and biochemical profile, a normal sigmoidoscopic appearance and no histological evidence of acute and chronic inflammation on rectal biopsies were considered to be in remission (responders). Participants who either did not achieved remission, or who deteriorated or developed complications during the study were considered as non‐responders Number failing to achieve/maintain remission IG: 10 (the other participants did not relapse within 2 months) Number failing to achieve/maintain remission CG: 10 (the other participants did not relapse within 2 months) Secondary outcomes Serious adverse events IG: 0 Serious adverse events CG: 0 Total adverse events IG: 0 Total adverse events CG: 0 Withdrawal from the study due to adverse events IG: 0 Withdrawal from the study due to adverse events CG: 0 Clinical response definition by the authors: same as primary outcome Number failing to achieve clinical response IG: 10 Number failing to achieve clinical response CG: 10 Endoscopic response definition by the authors: n/a Number failing to achieve endoscopic response IG: n/a Number failing to achieve endoscopic response CG: n/a Histological response definition by the authors: n/a Number failing to achieve histological response IG: n/a Number failing to achieve histological response CG: n/a Clinical relapse definition by the authors: n/a Endoscopic relapse definition by the authors: n/a Number who had an endoscopic relapse IG: n/a Number who had an endoscopic relapse CG: n/a HRQoL measurement tool: n/a HRQoL IG: n/a HRQoL CG: n/a
Notes	Funding source: none (author's response) Conflict of interest: none (author's response) Author contact: Gerassimos Mantzaris <gjmantzaris@gmail.com>
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The author provided the information that the trial used envelopes to draw treatment or no treatment.
Allocation concealment (selection bias)	Low risk	The author provided the information that sealed envelopes were used which were handled by a senior member of staff at the hospital who was unrelated to the study and distributed by the nursing staff who were also unrelated to the study.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Identical placebo and participants were assessed using a blinded method, as confirmed by the author.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Participants were assessed used a blinded method by a single investigator.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Data presented for all randomised participants.
Selective reporting (reporting bias)	Low risk	All outcomes presented in the results.
Other bias	Low risk	None.

Mantzaris 2001.

Study characteristics
Methods	Study design: RCT Setting: A' Gastroenterology Clinic, Evangelismos Hospital, Athens, Greece, single‐centre Study period: January 1995 and October 1999
Participants	Inclusion criteria: recent acute, severe attack of UC, according to the criteria of Truelove and Witts (Truelove 1955): passing > 8 bowel motions daily with mucous and blood, and presented with fever, tachycardia, a raised erythrocyte sedimentation rate (> 30 mmHg), and a low level of haemoglobin (< 10 g/dL) and serum albumin (< 3.5 g/L) Exclusion criteria: the isolation of pathogens from obtained faecal samples was an exclusion criterion Disease activity: acute, severe Disease duration IG: median 4 (range 3–7) years Disease duration CG: median 4.5 (range 2–8) years Extent of disease IG: extensive colitis 15, left sided colitis 11, proctosigmoiditis 3 Extent of disease CG: extensive colitis 13, left sided colitis 11, proctosigmoiditis 2 Concurrent therapy IG: n/a Concurrent therapy CG: n/a Number randomised to IG: 29 Number randomised to CG: 26 Number assessed in IG: 29 Number assessed in CG: 26 Age at randomisation IG: median 42 (range 18–65) years Age at randomisation CG: median 41 (range 19–65) years Sex (M/F) IG: 14/15 Sex (M/F) CG: 12/14
Interventions	Length of intervention: 10 days (follow‐up at 3 months) IG: intravenous ciprofloxacin (400 mg twice daily CG: identical placebo Participants received parenteral nutrition (2500 kcal and nitrogen 14 g daily) and were not allowed to eat or drink anything except water and oral glucose solutions. Care was taken to substitute for electrolyte losses. Blood was transfused to maintain haemoglobin levels over 10 g/dL All participants received a standard therapeutic regimen for 10 days consisting of intravenous hydrocortisone 400 mg daily in divided doses every 6 hours and hydrocortisone 100 mg in rectal drips twice daily
Outcomes	Primary outcomes Definition of treatment success as defined by authors: participants passing < 3 forming or formed stools daily without any visible blood or systemic signs of severe colitis, were considered to be improved (responders) and were given food orally and oral corticosteroids. Participants who either did not improve or deteriorated progressively were considered as non‐responders, and were offered intravenous ciclosporin if consented; otherwise, they proceeded to colectomy. Participants who showed a dramatic deterioration or developed complications at any time during the 10‐day treatment regimen were also considered as non‐responders and proceeded to emergency colectomy Number failing to achieve/maintain remission IG: 6 (none of the other participants relapsed within 3 months) Number failing to achieve/maintain remission CG: 6 (none of the other participants relapsed within 3 months) Secondary outcomes Serious adverse events IG: 2 had toxic megacolon Serious adverse events CG: 2 had toxic megacolon Total adverse events IG: mild, transient, steroid related adverse events Total adverse events CG: mild, transient, steroid related adverse events Withdrawal from the study due to adverse events IG: 0 (2 had toxic megacolon but are counted in the results) Withdrawal from the study due to adverse events CG: 0 (2 had toxic megacolon but are counted in the results) Clinical response definition by the authors: same as primary outcome Number failing to achieve clinical response IG: 6 Number failing to achieve clinical response CG: 6 Endoscopic response definition by the authors: unclear but they mention that 9/12 non‐responders underwent sigmoidoscopy and revealed a mildly inflamed rectum but [no] severe inflammation proximal to the sigmoid colon with numerous pseudopolyps, deep ulcers and extensive areas of denuded mucosa Number failing to achieve endoscopic response IG: n/a Number failing to achieve endoscopic response CG: n/a Histological response definition by the authors: n/a Number failing to achieve histological response IG: n/a Number failing to achieve histological response CG: n/a Clinical relapse definition by the authors: n/a Endoscopic relapse definition by the authors: n/a Number who had an endoscopic relapse IG: n/a Number who had an endoscopic relapse CG: n/a HRQoL measurement tool: n/a HRQoL IG: n/a HRQoL CG: n/a
Notes	Funding source: none (author's response) Conflict of interest: none (author's response) Author contact: Gerassimos Mantzaris; gjmantzaris@gmail.com
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The author provided the information that the trial used envelopes to draw treatment or no treatment.
Allocation concealment (selection bias)	Low risk	The author provided the information that sealed envelopes were used that were handled by a senior member of staff at the hospital who was unrelated to the study and distributed by the nursing staff who were also unrelated to the study.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"Double‐blinded" and identical placebo. The author confirmed that both participants and personnel were blinded.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	The author provided the information that the outcome assessor was blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Data presented for all randomised participants.
Selective reporting (reporting bias)	Low risk	All outcomes presented in the results.
Other bias	Low risk	None.

Mishra 2020.

Study characteristics
Methods	Study design: RCT Setting: single centre, Chandigarh, India Study period: April 2019 to March 2020
Participants	Inclusion criteria: acute severe UC as defined by modified Truelove and Witt's criteria (Truelove 1955) () : ≥ 6 bloody stools with 1 of the following: temperature > 37.8 °C, haemoglobin < 10.5 g/dL, pulse rate > 90/minute, CRP > 30 mg/L Exclusion criteria: people who refused consent; aged ≤ 12 years; pregnant or lactating women; severe uncontrolled comorbidities such as diabetes mellitus, hypertension, coronary artery disease, chronic liver disease, chronic kidney disease; history of antibiotic use in the past 1 month, excluding those who received a single dose of antibiotic in past 24 hours; history of hypersensitivity or contraindication to steroids or the test drugs; presence of toxic megacolon as defined by transverse colon diameter > 5.5 cm on abdominal X‐ray; people with evidence of sepsis or active infection Disease activity: active acute severe UC (bloody diarrhoea) Disease duration IG: median 24 (IQR 6 to 55) months Disease duration CG: median 24 (IQR 4.5 to 48) months Extent of disease IG: proctitis 2; left sided colitis 7, extensive colitis 11 Extent of disease CG: proctitis 0; left sided colitis 13, extensive colitis 11 Concurrent therapy for both groups (quote): "methylprednisolone 60 mg intravenous once daily, anticoagulation (unfractionated heparin 5000 IU subcutaneously twice daily), intravenous fluid supplementation, and correction of electrolytes. Participants were allowed orally and were prescribed a diet of 25–30 kcal/kg/day; oral 5‐ASA". Number randomised to IG: 25 Number randomised to CG: 25 Number assessed in IG: 25 Number assessed in CG: 25 Age at randomisation IG: median 33 (IQR 25 to 45) years Age at randomisation CG: median 33 (IQR 20 to 50) years Sex (M/F) IG: 13/12 Sex (M/F) CG: 10/15
Interventions	Length of intervention: therapy for 5 days and follow‐up at 28 days from beginning of study IG: intravenous ceftriaxone 1 g twice daily + intravenous metronidazole 500 mg 3 times daily. Ceftriaxone was prepared by reconstitution of the crystallised powder with 100 mL of normal saline. Metronidazole was available as a 100 mL solution in prepacked containers CG: intravenous placebo infusions 5 times daily of normal saline of similar quantity, colour, and from similarly shaped bottles Standard care consisting of steroids, prophylactic anticoagulation, and oral nutrition was administered to both groups Those participants who had an incomplete or no response as per Oxford criteria (Travis 1996) were managed as per the preferences of the treating clinician and the participants. Non‐responsive participants were eligible for either second line medical rescue therapy (ciclosporin and thiopurines or infliximab) or colectomy.
Outcomes	Primary outcomes Definition of treatment success as defined by authors: clinical response as defined by the Oxford criteria (Travis 1996) at day 3 of therapy: complete response = stool frequency < 3; partial response = stool frequency 3–8 and CRP < 45 mg/L; no response = stool frequency > 8 or stool frequency 3–8 with CRP > 45 mg/L Number failing to achieve/maintain remission IG: non‐response: 18 participants (non‐responders + partial responders) Number failing to achieve/maintain remission CG: non‐response: 19 participants (non‐responders + partial responders) Secondary outcomes Serious adverse events IG: 0 Serious adverse events CG: 0 Total adverse events IG: 0 Total adverse events CG: 1 participant had evidence of Entamoeba cyst on stool microscopy and was started intravenous metronidazole for the same Withdrawal from the study due to adverse events IG: 0 Withdrawal from the study due to adverse events CG: 0 Clinical response definition by the authors: partial response = stool frequency 3–8 and CRP < 45 mg/L Number failing to achieve clinical response IG: 9 (non‐responders) Number failing to achieve clinical response CG: 7 (non‐responders) Endoscopic response definition by the authors: n/a Number failing to achieve endoscopic response IG: n/a Number failing to achieve endoscopic response CG: n/a Histological response definition by the authors: n/a Number failing to achieve histological response IG: n/a Number failing to achieve histological response CG: n/a Clinical relapse definition by the authors: n/a Endoscopic relapse definition by the authors: n/a Number who had an endoscopic relapse IG: n/a Number who had an endoscopic relapse CG: n/a HRQoL measurement tool: n/a HRQoL IG: n/a HRQoL CG: n/a
Notes	Funding source: internal as part of a medical education and research institution. Conflict of interest: none according to authors Author contact: Vishal Sharma; docvishalsharma@gmail.com
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer generated randomised table with variable block sizes.
Allocation concealment (selection bias)	Low risk	Sequentially numbered opaque sealed envelopes.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Placebo infusion in similar containers and amount. The nursing personnel involved in delivering the regimen were aware of the allocation as the allocation was revealed to them after opening the envelope.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	The treating clinicians or the assessors were not aware of the allocation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Date presented for all randomised participants.
Selective reporting (reporting bias)	Low risk	All presented.
Other bias	Low risk	None.

Ohkusa 2005.

Study characteristics
Methods	Study design: RCT Setting: Department of Gastroenterology, Juntendo University Hospital, Japan, single‐centre Study period: participants enrolled in the study between August and October 2001
Participants	Inclusion criteria: people with chronic relapsing UC (1 relapse per year) who had been followed up for > 1 year. Eligibility criteria for study entry were high serum titres of immunoglobulins (IgG, IgA and IgM) to Fusobacterium varium, measured using a previously described enzyme‐linked immunosorbent assay, and mild‐to‐moderate disease with a score for rectal histology of ≥ 2, based on a scale of 1 (normal) to 5 (ulceration, erosion, or necrosis of the mucosa with cellular infiltration) Exclusion criteria: low serum titres of immunoglobulins (< 5 U/mL) to F varium, or if they required intravenous corticosteroid therapy. In addition, participants who had taken antibiotics within 4 weeks of study entry and had C diff or other stool pathogens at entry were excluded Disease activity: mild to moderate according to the published paper, however, in our correspondence they mentioned that they defined remission as a symptom score < 3, and based on this definition 7/10 participants in the IG were in remission while only 1/10 in the CG was in remission Disease duration IG: median 13 (range 3–45) years Disease duration CG: median 11 (range 3–26) years Extent of disease IG: pancolitis 7, left sided colitis 2, proctitis 1 Extent of disease CG: pancolitis 7, left sided colitis 3, proctitis 0 Concurrent therapy IG: sulfasalazine 9, 5‐ASA 1, glucocorticoid 1, probiotics 4 Concurrent therapy CG: sulfasalazine 9, 5‐ASA 1, glucocorticoid 2, probiotics 5 Number randomised to IG: 10 Number randomised to CG: 10 Number assessed in IG: 10 Number assessed in CG: 10 Age at randomisation IG: median 38 (range 21–58) years Age at randomisation CG: median 41 (range 16–73) years Sex (M/F) IG: 5/5 Sex (M/F) CG: 7/3
Interventions	Length of intervention: 2 weeks and follow‐up at 3–5 months and 12–14 months IG: amoxicillin 500 mg 3 times daily + tetracycline 500 mg 3 times daily + metronidazole 250 mg 3 times daily CG: no treatment Any concomitant medication for UC, such as sulfasalazine, 5‐ASA, prednisolone, probiotics, or a combination of these being taken at the start of the study was continued at a stable dose for a set time prior to inclusion in the study
Outcomes	Primary outcomes Definition of treatment success as defined by authors: based on the Lichtiger symptom score of disease activity (Lichtiger 1990), with scores range 0–21. In our correspondence the author mentioned that their definition of remission was a symptom score < 3. Clinical relapse was defined as the reappearance of visible blood in the stools for 2 consecutive days or the recurrence of frequent diarrhoea (> 6 bowel movements/day), or both; nocturnal diarrhoea or abdominal cramping Number failing to achieve/maintain remission IG: 5 at 3 months and 2 at 12 months Number failing to achieve/maintain remission CG: 9 at 3 months and 9 at 12 months Secondary outcomes Serious adverse events IG: 0 Serious adverse events CG: 0 Total adverse events IG: 3 out of 10 participants experienced metallic taste, 2 had malaise, 1 had watery diarrhoea without blood, and 1 had urticaria and fever Total adverse events CG: 0 Withdrawal from the study due to adverse events IG: 0 Withdrawal from the study due to adverse events CG: 0 Clinical response definition by the authors: same as primary outcome Number failing to achieve clinical response IG: 5 at 3 months and 9 at 12 months Number failing to achieve clinical response CG: 2 at 3 months and 9 at 12 months Before treatment IG: mean 3 (range 2–10) Before treatment CG: mean 3 (range 2–9) At 3–5 months IG: mean 3 (range 0–3) At 3–5 months CG: mean 4 (range 2–8) At 12–14 months IG: mean 2 (range 1–5) At 12–14 months CG: mean 4 (range 3–8) Endoscopic response definition by the authors: endoscopic findings scored according to the Matts grading score (Matts 1961) Number failing to achieve endoscopic response: Before treatment IG: mean 2 (range 2–4) Before treatment CG: mean 2 (range 1–4) At 3–5 months IG: mean 1 (range 1–2) At 3–5 months CG: mean 3 (range 1–4) At 12–14 months IG: mean 1 (range 1–3) At 12–14 months CG: mean 3 (range 1–4) Histological response definition by the authors: histological findings scored according to the Matts grading score (Matts 1961) Number failing to achieve histological response: Before treatment IG: mean 15 (range 13–21) Before treatment CG: mean 14 (range 11–21) At 3–5 months IG: mean 14 (range 11–18) At 3–5 months CG: mean 17 (range 11–20) At 12–14 months IG: mean 13 (range 9–21) At 12–14 months CG: mean 16 (range 9–20) Clinical relapse definition by the authors: clinical relapse was defined as the reappearance of visible blood in the stools for 2 consecutive days or the recurrence of frequent diarrhoea (> 6 bowel movements/day) (or both), nocturnal diarrhoea or abdominal cramping Endoscopic relapse definition by the authors: endoscopic relapse was defined as the reappearance of erosion or an ulcer Number who had an endoscopic relapse IG: unclear Number who had an endoscopic relapse CG: 2 participants relapsed at 3–5 months to a score of 4 and 1 more at 12–14 months HRQoL measurement tool: assessment of "general well‐being" HRQoL IG: 8/10 participants had improved well‐being HRQoL CG: n/a
Notes	Funding source: none (author's response) Conflict of interest: none (author's response) Author contact: ohkusa@med.juntendo.ac.jp
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants were randomly assigned to 1 of 2 groups and numbered sequentially. No more information given after contacting authors.
Allocation concealment (selection bias)	Low risk	The author mentioned in his response that allocation concealment was ensured via the envelope method.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Lack of placebo. No more information given after contacting authors.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Symptom assessment, endoscopic and histological evaluations were performed blind.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Data presented for all randomised participants.
Selective reporting (reporting bias)	Unclear risk	All outcomes presented, but some data were missing (mean and SDs instead of median and range). The reported results did not consistently present numbers of responders/non‐responders at 3 and 12 months and results are presented for "improved well‐being" for only 1 of the groups, which had not been properly defined in the methods section. When we contacted the authors, they provided us with the numbers of responders/non‐responders at 3 and 12 months but not the SD values.
Other bias	High risk	Despite mentioning in their paper that the included participants had active, mild to moderate UC, in our correspondence they mentioned that they defined remission as a symptom score < 3 and based on this definition 7/10 participants in the IG were in remission while only 1/10 in the CG was in remission.

Ohkusa 2010.

Study characteristics
Methods	Study design: RCT Setting: multicentre (11 centres), Japan Study period: January 2004 to July 2006
Participants	Inclusion criteria: eligibility criteria for study entry were mild to severe disease with a colonoscopy score of ≥ 1 (erythema, decreased vascular pattern, and mild friability) based on a scale of 0 (normal or inactive) to 3 (spontaneous bleeding and ulceration), or watery diarrhoea ≥ 5 times/day with visible blood in stools, or both Exclusion criteria: people with toxic megacolon or penicillin allergy, who were pregnant, or who had serious liver or renal disease or any psychological illness. People who had taken antibiotics within 4 weeks before study entry or had C diff or other stool pathogens (Salmonellae, pathogenic Escherichia coli, Campylobacter jejuni, Shigellae, etc.) at entry were also excluded. Disease activity: mild to severe Disease duration IG: mean 6.6 (SD 5.3) years Disease duration CG: mean 7.3 (SD 6.9) years after dropouts Extent of disease IG: extensive 48, left sided 45, proctitis 12 Extent of disease CG: extensive 41, left sided 57, proctitis 3 (after dropouts) Concurrent therapy IG: corticosteroid 59, sulfasalazine 27, 5‐ASA 71, immunosuppressant 6 Concurrent therapy CG: corticosteroid 52, sulfasalazine 29, 5‐ASA 71, immunosuppressant 2 (after dropouts) Number randomised to IG: 105 Number randomised to CG: 105 Number assessed in IG: 105 at 2 weeks, 81 at 3 months, 55 at 12 months. Primary result analysis was ITT, 105 participants Number assessed in CG: 101 at 2 weeks, 69 at 3 months, 39 at 12 months. Primary result analysis was ITT, 101 participants Age at randomisation IG: mean 38.5 (SD 14.1) years Age at randomisation CG: mean 39.3 (SD 13.1) years Sex (M/F) IG: 69/36 Sex (M/F) CG: 70/31 (after dropouts)
Interventions	Length of intervention: 2 weeks and follow‐up at 3 months (primary endpoint) and 12 months IG: combination therapy consisting of amoxicillin 500 mg 3 times daily, tetracycline 500 mg 3 times daily, and metronidazole 250 mg 3 times daily CG: placebo preparations consisting of identical appearing capsules and tablets containing sugar Doses of any concomitant oral or rectal medication for UC remained constant, except for corticosteroids, which were tapered after week 8 by 5 mg/week until it reached a dose of 20 mg/day. Thereafter, the dose was reduced by 2.5 mg/week until discontinuation. Corticosteroid discontinuation for > 3 months in steroid dependent UC was defined as steroid withdrawal. Steroid withdrawal rates at 6, 9, and 12 months after treatment completion were calculated.
Outcomes	Primary outcomes Definition of treatment success as defined by authors: clinical remission was defined as a total Mayo score of ≤ 2 points with no individual subscore exceeding 1 point. Clinical relapse was defined as reappearance of visible blood in stools for 2 consecutive days or recurrence of frequent diarrhoea (≥ 5 bowel movements/day), or both, nocturnal diarrhoea, or abdominal cramps. If a participant relapsed or showed exacerbation with severe or fulminant UC symptoms, study participation was stopped and the participant was treated appropriately Number failing to achieve/maintain remission IG: 85 at 3 months, 77 at 12 months Number failing to achieve/maintain remission CG: 89 at 3 months, 90 at 12 months (with the dropouts) Secondary outcomes Serious adverse events IG: 0 Serious adverse events CG: 0 Total adverse events IG: in total 55 (52.4%). Reported: nausea 18, fever 13, watery diarrhoea 8, urticaria and fever 6 Total adverse events CG: in total 16 (14.9%). No specifics reported Withdrawal from the study due to adverse events IG: 2 (unspecified) Withdrawal from the study due to adverse events CG: 0 Clinical response definition by the authors: the primary study endpoint was a clinical response documented by the Mayo score at 3 months after treatment completion; secondary endpoints were clinical and endoscopic score improvements at 12 months. For the Mayo system, clinical response was defined as a decrease from baseline in the total score of ≥ 3 points if the baseline score was > 3 points, or a decrease of ≥ 2 points if the baseline disease activity index was 3, or a decrease of ≥ 1 point if the baseline disease activity index was 1 or 2 Number failing to achieve clinical response IG: 58 at 3 months, 53 at 12 months Number failing to achieve clinical response CG: 82 at 3 months, 83 at 12 months (with dropouts) Baseline IG score: mean 6.38 Baseline CG score: mean 5.79 3 months IG score: mean 4.28 3 months CG score: mean 5.24 12 months IG score: mean 4.02 12 months CG score: mean 5.18 Endoscopic response definition by the authors: evaluated according to the Mayo system sigmoidoscopy subscore, with scores of 0–3 Number failing to achieve endoscopic response: Baseline IG score: mean 1.91 Baseline CG score: mean 1.78 3 months IG score: mean 1.33 3 months CG score: mean 1.59 12 months IG score: mean 1.29 12 months CG score: mean 1.56 Histological response definition by the authors: n/a Number failing to achieve histological response IG: n/a Number failing to achieve histological response CG: n/a Clinical relapse definition by the authors: clinical relapse was defined as reappearance of visible blood in stools for 2 consecutive days or recurrence of frequent diarrhoea (≥ 5 bowel movements/day), nocturnal diarrhoea, or abdominal cramps Endoscopic relapse definition by the authors: n/a Number who had an endoscopic relapse IG: n/a Number who had an endoscopic relapse CG: n/a HRQoL measurement tool: n/a HRQoL IG: n/a HRQoL CG: n/a
Notes	Funding source: Department of Gastroenterology, Juntendo University School of Medicine, and the Kanagawa Nanbyo Foundation. Conflict of interest: the authors declared none Author contact: ohkusa@jikei.ac.jp
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The study used central randomisation. After random allocation, 4 participants in the placebo CG did not meet the inclusion criteria (inactive UC, no mucosal inflammatory findings, and no symptoms) and were excluded because further improvement in inactive UC would be unlikely (i.e. inclusion could have biased the CG in the no‐effect direction)
Allocation concealment (selection bias)	Low risk	Treatment allocation was produced by a computer‐generated randomisation list. Randomisation was carried out using numbered containers. The blinding was not broken until the entire study was completed.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double blind: placebo preparations consisting of identical‐appearing capsules and tablets containing sugar. Participants received identical drugs in identical bottles and remained blinded throughout the study.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	All clinical evaluations and endoscopies were carried out by clinicians who were blinded to the participants' therapy.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Data presented for all randomised participants.
Selective reporting (reporting bias)	Low risk	All outcomes presented, but some data are missing (mean, SD endoscopy and histology scores instead of only the mean). After we contacted them, the authors provided us the numbers for responders/non‐responders at 3 and 12 months but not the SD values.
Other bias	Low risk	Despite randomisation, males were over‐represented in the sample but according to authors the distributions of age, sex, disease duration, disease severity, extent of disease, concomitant medication, rates of corticosteroid‐refractory and ‐dependent disease, and Mayo scores were well‐balanced between the 2 groups.

Petersen 2014.

Study characteristics
Methods	Study design: RCT Setting: Department of Gastroenterology, Hvidovre University Hospital, Hvidovre, Denmark, single‐centre Study period: n/a
Participants	Inclusion criteria: aged ≥ 18 years, with a CAI score ≥ 6. An endoscopic score was not performed as part of the study protocol; however, all participants had a faecal calprotectin performed to confirm disease activity among included participants, cut‐off 50 mg/kg Exclusion criteria: pregnant or breastfeeding women, known hypersensitivity to ciprofloxacin, people who participated in another clinical trial, positive stool sample with any enteric pathogens, parasites or C diff, current treatment with systemic corticosteroids or biologic therapy Disease activity: active Disease duration IG 1: median 4 (range 0–13) years Disease duration IG 2: median 6 (range 0–45) years Disease duration IG 3: median 4 (range 0–24) years Disease duration IG 4: median 5 (range 0–38) years Extent of disease IG 1: proctitis 3, left sided 13, pancolitis 9 Extent of disease IG 2: proctitis 4, left sided 18, pancolitis 3 Extent of disease IG 3: proctitis 3, left sided 18, pancolitis 4 Extent of disease IG 4: proctitis 4, left sided 14, pancolitis 7 Concurrent therapy IG 1: systemic 5‐ASA 19, topical 5‐ASA 17, azathioprine/6‐mercaptopurine 8, topical prednisolone 9 Concurrent therapy IG 2: systemic 5‐ASA 15, topical 5‐ASA 15, azathioprine/6‐mercaptopurine 6, topical prednisolone 6 Concurrent therapy IG 3: systemic 5‐ASA 16, topical 5‐ASA 13, azathioprine/6‐mercaptopurine 2, topical prednisolone 1 Concurrent therapy IG 4: systemic 5‐ASA 15, topical 5‐ASA 18, azathioprine/6‐mercaptopurine 6, topical prednisolone 1 Number randomised to IG 1: 25 Number randomised to IG 2: 25 Number randomised to IG 3: 25 Number randomised to IG 4: 25 Number assessed in IG 1: 25 at week 0, 25 at week 1, 22 at week 4, 20 at week 8, 19 at week 12 Number assessed in IG 2: 25 at week 0, 25 at week 1, 25 at week 4, 22 at week 8, 21 at week 12 Number assessed in IG 3: 25 at week 0, 23 at week 1, 18 at week 4, 16 at week 8, 14 at week 12 Number assessed in IG 4: 25 at week 0, 25 at week 1, 22 at week 4, 20 at week 8, 20 at week 12 Age at randomisation IG 1: 38.4 years Age at randomisation IG 2: 39.3 years Age at randomisation IG 3: 38.7 years Age at randomisation IG 4: 38.1 years Sex (M/F) IG 1: 9/16 Sex (M/F) IG 2: 9/16 Sex (M/F) IG 3: 11/14 Sex (M/F) IG 4: 9/16
Interventions	Length of intervention: 12 weeks (assessments at week 0, 1, 4, 8, and 12) IG 1: ciprofloxacin + Nissle: ciprofloxacin 500 mg twice daily for 1 week followed by EcN for 7 weeks (100 mg once daily for 4 days followed by 100 mg twice daily for the rest of the period) IG 2: ciprofloxacin + placebo: ciprofloxacin for 1 week followed by placebo for 7 weeks IG 3: placebo + Nissle: placebo for 1 week followed by EcN for 7 weeks IG 4: placebo + placebo: placebo for 1 week followed by placebo for 7 weeks Standard medical care and therapies, and relevant changes in these based on the treating physician's discretion, were allowed throughout the study; however, people requiring current treatment with systemic steroids or tumour necrosis factor‐α inhibitors were excluded from the study, topical steroids were allowed.
Outcomes	Primary outcomes Definition of treatment success as defined by authors: remission, evaluated by CAI score ≤ 4, among participants treated with ciprofloxacin for 1 week or EcN for 7 weeks (or both) as an add‐on treatment to conventional therapies Number failing to achieve/maintain remission IG 1: PP at 12 weeks 6 participants (66% of 19 achieved remission), ITT 10 participants (60% of 25 achieved remission) Number failing to achieve/maintain remission IG 2: PP at 12 weeks 5 participants (78% of 21 achieved remission), ITT 7 participants (72% of 25 achieved remission) Number failing to achieve/maintain remission IG 3: PP at 12 weeks 9 participants (54% of 19 achieved remission), ITT 15 participants (41% of 25 achieved remission) Number failing to achieve/maintain remission IG 4: PP at 12 weeks 2 participants (89% of 20 achieved remission), ITT 5 participants (80% of 25 achieved remission) Secondary outcomes Serious adverse events: 0 Total adverse events: 4 participants treated with ciprofloxacin experienced itching, nausea, or bloating and abdominal pain and 2 with placebo experienced nausea or bloating and abdominal pain. During the 7 weeks of EcN + placebo treatment, 18 participants reported adverse effects, 9 treated with EcN (itching, nausea, bloating, abdominal pain) and 9 treated with placebo (nausea, bloating, abdominal pain, sensitivity towards smells) Withdrawal from the study due to adverse events: 0 Clinical response definition by the authors: improvement in CAI scores Number failing to achieve clinical response IG 1: CAI0 10.52, CAI1 8.24, CAI4 5.74, CAI8 5.38, CAI12 4.1 Number failing to achieve clinical response IG 2: CAI0 8.88, CAI1 7.16, CAI4 4.6, CAI8 2.77, CAI12 4.14 Number failing to achieve clinical response IG 3: CAI0 9.28, CAI1 7.57, CAI4 6.78, CAI8 4.44, CAI12 5.29 Number failing to achieve clinical response IG 4: CAI0 8.88, CAI1 6.84, CAI4 5.18, CAI8 3.95, CAI12 4.9 Endoscopic response definition by the authors: n/a Number failing to achieve endoscopic response: n/a Histological response definition by the authors: n/a Number failing to achieve histological response: n/a Clinical relapse definition by the authors: n/a Endoscopic relapse definition by the authors: n/a Number who had an endoscopic relapse: n/a HRQoL measurement tool: n/a HRQoL: n/a
Notes	Funding source: Danish Colitis–Crohn's Associations Research Fund and the Danish Medical Associations Research Fund (Mimi and Victor Larsens Fond) Conflict of interest: the authors declare no conflict of interest. Author contact: munk@dadlnet.dk (AM Petersen)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants were randomly assigned, according to a randomisation carried out at the central hospital pharmacy (Region Hovedstadens Apotek, Marielundvej 25, 2730 Herlev, Denmark) in Copenhagen, to treatment groups; in ratio of 1:1:1:1, allowing 25 participants to be included in each group.
Allocation concealment (selection bias)	Unclear risk	Not mentioned. No more information given after contacting authors.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	The placebo preparations were manufactured and placebo and active treatments were packaged and distributed by the central hospital pharmacy (Region Hovedstadens Apotek, Marielundvej 25, 2730 Herlev, Denmark) securing double blinding, which was upheld until after data‐analysis.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not mentioned. No more information given after contacting authors.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Data presented for all participants at all time points.
Selective reporting (reporting bias)	Unclear risk	All outcomes reported but SDs were missing for CAI scores. Authors did not provide a response.
Other bias	Low risk	Males under‐represented in the sample but in general a well‐balanced study at baseline.

Turunen 1998.

Study characteristics
Methods	Study design: RCT Setting: Gastroenterological Unit, Department of Medicine, University of Helsinki, Finland, single‐centre Study period: April 1990 to May 1992
Participants	Inclusion criteria: histologically confirmed UC with moderate to severe endoscopic activity over ≥ 2 segments of the large bowel; aged 18–65 years; were willing to participate were considered for enrolment in the study. Women of child‐bearing potential were requested to practice reliable contraception during the treatment period Exclusion criteria: people with ulcerative proctitis only were not included in the study; positive blood cultures requiring systemic antibiotic treatment at the time of enrolment; treatment with any systemic antibiotic other than metronidazole 2 weeks before entry; any condition requiring emergency surgery, including peritonitis and fulminating colitis; severe central nervous system disorders or history of epilepsy; renal failure (limit, serum creatinine 400 mmol/L) or abnormal liver function test results (limit, liver transaminase levels of more than twice the upper normal level); diabetes or other metabolic disease; or pregnancy or lactation Disease activity: moderate to severe Disease duration IG: mean 5.0 (SEM 1.3) years Disease duration CG: mean 5.7 (SEM 0.8) years Extent of disease IG: n/a Extent of disease CG: n/a Concurrent therapy IG: steroids 14, mesalamine 16, sulfasalazine 12 Concurrent therapy CG: steroids 26, mesalamine 19, sulfasalazine 11 Number randomised to IG: 85 participants were randomised in total but it was not mentioned in which group the dropouts occurred Number randomised to CG: 85 participants were randomised in total but it was not mentioned in which group the dropouts occurred Number assessed in IG: 38 Number assessed in CG: 45 Age at randomisation IG: mean 33.9 (SEM 10.5) years Age at randomisation CG: mean 34.4 (SEM 9.9) years Sex (M/F) IG: 30/8 Sex (M/F) CG: 28/17
Interventions	Length of intervention: 6 months and 6 months' follow‐up (clinical assessment and colonoscopic evaluation were performed at 0, 3, 6, and 12 months) IG: ciprofloxacin 500–750 mg twice daily CG: placebo Therapy with ciprofloxacin or placebo was started after the initial colonoscopy at entry and continued until the 6‐month colonoscopy. The dosage of the active drug was 500 mg twice daily in participants weighing < 70 kg and 750 mg twice daily for those weighing > 70 kg Prednisone was administered to all participants at 0.75 mg/kg bodyweight for 4 weeks and was continued at 0.5 mg/kg during the next 4 weeks and then at 0.25 mg/kg for up to 12 weeks; the dose was then tapered off, if possible, meaning that in cases of partial response or recurrence of the symptoms, prednisone was reduced more slowly or increased moderately (maximally 0.5 mg/kg for 2 weeks). If disease activity was limited to the rectum, rectal administration of corticosteroids and mesalamine was used when steroids were tapered off. Hospitalisation and intravenous treatment was used at entry if necessary. Mesalamine (Asacol; Wulfing Pharma, Gronau, Germany; 800 mg twice daily) administration was started or continued at entry and used as a maintenance treatment during the entire study period
Outcomes	Primary outcomes Definition of treatment success as defined by authors: treatment failure, the primary endpoint, was defined as both symptomatic and endoscopic failure to respond. The primary endpoint of the study was treatment failure during the 6‐month treatment period. The decision regarding treatment failure was made by 2 study physicians. The opposite was treatment success Number failing to achieve/maintain remission IG: n/a Number failing to achieve/maintain remission CG: n/a Secondary outcomes Serious adverse events IG: 0 Serious adverse events CG: 0 Total adverse events IG: 2 cases of deep venous thrombosis Total adverse events CG: 1 case of deep venous thrombosis Withdrawal from the study due to adverse events IG: 0 Withdrawal from the study due to adverse events CG: 0 Clinical response definition by the authors: as in primary definition Number failing to achieve clinical response IG: 8 at 3 months and 17 at 6 months (after the 6‐month [note: possible error here and the authors may mean 3‐month as they mention 6 months later in the sentence] treatment period, 30/38 participants in the IG had responded and maintained their response). During the 6‐month follow‐up period after termination of treatment with the study drug, 9/30 participants in the ciprofloxacin group failed to respond Number failing to achieve clinical response CG: 20 at 3 months 27 at 6 months (after the 6‐month [note: possible error here and the authors may mean 3‐month as they mention 6 months later in the sentence] treatment period, 25/45 participants in the placebo group had responded and maintained their response). During the 6‐month follow‐up period after termination of treatment with the study drug 7/25 in the placebo group failed to respond Endoscopic response definition by the authors: treatment failure was defined as a colonoscopic finding of moderate to severe activity in ≥ 2 segments of the colon after failure to respond to this regimen. Colonoscopy included assessment of the presence or absence of erythema, friability, small and large ulcerations or mucopurulent exudate, and visibility of the vascular pattern. These were assessed in the caecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum. A semiquantitative score of 0–114 was then computed from the severity of the inflammatory changes in each segment. In addition, a global assessment of severity was made by the endoscopist (0, normal; 1, quiescent; 2, mild; 3, moderate; and 4, severe). Number failing to achieve endoscopic response IG: 7 at 3 months (5/36 in the ciprofloxacin group had mild, moderate, or severe inflammation) were responders Number failing to achieve endoscopic response CG: 23 at 3 months (17/39 in the placebo group had mild, moderate, or severe inflammation) were responders. Histological response definition by the authors: the histological score was assessed from 16 biopsy specimens according to the area with the most severe inflammation as follows: 0, normal or only fibrosis or scarring; 1, mild with mononuclear cell infiltration; 2, moderate with neutrophils present in crypts or epithelium; and 3, severe inflammation with neutrophil infiltration and crypt abscesses. Number failing to achieve histological response IG: 4 at 3 months (2/36 participants treated with ciprofloxacin had a histological score of moderate or severe inflammation) Number failing to achieve histological response CG: 23 at 3 months (15/37 participants treated with placebo had a histological score of moderate or severe inflammation) Clinical relapse definition by the authors: n/a Endoscopic relapse definition by the authors: n/a Number who had an endoscopic relapse IG: n/a Number who had an endoscopic relapse CG: n/a HRQoL measurement tool: n/a HRQoL IG: n/a HRQoL CG: n/a
Notes	Funding source: supported by Department of Research and Development, Northern Europe, Bayer AG, Gothenburg, Sweden Conflict of interest: not mentioned Author contact: Ulla Turunen, MD; martti.farkkila@hus.fi
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was performed with a computer program integrated in the database system by the biased coin method (weight, 0.6) to ensure approximately equivalent participant groups with respect to bodyweight (≥ 70 vs < 70 kg) and extent of UC (total or extensive vs left sided).
Allocation concealment (selection bias)	Low risk	Author responded that allocation concealment was ensured with the use of sealed envelopes and the person who handled the sealed envelopes had no further connection to the study.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	The study was double blind, and records of the codes were kept at the Bayer AG office in Helsinki, Finland. 3 physicians were responsible for the clinical part of the study. Efforts were made for successful blinding also by keeping the microbiological results separate from patient files and out of the hospital computer system.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Author responded that personnel providing participant care who assessed the data were blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	All data reported for the 83 assessed participants, but it is not mentioned where the 2 dropouts occurred.
Selective reporting (reporting bias)	Low risk	All reported but the definitions for responses/failures were complicated and not all endpoints were reported clearly.
Other bias	Unclear risk	Well‐balanced in all but males were over‐represented despite randomisation.

5‐ASA: 5‐aminosalicylic acid; C diff (Clostridioides difficile, previously known as Clostridium difficile); CG: control group; CRP: C‐reactive protein; EcN: Escherichia coli Nissle; F: female; GI: gastrointestinal; HRQoL: health related quality of life; IG: intervention group; IQR: interquartile range; ITT: intention to treat; M: male; n/a: not available; PP: per protocol; RCT: randomised controlled trial; SD: standard deviation; SEM: standard error of the mean; UCDAI: Ulcerative Colitis Disease Activity Index.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Abdallah 1970	Wrong design – not RCT
Aoki 2012	Wrong intervention
Borok 1996	Wrong study type – not RCT
Breton 2019	Wrong study design – not RCT
Casellas 1998	Wrong study design – no data for induction of remission > 14 days
Chapman 1986	Wrong study design – no data for induction of remission > 14 days
Dickinson 1985	Wrong participants and intervention – they did not receive 14 days of antibiotics, only received 7 days of treatment
El‐Nachef 2020	Wrong intervention
Gardner 1981	Wrong study design – not RCT
Gionchetti 1999	Wrong study design – no data for induction of remission > 14 days
Guslandi 2006	Wrong study design – not RCT
Henn 2021	Wrong intervention (antibiotic only given as pretreatment)
Hillaire 2007	Wrong participants – not ulcerative colitis
Inoue 2007	Wrong participants – not ulcerative colitis
Ishikawa 2017	Wrong design – not RCT
Järnerot 1989	Wrong study – not ulcerative colitis
Jena 2020	Wrong intervention (antifungal)
Jigaranu 2014	Wrong participants – not ulcerative colitis
JPRN‐UMIN000025846	Trial registration for wrong study design – not RCT
Kato 2014	Wrong study design – not RCT
Koido 2014	Wrong population – not for induction or maintenance
Kordy 2018	Wrong design – not RCT
Kuenstner 2007	Wrong design – not RCT
Kumar 2010	Wrong design – not RCT
Lee 2018	Wrong population – not ulcerative colitis
Leiper 2000	Wrong population – not ulcerative colitis
Leiper 2008	Wrong design – not RCT
Levine 2016	Wrong population – not ulcerative colitis
Levine 2019	Wrong population – not ulcerative colitis
Lobo 1993	Wrong design – not RCT
Minami 2009	Wrong population – not for induction and maintenance
NCT04898348	Wrong intervention
Nomura 2005	Wrong population – not for induction and maintenance
Okahara 2020	Wrong design – not RCT
Sato 2009	Wrong population – not for induction and maintenance
Schoonees 1996	Wrong design – not RCT
Selby 2007	Wrong population – not ulcerative colitis
Terao 2011	Wrong design – not RCT
Turner 2019	Not induction trial – no data for induction of remission > 14 days
Turunen 1995	Wrong population – not ulcerative colitis
Turunen 1999	Wrong design – not RCT
Uehara 2010	Wrong design – not RCT
Wu 2013	Not induction trial – no data for induction of remission > 14 days
Yukawa 2013	Wrong population – not induction or maintenance trial

RCT: randomised controlled trial.

Characteristics of studies awaiting classification [ordered by study ID]

JPRN‐UMIN000009811.

Methods	RCT
Participants	40
Interventions	Group 1: 2‐week combination therapy consisting of amoxicillin (Sawacillin) 500 mg 3 times daily, tetracycline (Acromycin) 500 mg 3 times daily, and metronidazole (Flagyl) 250 mg 3 times daily with an inhibitor of gastric acid secretion. Participants are to be followed for 3 months (first endpoint). Group 2: 2‐week combination therapy consisting of amoxicillin (Sawacillin) 500 mg 3 times daily, fosfomycin (Fosmicin) 1000 mg 3 times daily, and metronidazole (Flagyl) 250 mg 3 times daily, with an inhibitor of gastric acid secretion. Participants are to be followed for 3 months.
Outcomes	Primary outcome Clinical remission and clinical response Secondary outcome Treatment failure rate
Notes	Date of first enrolment: 23 January 2013, completion date: 30 August 2018

JPRN‐UMIN000017844.

Methods	RCT
Participants	20
Interventions	Amoxicillin, tetracycline, and metronidazole
Outcomes	Clinical remission in 3 months
Notes	Date of first enrolment: 22 June 2015, completion date: none

JPRN‐UMIN000018157.

Methods	RCT
Participants	20
Interventions	Oral antibiotics
Outcomes	Improvement rate after 3 months
Notes	Completion date: 31 December 2019

JPRN‐UMIN000024520.

Methods	RCT
Participants	212
Interventions	Group 1: placebo Group 2: metronidazole Group 3: metronidazole + amoxicillin Group 4: metronidazole + amoxicillin + tetracycline
Outcomes	Primary outcomes Response rate after 12 weeks (Mayo score) when the Mayo score decreases > 30% and, > 3 decreases; bleeding subscore from the rectum is > 1 decrease or < 1 Secondary outcomes Remission rate after 12 weeks (Mayo score) Response rate and remission rate after 12 weeks (Mayo score) with refractory/"inrefractory" (understood to mean non‐refractory) Endoscopic remission rate after 12 weeks (mucosal healing)
Notes	Date of first enrolment: 15 December 2016, completion date: none

Mishra 2021.

Methods	RCT
Participants	Group 1: 25 participants Group 2: 25 participants
Interventions	Group 1: intravenous ceftriaxone Group 2: intravenous placebo
Outcomes	Primary outcome Response on day 3 according to the Oxford criteria (Travis 1996) Secondary outcomes Changes in partial Mayo score; CRP levels, faecal calprotectin (day 3), and need for second line therapy, hospital stay, and mortality (day 28)
Notes	This RCT will be included for analysis in the next review update.

NCT01772615.

Methods	RCT, quadruple blinded (participant, care provider, investigator, outcomes assessor)
Participants	100
Interventions	Group 1: ciprofloxacin + Escherichia coli Nissle Group 2: ciprofloxacin + placebo Group 3: placebo + Escherichia coli Nissle Group 4: placebo + placebo
Outcomes	Primary outcomes Number of participants in remission at 12 weeks; number of participants in remission (CAI ≤ 4); time for participants to achieve remission (CAI ≤ 4) Secondary outcome Numbers of participants who completed the study at 12 weeks; numbers of participants who completed the study; IBDQ score; dietary intake; adverse effects of treatment Other outcomes Escherichia coli strains at 12 weeks; changes in Escherichia coli strains
Notes	Study completion date: August 2013, perhaps linked to Petersen 2014.

NCT02606032.

Methods	RCT, quadruple blinded (participant, care provider, investigator, outcomes assessor)
Participants	80
Interventions	Group 1: metronidazole 500 mg twice daily + doxycycline 100 mg twice daily + terbinafine 250 mg once daily all for 14 days Group 2: placebo + placebo + placebo all twice daily for 14 days
Outcomes	Primary outcomes Remission of ulcerative colitis defined as a Mayo score < 3 with an endoscopic Mayo score = 0 at the end of the trial; transplant therapy in ulcerative colitis at 9 weeks
Notes	Estimated completion date: 30 April 2020

Turner 2020.

Methods	RCT
Participants	Group 1: 16 children Group 2: 12 children
Interventions	Group 1: antibiotics + IVCS (amoxicillin, vancomycin, metronidazole, doxycycline/ciprofloxacin) Group 2: IVCS (amoxicillin, vancomycin, metronidazole, doxycycline/ciprofloxacin)
Outcomes	Primary outcome Disease activity (Pediatric Ulcerative Colitis Activity Index) at day 5. Microbiome was analysed using 16S rRNA gene and metagenome.
Notes	This RCT will be included for analysis in the next review update.

CAI: Clinical Activity Index; CG: control group; CRP: C‐reactive protein; IBDQ: Inflammatory Bowel Disease Questionnaire; IG: intervention group; RCT: randomised controlled trial.

Characteristics of ongoing studies [ordered by study ID]

NCT03986996.

Study name	Antimicrobial therapy for ulcerative colitis (UC)
Methods	RCT, investigator‐blinded, 2 groups with 1:1 randomisation
Participants	40
Interventions	Group 1: amoxicillin + metronidazole + tetracycline twice daily for 2 weeks Group 2: amoxicillin + doxycycline twice daily for 2 weeks
Outcomes	Primary outcome Efficacy – clinical response at week 3; response defined as a 3 point drop in SCCAI/20 point drop in PUCAI or drop in < 3/20 point but entering clinical remission, defined as a SCCAI score < 5/PUCAI score < 10 Secondary outcomes Efficacy – clinical response at week 6; mean SCCAI/PUCAI; efficacy – remission at week 6; SCCAI score < 5/PUCAI score < 10; efficacy – Physicians Global Assessment at week 6
Starting date	Estimated study start date: July 2019 Estimated completion date: June 2022
Contact information	Michal Yaakov, PhD; MICHALY@wmc.gov.il
Notes

PUCAI: Pediatric Ulcerative Colitis Activity Index; RCT: randomised controlled trial; SCCAI: Simple Clinical Colitis Activity Index.

Differences between protocol and review

Within our protocol (Gordon 2020), we did not include the proportion of participants who had a clinical relapse (as defined by the original studies), as a secondary outcome. We have now included it as our fourth secondary outcome.

If enough data were available, we planned to perform subgroup analyses of the primary outcomes. The following variables were selected:

1. Age of participants (children versus adults).

2. Different antibiotic dose.

3. Duration of the intervention.

4. Sex (male versus female).

5. Route of administration (oral versus intravenous).

Not enough data were available, so no subgroup analyses were performed.

Following advice, we decided that if one treatment arm reported adverse events, while the other arm reported zero events, that we would not calculate absolute risk based on 0 events, but rather 0.5 events. This was because zero events were not considered realistic. Since the value 0.5 cannot be used within Review Manager 5 analyses, this method was only employed in our summary of findings tables and does not appear in the forest plots and corresponding analyses, where adverse events have been left as zero.

This approach was not deemed necessary when both intervention arms report zero events. In these instances, we did not make any changes.

We have inserted footnotes into the summary of finding tables to explain this approach. We have also inserted a statement within the methods section.

Contributions of authors

MG: initiated and conceptualised the review; secured funding; performed screening of titles and abstracts and full‐text articles, data extraction and contacted authors, analysed and interpreted data; checked quality assessment; checked the quality of statistical analysis; contributed to writing and editing the review; made an intellectual contribution to the review; approved the final review prior to submission.

VS: performed screening of titles and abstracts and full‐text articles, extracted data and contacted authors; checked the quality of data extraction; analysed and interpreted data; undertook and checked quality assessment; performed statistical analysis; checked the quality of the statistical analysis; contributed to writing and editing the review; made an intellectual contribution to the review; approved the final review prior to submission.

CGC: performed screening of titles and abstracts and full‐text articles, extracted data and contacted authors; checked the quality of data extraction; analysed and interpreted data; undertook and checked quality assessment; performed statistical analysis; checked the quality of the statistical analysis; contributed to writing and editing the review; made an intellectual contribution to the review; approved the final review prior to submission.

AKA: initiated and conceptualised the review; contributed in the resolution of disagreements and offered general guidance; approved the final review prior to submission.

Sources of support

Internal sources

• NIHR grant, UK

  This review was supported by an NIHR Evidence Synthesis grant

External sources

• None, Other

  None

Declarations of interest

MG: since February 2018, I have received travel fees to attend international scientific and training meetings from Pharma companies. These grants included no honoraria, inducement, advisory role, or any other relationship and were restricted to the travel and meeting related costs of attending such meetings.

These include: DDW May 2018; Advances in IBD December 2018; DDW May 2019

The companies include: Biogaia (2018 to 2019), Ferring (2018), synergy (bankrupt in 2018) and Tillots (2018 to 2019).

None of these companies have had any involvement in any works completed by me and I have never had any payments for any other activities for them, as confirmed below.

From these date onwards, I have made a personal undertaking to take no further funds from any pharmaceutical or formula company in any form for travel or other related activities.

This is to lift the limitations such funding has on my ability to act as a first and corresponding author on reviews, in line with the Cochrane policies on such matters and is reported in line with these policies.

These current declarations will expire over the next three years and this statement updated regularly to reflect this.

VS: none.

CGC: none.

AKA: none.

New