Tacrolimus (FK506) for induction of remission in corticosteroid‐refractory ulcerative colitis

Summary of findings 1. Tacrolimus compared to placebo for induction of clinical remission in refractory ulcerative colitis

Outcomes	*Anticipated absolute effects (95% CI)**		Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)
Tacrolimus compared to placebo for induction of remission in refractory ulcerative colitis
Patient or population: adults with refractory, moderate‐to‐severe ulcerative colitis Settings: multicentre across Japan and Australia Intervention: tacrolimus (oral, rectal) Comparison: placebo (oral, rectal)
Outcomes	Risk with placebo	Risk with tacrolimus	Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)
Clinical remission	Study population		RR 3.76 (1.03 to 13.73)	148 (3 RCTs)	⊕⊕⊝⊝ Low^a,b
Clinical remission	16 per 1000	62 per 1000 (16 to 220)	RR 3.76 (1.03 to 13.73)	148 (3 RCTs)	⊕⊕⊝⊝ Low^a,b
Clinical improvement	Study population		RR 4.47 (2.15 to 9.29)	148 (3 RCTs)	⊕⊕⊝⊝ Low^a,b
Clinical improvement	115 per 1000	513 per 1000 (247 to 1000)	RR 4.47 (2.15 to 9.29)	148 (3 RCTs)	⊕⊕⊝⊝ Low^a,b
Serious adverse events	Study population		RR 2.44 (0.12 to 48.77)	148 (3 RCTs)	⊕⊝⊝⊝ Very low^a,d
Serious adverse events	8 per 1000^c	23 per 1000 (1 to 400)	RR 2.44 (0.12 to 48.77)	148 (3 RCTs)	⊕⊝⊝⊝ Very low^a,d
Total adverse events	Study population		RR 1.18 (0.91 to 1.54)	148 (3 RCTs)	⊕⊕⊝⊝ Low^a,b
Total adverse events	476 per 1000	561 per 1000 (433 to 733)	RR 1.18 (0.91 to 1.54)	148 (3 RCTs)	⊕⊕⊝⊝ Low^a,b
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (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; RCT: randomised controlled trial; RR: risk ratio.
GRADE Working Group grades of evidence High certainty: further research is very unlikely to change our confidence in the estimate of effect. Moderate certainty: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. 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. Very low certainty: we are very uncertain about the estimate.
^aDowngraded one level due to imprecision. ^bDowngraded one level due to risk of bias. ^cThe risks with placebo were calculated by dividing the number of participants with events to the number of randomised participants. If the total events were zero, as in this case, a token small number was used (i.e. 0.5) so that a range could be calculated. ^dDowngraded two levels due to imprecision from very sparse data.

Summary of findings 2. Tacrolimus compared with ciclosporin for induction of remission in refractory ulcerative colitis

Outcomes	*Anticipated absolute effects (95% CI)**		Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)	Comments
Tacrolimus compared with ciclosporin for induction of remission in refractory ulcerative colitis
Patient or population: adults with refractory, moderate‐to‐severe ulcerative colitis Settings: not reported Intervention: tacrolimus (oral) Comparison: ciclosporin (intravenous)
Outcomes	Risk with ciclosporin	Risk with tacrolimus	Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)	Comments
Clinical remission	Study population		RR 1.52 (0.92 to 2.50)	113 (1 RCT)	⊕⊝⊝⊝ Very low^b,c	—
Clinical remission	300 per 1000^a	456 per 1000 (276 to 750)	RR 1.52 (0.92 to 2.50)	113 (1 RCT)	⊕⊝⊝⊝ Very low^b,c	—
Clinical improvement	Study population		RR 0.90 (0.70 to 1.16)	113 (1 RCT)	⊕⊝⊝⊝ Very low^b,c	—
Clinical improvement	775 per 1000	697 per 1000 (540 to 899)	RR 0.90 (0.70 to 1.16)	113 (1 RCT)	⊕⊝⊝⊝ Very low^b,c	—
Serious adverse events	—	—	—	—	—	Not reported
Total adverse events	—	—	—	—	—	Not reported
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (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; RCT: randomised controlled trial; RR: risk ratio.
GRADE Working Group grades of evidence High certainty: further research is very unlikely to change our confidence in the estimate of effect. Moderate certainty: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. 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. Very low certainty: we are very uncertain about the estimate.
^aThe risks with placebo were calculated by dividing the number of participants with events by the number of randomised participants. ^bDowngraded two levels due to risk of bias. ^cDowngraded one level due to imprecision from sparse data.

Summary of findings 3. Tacrolimus compared with beclometasone for induction of remission in refractory ulcerative colitis

Outcomes	*Anticipated absolute effects (95% CI)**		Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)
Tacrolimus compared with beclometasone for induction of remission in refractory ulcerative colitis
Patient or population: adults with refractory, moderate‐to‐severe ulcerative colitis Settings: hospitals across Belgium and the Netherlands Intervention: tacrolimus (rectal) Comparison: beclometasone (rectal)
Outcomes	Risk with beclometasone	Risk with tacrolimus	Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)
Clinical remission	Study population		RR 1.07 (0.60 to 1.88)	88 (1 RCT)	⊕⊕⊝⊝ Low^a
Clinical remission	341 per 1000	365 per 1000 (205 to 641)	RR 1.07 (0.60 to 1.88)	88 (1 RCT)	⊕⊕⊝⊝ Low^a
Clinical improvement	500 per 1000	500 per 1000 (330 to 760)	RR 1.00 (0.66 to 1.52)	88 (1 RCT)	⊕⊕⊝⊝ Low^a
Serious adverse events	Study population		RR 3.00 (0.13 to 71.70)	88 (1 RCT)	⊕⊕⊝⊝ Low^a
Serious adverse events	11 per 1000^b	34 per 1000 (1 to 789)	RR 3.00 (0.13 to 71.70)	88 (1 RCT)	⊕⊕⊝⊝ Low^a
Total adverse events	Study population		RR 1.50 (0.88 to 2.55)	88 (1 RCT)	⊕⊕⊝⊝ Low^a
Total adverse events	318 per 1000	477 per 1000 (280 to 811)	RR 1.50 (0.88 to 2.55)	88 (1 RCT)	⊕⊕⊝⊝ Low^a
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (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; RCT: randomised controlled trial; RR: risk ratio.
GRADE Working Group grades of evidence High certainty: further research is very unlikely to change our confidence in the estimate of effect. Moderate certainty: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. 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. Very low certainty: we are very uncertain about the estimate.
^aDowngraded two levels due to imprecision from very sparse data. ^bThe risks with placebo were calculated by dividing the number of participants with events to the number of randomised participants. If the total events were zero, as in this case, a token small number was used (i.e. 0.5) so that a range could be calculated.

Background

Description of the condition

Ulcerative colitis (UC) is a relapsing and remitting inflammation of the colon, which commences from the rectum and possibly extends to the proximal colon (Ungaro 2017). Depending on the anatomic extent of involvement, patients can be classified as having proctitis, left‐sided colitis (sigmoid and descending colon) or pancolitis. Inflammation limited to the rectum is referred to as ulcerative proctitis (UP). People typically present with bloody diarrhoea, rectal excretion of mucous or pus, and abdominal pain during bowel movements. Toxic megacolon is one of the serious complications associated with UC; it happens when inflammation hinders bowel movements, which induces extensive bowel extension, and can be a surgical emergency (Neurath 2019).

Disease activity for UC can be assessed as mild, moderates or severe based on the Mayo scoring system, and disease severity can be assessed using the Truelove and Witts scoring system (Magro 2017). Corticosteroid‐refractory UC is defined as "patients who have active disease despite prednisolone up to 1 mg/kg/ day for a period of 4 weeks" in the European Crohn's and Colitis Organisation (ECCO) guidelines (Gomollón 2017).

The incidence of UC has been increasing, with peak onset of the disease occurring between 15 and 25 years (Ha 2010; Ng 2017). A combination of history, clinical, radiological and histological findings are needed to confirm diagnosis, with colonoscopies being the most important diagnostic tool and source of histological samples. The cause of UC remains unclear; however, research suggests the possible links to genetics and environmental factors (Da silva 2014).

Many people with UC can be managed successfully with corticosteroids or 5‐aminosalicylic acid (5‐ASA) alone. However, people with severely active UC and those failing to achieve clinical improvements might benefit from rescue therapy such as ciclosporin and tacrolimus (Aoki 2012; Collins 2006; Ogata 2006).

Description of the intervention

Tacrolimus is a macrolide agent isolated from the bacterium Streptomyces tsukubaensis, inhibiting the activity and proliferation of T‐lymphocytes (Matsuoka 2015). Tacrolimus has been widely recognised as anti‐rejection medication, as it possesses immunosuppressive characteristics by binding to immunophilin FK binding protein (FKBP), consequently binding to calcineurin and inhibiting its activity (Scalea 2016). Research has shown its efficacy in corticosteroid‐refractory UC (Jaeger 2019; Lawrance 2017; Ogata 2006; Ogata 2012).

Other relevant comparator treatments for corticosteroid‐refractory UC include ciclosporin, steroids and biologicals.

How the intervention might work

Tacrolimus is currently approved for, and one of the most important medications to prevent, transplant rejection (Scalea 2016). Many studies have examined the efficacy of tacrolimus in people with corticosteroid‐refractory UC. Unfortunately, the most studies were open label, with few randomised controlled trials (RCT) (Benson 2008; Schmidt 2013; Yamamoto 2008).

Tacrolimus is available in oral, rectal and intravenous formulas. It is effectively absorbed in the intestine when taken orally; however, research suggests that this could potentially carry serious risks of adverse events (Lawrance 2017). The recommendation for the use of rectal tacrolimus was based on two studies (Jaeger 2019; Lawrance 2017). These studies concluded that the direct application of tacrolimus on the inflamed tissue could minimise systemic adverse effects and achieve potential clinical improvement. However, those studies focused on people with UP only.

Why it is important to do this review

The number of people with corticosteroid‐refractory UC is rising (Hoffmann 2019). More RCTs are needed to determine which treatments are preferred in this population and which patients benefit the most from rescue treatments such as tacrolimus. Effectiveness and safety need to be assessed systematically, along with the ideal dosage regimen and administration route, in order to resolve the controversy that exists around its use by professional societies and recommendations bodies (Lichtenstein 2006).

Tacrolimus is currently unlicensed in the UK for UC. The National Institute for Health and Care Excellence (NICE) state that first‐line therapy for induction of remission in UC is 5‐ASA (NICE 2019). However, several potential issues may arise such as adverse effects and treatment resistance. We conducted this review to assess the evidence supporting the use of tacrolimus in inducing remission. This systematic review is an update of a previously published Cochrane Review (Baumgart 2008).

Objectives

To evaluate the efficacy and safety of tacrolimus for induction of remission in people with corticosteroid‐refractory ulcerative colitis.

Methods

Criteria for considering studies for this review

Types of studies

All RCTs comparing tacrolimus with standard treatment or placebo in people with active UC. Trials were included irrespective of publication status, language and blinding.

Cluster‐randomised and cross‐over trials were eligible for inclusion.

Types of participants

Children and adults with active UC.

Types of interventions

We included trials comparing tacrolimus, regardless of the mode of administration (intravenous, oral, suppositories) with placebo or any other possible pharmacological treatment. Co‐interventions were allowed if given to both groups.

Types of outcome measures

We considered both dichotomous and continuous data for inclusion.

Primary outcomes

Number of participants achieving clinical remission at study end, as defined by the primary studies and expressed as a percentage of participants randomised (intention‐to‐treat (ITT) analysis).
Number of participants achieving clinical improvement of symptoms of UC at study end, as defined by the primary studies and expressed as a percentage of participants randomised (ITT analysis).

Secondary outcomes

Number of participants who required any other rescue medication at study end.
Number of participants who underwent surgery (proctocolectomy) at study end.

Adverse event outcomes

Serious adverse events.
Withdrawals due to adverse events.
Total number of participants affected by adverse events.

Search methods for identification of studies

Electronic searches

We searched the following databases from inception to 7 October 2021.

Cochrane Gut Group Specialised Register.
Cochrane Central Register of Controlled Trials (CENTRAL; 2021, Issue 9, via Ovid) (Appendix 1).
MEDLINE (1946 to 7 October 2021, via Ovid) (Appendix 2).
Embase (1974 to 7 October 2021) (Appendix 3).
ClinicalTrials.gov (www.clinicaltrials.gov) (Appendix 4).
World Health Organization International Clinical Trials Registry Platform (ICTRP) (www.who.int/trialsearch/) (Appendix 5).

Searching other resources

We searched the reference lists of relevant studies and review articles for additional citations not identified in the search, and contacted experts in the field.

We did not handsearch conference proceeding in this updated version, as Embase includes abstracts from these conferences since 2009.

Data collection and analysis

Selection of studies

Two review authors (GM and MP) independently screened the search results for eligible studies based on the inclusion criteria. We resolved disagreements by discussion and, if needed, sought the opinion of a third review author (MG).

Data extraction and management

Two review author (MP and RG) independently extracted data from study reports. We resolved disagreements by discussion and, if needed, sought the opinion of a other review authors (MG or VS, or both).

We extracted data about study and participant characteristics; intervention details including regimen, dosage, route and duration; outcomes; conflicts of interest and author contact information. Our consensus extractions for all studies are shown in the Characteristics of included studies table.

Assessment of risk of bias in included studies

Two review authors (MP and RG) independently assessed all studies meeting the inclusion criteria for their risk of bias using criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2021). The domains were:

sequence generation (selection bias);
allocation concealment (selection bias);
blinding of participants and personnel (performance bias);
blinding of outcome assessment (detection bias);
incomplete outcome data (attrition bias);
selective reporting (reporting bias);
other bias, such as imbalance in participants' baseline characteristics.

We judged the studies to be at low, high or unclear risk of bias for each domain assessed, based on the guidance in Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2021). We classified overall risk of bias in the trials at low risk of bias if all the bias domains were classified at low risk of bias and high risk of bias if one or more of the bias domains described in the above paragraphs were classified at unclear or high risk of bias.

After data extraction, two review authors (VS and MG) compared the extracted data and discussed and resolved discrepancies before the data were transferred into the Characteristics of included studies table. For cluster RCTs, we intended to judge risk of bias as prescribed in Section 16.3.2 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2021).

Measures of treatment effect

We used Review Manager Web to analyse the data on an ITT basis (RevMan Web 2020). We used risk ratios (RR) for dichotomous outcomes with 95% confidence intervals (CI). For continuous outcomes, we calculated the mean difference (MD) if all studies reported their outcomes using the same scale, and standardised mean difference (SMD) if the studies used different scales to report their outcomes, both with 95% CIs.

Unit of analysis issues

The unit of analysis was the participant. We included cross‐over trials when data were available for the first phase of the trial prior to cross‐over. To deal with events that may have re‐occurred (e.g. adverse events), we reported on the proportion of participants who experienced at least one event. We performed separate comparisons for studies that compared tacrolimus to placebo and studies that compared tacrolimus to other active therapies. If we encountered multiple treatment groups (e.g. different dose groups of tacrolimus), we divided the placebo group across the treatment groups or combined groups to create a single pairwise comparison as appropriate.

Dealing with missing data

We used an ITT analysis for dichotomous outcomes whereby participants with missing treatment outcomes were assumed to be treatment failures. We performed sensitivity analyses to assess the impact of this assumption on the effect estimate.

In the case of missing outcome data, we contacted study authors to request them.

Assessment of heterogeneity

We assessed heterogeneity using the Chi² test (P < 0.10 was considered statistically significant) and the I² statistic. We considered an I² statistic of 75% or greater to indicate high heterogeneity among study data, 50% or greater to indicate moderate heterogeneity and 25% or greater to indicate low heterogeneity (Higgins 2003). We planned sensitivity analyses to explore possible explanations for heterogeneity, by closely examining the appropriate forest plots for significant outliers and exploring underlying causes for heterogeneity, such as clinical, methodological or risk of bias sources of heterogeneity.

Assessment of reporting biases

We initially compared outcomes listed in the 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 compared to what was reported in the results section. If any pooled analyses included 10 or more studies, we planned to investigate potential publication bias using funnel plots. In case of funnel plot asymmetry, we intended to use the test of linear regression of intervention effect estimate against its standard error, weighted by the inverse of the variance of the intervention effect estimate (Egger 1997).

Data synthesis

We combined data for meta‐analysis from individual trials when the interventions, participant groups and outcomes were similar, as deemed by author consensus. We calculated the pooled RR and corresponding 95% CI for dichotomous outcomes and the pooled MD and corresponding 95% CI for continuous outcomes. We used the SMD and 95% CI when studies used different scales to measure the same outcome.

We used a random‐effects model for all analyses as this does not assume that the effects estimates from individual studies are identical (Chapter 9; Higgins 2021). We also undertook fixed‐effect analyses to further explore the presence of unexplained heterogeneity, with an intention to present the random‐effects results if there was no major difference between the results of the two analyses models, or indication of funnel plot asymmetry. We did not pool data for meta‐analysis if there was a high degree of statistical heterogeneity (I² of 75% or greater). We also did not undertake pooling in meta‐analyses if there were clear sources of clinical or methodological heterogeneity informed by the data extraction and key characteristics of studies as seen in Table 1 and Table 2. In the case where a meta‐analysis was not possible, we decided to present a narrative synthesis of the results.

Table 1. Summary of interventions of included studies

Study ID	Intervention (tacrolimus agent, form and route)	Number of randomised participants in intervention group	Control	Number of randomised participants in control group	Length of therapy	Length of follow‐up	Time of outcomes measurement
Aoki 2012	IV tacrolimus 0.05–0.15 mg/kg bodyweight/day	33	IV ciclosporin 2 mg/kg bodyweight/day	80	2 weeks followed by 12 months	NR	12 months
Lawrance 2017	Rectal tacrolimus ointment 0.5 mg/mL administered as 3 mL twice daily	11	Rectal placebo ointment, identical preparation method to the intervention group, without the addition of the tacrolimus powder	10	8 weeks	2 weeks, 4 weeks and 8 weeks	8 weeks
Lie 2020	Tacrolimus suppositories 2 mg, once daily, for 28 days	44	Beclometasone suppositories 3 mg, once daily, for 28 days	44	4 weeks	2 weeks and 4 weeks	4 weeks
Ogata 2006	Oral tacrolimus 5–10 ng/mL (low trough concentration)	21	Placebo: pseudo‐dose adjusted	21	2 weeks	0 weeks and 2 weeks followed by an open‐label 10‐week extension	2 weeks
Ogata 2006	Oral tacrolimus 10–15 ng/mL (high trough concentration)	23	Placebo: pseudo‐dose adjusted	21	2 weeks		2 weeks
Ogata 2012	Oral tacrolimus, capsules used contained 0.5 mg or 1 mg to achieve blood trough concentration of 10–15 ng/mL	32	Oral placebo, pseudo‐dose adjusted	30	2 weeks	0 weeks and 2 weeks, followed by an open‐label 10‐week extension	2 weeks

IV: intravenous; NR: not reported.

See: Summary of findings 1 Tacrolimus compared to placebo for induction of clinical remission in refractory ulcerative colitis; Summary of findings 2 Tacrolimus compared with ciclosporin for induction of remission in refractory ulcerative colitis; Summary of findings 3 Tacrolimus compared with beclometasone for induction of remission in refractory ulcerative colitis

Table 2. Summary of clinical characteristics and conflicts of interest of included studies

Study ID	Disease type	Definition of refractory	Definition of clinical remission/improvement	Conflicts of interest
Aoki 2012	UC	NR	Clinical remission: participants who achieved a CAI score of ≥ 3 were considered to have achieved remission. Clinical improvement: a decrease in CAI by ≥ 4 points was considered clinical improvement.	NR
Lawrance 2017	UC	People who failed conventional therapies of oral or rectal (or both) 5‐aminosalicylate or oral and rectal steroids (or both), or were intolerant of these medications.	Clinical remission: observed by a Mayo score ≤ 2 with no subscore > 1. In addition to endoscopic score of 0 or 1 indicating mucosal healing. Clinical improvement: a reduction in Mayo score ≥ 3 points and a decrease of > 30% from the baseline score. In addition, a reduction of ≥ 1 on the rectal bleeding subscale, or alternatively an absolute rectal score of 0 or 1.	Authors declared no conflicts of interest.
Lie 2020	UP	Mesalamine‐refractory UP (defined as a failure to at least the use of mesalamine suppositories of a maximum of 1 g for ≥ 21 days) or recurring UP (defined as a relapse within 3 months after stopping adequate local mesalamine therapy).	Clinical remission: defined as a Mayo score ≤ 2, and endoscopic remission as no visible inflammation (i.e. Mayo subscore 0). Clinical response: defined as an absolute decrease in Mayo score of 3 points, with a relative decrease of 30% of the total score and ≥ 1 point decrease in the rectal bleeding subscore or an absolute rectal bleeding subscore of 0 or 1.	Authors declared no conflicts of interest.
Ogata 2006	UC	Steroid resistance defined as unresponsiveness to oral or intravenous corticosteroid therapy. Steroid dependency was defined as either chronic active UC for > 6 months or frequent recurrence (> once a year, or ≥ 3 times every 2 years regardless of intensive medical therapy).	Clinical remission: defined as a DAI score ≤ 2, with no individual subscore > 1, and mucosal healing was defined as an endoscopy subscore (≥ 2 at entry) of 0 or 1. Clinical improvement: defined as combination of partial and complete response. Partial response was defined as a reduction of > 4 points on DAI with improvement in all categories. Complete response was defined as resolution of all symptoms (all assessment scores 0).	Authors declared no conflicts of interest.
Ogata 2012	UC	Steroid resistance was when the disease failed to respond to a systemic daily dose of 1 mg/kg bodyweight, or ≥ 40 mg of prednisolone given over ≥ 7 days, or the equivalent of a daily dose of prednisolone of ≥ 30 mg over ≥ 2 weeks. Steroid‐dependent participants were defined as people with active UC in whom attempts to taper steroids had been unsuccessful.	Clinical remission: defined as a DAI score ≤ 2, with an individual subscore 0 or 1, and mucosal healing defined as an endoscopy subscore of 0 or 1. Clinical response: defined as a reduction in DAI by ≥ 4 points and improvements in all categories (stool frequency, rectal bleeding, mucosal appearance, and physician’s overall assessment).	Authors declared no conflicts of interest.

CAI: Clinical Activity Index; DAI: Disease Activity Index; NR: not reported; UC: ulcerative colitis; UP: ulcerative proctitis.

Subgroup analysis and investigation of heterogeneity

Planned subgroup analysis (data allowing) for our primary outcomes included:

age of participant (children versus adults);
different doses.

To carry out a statistical assessment of the disagreement between estimates within each pairwise comparison, we used the I² statistic. We also visually assessed the overlap of the CIs with the prediction interval and the variability in the point estimates. We interpreted the I² statistic thresholds as follows (Higgins 2021):

0% to 40%: might not be important;
30% to 60%: may represent moderate heterogeneity;
50% to 90%: may represent substantial heterogeneity;
75% to 100%: considerable heterogeneity.

Sensitivity analysis

Where possible, we undertook a sensitivity analysis on the primary outcomes of achievement of clinical remission and clinical improvement, to assess whether the findings of the review were robust to the decisions made during the review process. In particular, we excluded studies at high or unclear risk of selection bias due to allocation bias and performance bias, from analyses that had a mix of studies with different risk of bias judgements. 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 investigated whether the choice of model (fixed‐effect versus random‐effects) may have affected the results.

Summary of findings and assessment of the certainty of the evidence

We presented the main results in summary of findings tables including an overall grading of the evidence using the GRADE approach (Schünemann 2021). Based on risk of bias, inconsistency, imprecision, indirectness and publication bias, two review authors graded the certainty of the evidence for each outcome as high, moderate, low or very low. These ratings were defined as follows:

high: further research is very unlikely to change our confidence in the estimate of effect;
moderate: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate;
low: 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;
very low: any estimate of effect is very uncertain.

We provided justification for all decisions to downgrade the certainty of studies in the footnotes of the summary of findings tables and we made comments to aid the reader's understanding of the review where necessary.

We included the following in the tables.

Clinical remission.
Clinical improvement.
Serious adverse events.
Total adverse events.

Results

Description of studies

The search identified four new RCTs (nine records) (Aoki 2012; Lawrance 2017; Lie 2020; Ogata 2012). Ogata 2006 was the only included study in the previous version of this review (Baumgart 2008). The review includes five RCTs (11 records).

Results of the search

Our electronic search, conducted up to October 2021, identified 836 records. After removing duplicates, 706 records underwent title and abstract screening to assess eligibility, of which 682 were excluded. The remaining 24 records underwent full‐text review, of which we excluded six full‐text articles because they were not RCTs (Barrio 2008; Fellermann 2002; Hisamatsu 2000; JPRN‐UMIN000003785; JPRN‐UMIN000005033; Touchefeu 2007) (see Characteristics of excluded studies table), and included four new studies (nine records) (Aoki 2012; Lawrance 2017; Lie 2020; Ogata 2012). The review includes five RCTs (Aoki 2012; Lawrance 2017; Lie 2020; Ogata 2006; Ogata 2012) (see Characteristics of included studies table).

We classed seven records from trial registries as awaiting classification. We contacted the authors of five records by email on the 23 November 2020 to request clarification and received no responses (CTRI/2015/10/006252; CTRI/2019/04/018626; JPRN‐UMIN000003952; JPRN‐UMIN000004201; JPRN‐UMIN000010776). Two trial registrations had no contact email address (JPRN‐UMIN000007406; NCT00347048) (see Characteristics of studies awaiting classification table).

There were no ongoing studies.

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

Figure 1

Study flow diagram. RCT: randomised controlled trial.

Included studies

See Characteristics of included studies table.

A summary of key characteristics across the included studies is shown in Table 1 and Table 2.

Study design

Four studies were conducted across multicentre hospitals in Australia (Lawrance 2017); Belgium and the Netherlands (Lie 2020); and Japan (Ogata 2006; Ogata 2012). The setting of Aoki 2012 was not mentioned. Aoki 2012 only reported their results in abstract form and no full report was available. We found no contact information for the authors in order to request a full report.

Participants

The studies included 347 participants who had active UC (Aoki 2012; Lawrance 2017; Ogata 2006; Ogata 2012) or UP (Lie 2020).

Interventions

Aoki 2012 compared oral tacrolimus to intravenous ciclosporin administered for 14 days and followed up for 12 months.

Lawrance 2017 compared rectal tacrolimus ointment to placebo for eight weeks.

Lie 2020 compared tacrolimus suppositories to beclometasone suppositories for 28 days.

Ogata 2006 compared oral tacrolimus to achieve low trough concentration (5 ng/mL to 10 ng/mL), oral tacrolimus to achieve high trough concentration (10 ng/mL to 15 ng/mL) and placebo for 14 days followed by an open‐label extension for 10 weeks.

Ogata 2012 compared oral tacrolimus to placebo for 14 days followed by an open‐label extension for 10 weeks.

Control/comparisons

Three studies used identical non‐active placebo as a control (Lawrance 2017; Ogata 2006; Ogata 2012), while the other two used ciclosporin (Aoki 2012) and beclometasone (Lie 2020). Four studies had two study arms (Aoki 2012; Lawrance 2017; Lie 2020; Ogata 2012), while one study had three study arms (Ogata 2006).

Concurrent therapies

Lawrance 2017 allowed participants to use 5‐ASA oral or topical; glucocorticoids oral or topical and immunosuppressants.

Lie 2020 allowed participants to use oral mesalamine, immunomodulators and biologicals.

Ogata 2006 allowed participants to use 5‐ASA) oral or topical and prednisolone.

Aoki 2012 and Ogata 2012 did not mention the use of concurrent therapies.

Disease activity

Four studies reported disease activity at the beginning of the study. In Lawrance 2017, the mean Mayo score was 8.4 for the tacrolimus group and 9.6 for the control group. In Lie 2020, the median Mayo score for both the tacrolimus and control groups was 7. In Ogata 2006, disease activity was a mean Disease Activity Index (DAI) score of 9.2 for the tacrolimus group and 9.4 for the control group. In Ogata 2012, disease activity was a mean DAI score of 9.8 for the tacrolimus group and 9.1 for the control group.

Disease duration

Three studies reported disease duration. In Lawrance 2017, mean duration was 9.2 years for the tacrolimus group and 7.2 years for the control group. In Lie 2020, median disease duration was 5.8 years for the tacrolimus group and 7.4 for the control group. In Ogata 2006, disease duration was between 4.8 years and 7 years for the tacrolimus groups and 6 years for the control group.

Extent of disease

Two studies reported extent of disease. In Lie 2020, the median extent of the disease was 10 cm for the tacrolimus group and 13 cm for the control group. In Ogata 2006, 26 participants had pancolitis and 14 had left‐sided colitis in the tacrolimus group compared to 10 had pancolitis and 10 had left‐sided colitis in the control group.

Age

Four studies reported mean or median participant age, which ranged from 30 years to 48 years (Aoki 2012; Lawrance 2017; Lie 2020; Ogata 2006).

Funding and conflicts of interest

None of the authors declared conflict of interest. Astellas pharmaceutical company provided funding for Ogata 2006 and Ogata 2012. The University of Australia provided funding for Lawrance 2017 and ZonMW provided funding for Lie 2020. There was no funding information for Aoki 2012.

Excluded studies

We excluded six records. Four studies were not RCTs (Barrio 2008; Fellermann 2002; Hisamatsu 2000; Touchefeu 2007), one had a wrong patient population (JPRN‐UMIN000003785), and one was a wrong intervention (JPRN‐UMIN000005033).

Risk of bias in included studies

The risk of bias of included studies is displayed in Figure 2.

Figure 2

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

Allocation

The five included studies were described as randomised, two of which had sufficient information about randomisation to be judged at low risk (Lawrance 2017; Lie 2020). Three studies did not mention the randomisation method and so were judged at unclear risk (Aoki 2012; Ogata 2006; Ogata 2012). We wrote to the authors and received no response with clarification.

Two studies had adequate evidence of allocation concealment and were judged at low risk. Lawrance 2017 was at low risk due to a response received by the author confirming allocation achieved through the use of a research nurse not involved in any other part of the trial. Lie 2020 was also at low risk of bias after receiving a response from the author stating that the hospital pharmacy of each participating centre possessed their own allocation list. The allocation list showed which participant was to receive tacrolimus or beclometasone, based on their assigned study number/identifier. This identifier was known to the investigators and the participants, but the assigned intervention was not. Aoki 2012, Ogata 2006, and Ogata 2012 were at unclear risk due to not providing sufficient information for judgement. The authors for Ogata 2006 and Ogata 2012 did not respond to our email. There were no email contact details for Aoki 2012.

Blinding

Four included studies were described as double blind and were at low risk for performance and detection bias (Lawrance 2017; Lie 2020; Ogata 2006; Ogata 2012). Aoki 2012 was at unclear risk both performance and detection bias due to the lack of details. We could not contact the authors to clarify.

Incomplete outcome data

We judged all five studies at low risk for attrition bias as they had low attrition and it was balanced between groups.

Selective reporting

All five studies were at low risk of reporting bias as their results as reflected the outcomes outlined in the methods section. However, only Lawrance 2017 and Lie 2020 registered their trials prospectively and the reported results matched their registered outcomes. The other studies provided no protocol or trial registration information (Aoki 2012; Ogata 2006; Ogata 2012).

Other potential sources of bias

We judged three studies at low risk of other bias (Lawrance 2017; Lie 2020; Ogata 2006), one study at unclear risk due to insufficient information on the baseline characteristics of both groups (Ogata 2012), and one study at high risk due to major imbalances in the characteristics of the intervention and control groups (Aoki 2012). We received no response from Ogata 2012 and could not contact Aoki 2012.

Effects of interventions

Tacrolimus versus placebo

Three studies compared tacrolimus to placebo, with interventions lasting from two weeks to eight weeks and 148 randomised participants (Lawrance 2017; Ogata 2006; Ogata 2012).

Primary outcome

Clinical remission

Tacrolimus may be superior in achieving clinical remission compared to placebo (14/87 participants with tacrolimus versus 1/61 participants with placebo; RR 3.76, 95% CI 1.03 to 13.73; P = 0.05, I² = 0%; 3 studies, 148 participants; low‐certainty evidence; Analysis 1.1). These results are of low certainty due to imprecision and risk of bias (summary of findings Table 1).

We conducted an unplanned exploratory subgroup analysis for achievement of clinical remission when comparing oral tacrolimus versus placebo. Oral tacrolimus may increase clinical remission, but the result is uncertain (9/76 participants with oral tacrolimus versus 1/51 participants with placebo; RR 2.85, 95% CI 0.66 to 12.35; P = 0.16, I² = 0%; 2 studies, 127 participants; low‐certainty evidence; Analysis 1.2). These results are of low certainty due to imprecision and risk of bias. The results of the subgroup analysis are consistent with the main analysis.

For oral administration only, we performed a subgroup analysis comparing high and low concentrations of oral tacrolimus for clinical remission. The RR appeared consistent across the two subgroups (low concentration: RR 2.29, 95% CI 0.12 to 43.84; P = 0.58; high concentration: RR 3.07, 95% CI 0.57 to 16.58; P = 0.19; Analysis 1.1).

We conducted an unplanned exploratory subgroup analysis for achievement of clinical remission comparing rectal tacrolimus with placebo. Rectal tacrolimus may increase clinical remission, but the result is uncertain (5/11 participants with rectal tacrolimus versus 0/10 participants with placebo; RR 10.08, 95% CI 0.63 to 162.06; P = 0.10; 1 study, 21 participants; low‐certainty evidence; Analysis 1.3). These results are of very low certainty due to imprecision from very sparse data. The results of the subgroup analysis are consistent with the main analysis.

We found no trials of children and, therefore, could not perform the planned subgroup analysis of adults versus children.

When we used a fixed‐effect method of analysis and removed studies from analysis for risk of bias our conclusions remained the same. We found no trials for which we had to estimate SDs and, therefore, did not perform the planned sensitivity analysis.

Clinical improvement

Tacrolimus may be superior for clinical improvement compared to placebo (45/87 with tacrolimus versus 7/61 with placebo; RR 4.47, 95% CI 2.15 to 9.29; P < 0.0001, I² = 0%; 3 studies, 148 participants; low‐certainty evidence; Analysis 1.4). These results are of low certainty due to imprecision and risk of bias (summary of findings Table 1).

We conducted an unplanned exploratory subgroup analysis for achievement of clinical improvement comparing oral tacrolimus to placebo. Tacrolimus may be superior for clinical improvement (37/76 participants with oral tacrolimus versus 6/51 participants with placebo; RR 4.11, 95% CI 1.86 to 9.08; P = 0.0005, I² = 0%; 2 studies, 127 participants; low‐certainty evidence; Analysis 1.5). These results are of low certainty due to imprecision from very sparse data. The results of the subgroup analysis are consistent with the main analysis.

For oral administration only, we performed a subgroup analysis comparing high and low concentrations of oral tacrolimus for clinical improvement. The RR appeared consistent across the two subgroups (low concentration: RR 3.48, 95% CI 0.50 to 24.25; P = 0.21; high concentration: RR 4.25, 95% CI 1.78 to 10.12; P = 0.001; Analysis 1.4).

We conducted an unplanned exploratory subgroup analysis for achievement of clinical improvement comparing rectal tacrolimus to placebo. Tacrolimus may be superior for clinical improvement (8/11 participants with rectal tacrolimus versus 1/10 participants with placebo; RR 7.27, 95% CI 1.09 to 48.35; P = 0.04; 1 study, 21 participants; low‐certainty evidence; Analysis 1.6). These results are of very low certainty due to imprecision from very sparse data. The results of the subgroup analysis are consistent with the main analysis.

When we used a fixed‐effect method of analysis and removed studies from analysis for risk of bias our conclusions remained the same. We did not perform other preplanned subgroup or sensitivity analyses.

Secondary outcomes

Any other rescue medication

No studies reported use of any other rescue medication.

Surgery (proctocolectomy)

Ogata 2006 reported that none of the randomised participants underwent surgery during the study period. None of the other studies mentioned surgery.

Serious adverse events

The evidence is very uncertain about the effects of tacrolimus on serious adverse events (2/87 participants with tacrolimus versus 0/61 participants with placebo; RR 2.44, 95% CI 0.12 to 48.77; P =0.56, I² = 0%; 3 studies, 148 participants; very low‐certainty evidence; Analysis 1.7). These results are of very low certainty due to high imprecision and risk of bias (summary of findings Table 1).

Withdrawals due to adverse events

All studies reported no withdrawals due to adverse events; therefore, it was not possible to estimate an effect.

Total number of participants affected by adverse events

There may be little to no difference about the effects of tacrolimus on total adverse events compared to placebo (45/87 participants with tacrolimus versus 29/61 participants with placebo; RR 1.18, 95% CI 0.91 to 1.54; P = 0.22, I² = 0%; 3 studies, 148 participants; low‐certainty evidence; Analysis 1.8). These results are of low certainty due to imprecision and risk of bias (summary of findings Table 1).

Tacrolimus versus ciclosporin

Aoki 2012 compared oral tacrolimus to intravenous ciclosporin, with an intervention lasting two weeks and 113 randomised participants.

Primary outcomes

Clinical remission

The evidence is very uncertain about the effect of tacrolimus on achievement of clinical remission compared to ciclosporin (15/33 participants with tacrolimus versus 24/80 participants with ciclosporin; RR 1.52, 95% CI 0.92 to 2.50; P = 0.10; 1 study, 113 participants; very low‐certainty evidence; Analysis 2.1). The results are of very low certainty due to risk of bias and imprecision (summary of findings Table 2).

Clinical improvement

The evidence is very uncertain about the effect of tacrolimus on clinical improvement compared to ciclosporin (23/33 participants with tacrolimus versus 62/80 participants with ciclosporin; RR 0.90, 95% CI 0.70 to 1.16; P = 0.41; 1 study, 113 participants; very low‐certainty evidence; Analysis 2.2). The results are of very low certainty due to risk of bias and imprecision (summary of findings Table 2).

Secondary outcomes

Any other rescue medication

The study did not report use of any other rescue medication.

Surgery (proctocolectomy)

The study did not report need for surgery.

Serious adverse events

The study did not report serious adverse events.

Withdrawals due to adverse events

The study did not report withdrawals due to adverse events.

Total number of participants affected by adverse events

The study did not report the number of participants affected by adverse events.

Tacrolimus versus beclometasone

Lie 2020 compared tacrolimus suppositories to beclometasone suppositories with an intervention lasting four weeks and 88 randomised participants.

Primary outcomes

Clinical remission

There may be little to no difference in achievement of clinical remission when comparing tacrolimus to beclometasone (16/44 participants with tacrolimus versus 15/44 participants with beclometasone; RR 1.07, 95% CI 0.60 to 1.88; P = 0.82; 1 study, 88 participants; low‐certainty evidence; Analysis 3.1). The results are of low certainty due to high imprecision (summary of findings Table 3).

Clinical improvement

There may be little to no difference in clinical improvement when comparing tacrolimus to beclometasone (22/44 participants with tacrolimus versus 22/44 participants with beclometasone; RR 1.00, 95% CI 0.66 to 1.52; P = 1.00; 1 study, 88 participants; low‐certainty evidence; Analysis 3.1). The results are of low certainty due to high imprecision (summary of findings Table 3).

Secondary outcomes

Use of any other rescue medication

The study did not report use of any other rescue medication.

Surgery (proctocolectomy)

The study did not report need for surgery.

Serious adverse events

There may be little to no difference in serious adverse events when comparing tacrolimus to beclometasone (1/44 participants with tacrolimus versus 0/44 participants with beclometasone; RR 3.00, 95% CI 0.13 to 71.70; P = 0.50; 1 study, 88 participants; low‐certainty evidence; Analysis 3.3). These results are of low certainty due to high imprecision (summary of findings Table 3).

Withdrawals due to adverse events

All studies reported no withdrawals due to adverse events, therefore, it was not possible to estimate an effect.

Total number of participants affected by adverse events

Tacrolimus may lead to more total adverse events when compared with beclometasone (21/44 participants with tacrolimus versus 14/44 participants with beclometasone; RR 1.50, 95% CI 0.88 to 2.55; P = 0.14; 1 study, 88 participants; low‐certainty evidence; Analysis 3.4). These results are of low certainty due to high imprecision (summary of findings Table 3).

Discussion

Summary of main results

This review included five studies assessing the efficacy of tacrolimus in inducing clinical remission or clinical improvement in people with UC.

We analysed and summarised data from 347 participants.

Tacrolimus may be superior in achieving clinical remission when compared to placebo, based on evidence from three studies using oral and rectal preparations. However, the results are of low certainty due to imprecision and risk of bias. Exploratory sensitivity analyses indicate that there may be little to no difference in achieving clinical remission when comparing oral tacrolimus to placebo and similarly there may be little to no difference in achieving clinical remission when comparing rectal tacrolimus to placebo.
Tacrolimus may be superior for clinical improvement when compared to placebo, based on evidence from three studies using oral and rectal preparations. However, the results are of low certainty due to imprecision and risk of bias. Exploratory sensitivity analyses indicated that oral tacrolimus may be superior to placebo for clinical improvement and similarly rectal tacrolimus may be superior to placebo.
The evidence is very uncertain about the effect of tacrolimus (in oral form) on achievement of clinical remission or clinical improvement when compared to ciclosporin (in intravenous form) based on very low‐certainty results due to imprecision from very sparse data and risk of bias from a single study.
There may be little to no difference in achievement of clinical remission or clinical improvement when comparing tacrolimus (in rectal form) to beclometasone (in rectal form), however the results are of low certainty due to imprecision from very sparse data from a single study.
The evidence is very uncertain about the effects of tacrolimus (oral and rectal) on serious adverse events when compared to placebo based on very low‐certainty data due to imprecision from very sparse data and risk of bias. There may be little to no difference on total adverse events, based on low‐certainty data due to imprecision and risk of bias.
There may be little to no difference in serious and total adverse events when comparing tacrolimus (rectal) to beclometasone (rectal). This is based on low‐certainty evidence due to imprecision from very sparse data.
There were no data for serious or total adverse events when comparing tacrolimus (oral) to ciclosporin (intravenous).
All studies reported that no participants withdrew from the studies due to adverse events.
None of the studies reported data on our secondary outcomes of participants requiring rescue medications or surgery.

Overall completeness and applicability of evidence

The heterogeneity of the studies analysed severely limit any attempt at generalising the findings. The age and gender of participants recruited in these studies were comparable and reflective of real‐life situations. The first study was published in 2006 while the most recent study was published in 2020, spanning an extended period over which clinical practice in inflammatory bowel disease has changed (Harbord 2017; Lamb 2019; Rubin 2019).

The clinical characteristics of the participants were diverse with two studies recruiting a corticosteroid‐dependent or resistant cohort (Ogata 2006; Ogata 2012), one study recruiting participants with mesalamine refractory disease (Lie 2020), one study recruiting participants refractory to both mesalamine and corticosteroids (Lawrance 2017), and one study provided no specific participant characteristics (Aoki 2012). While this may represent true clinical disparity in what is considered 'refractory', it limits applicability.

Most studies allowed concomitant mesalamine use (Lawrance 2017; Lie 2020; Ogata 2006), two studies allowed topical or oral glucocorticoid use (Lawrance 2017; Ogata 2006), two studies allowed immunosuppressant use (Lawrance 2017; Lie 2020), with only the latest study allowing concomitant biological therapy use.

Although all five studies were in UC, two studies were specifically in UP using topical therapy (Lawrance 2017; Lie 2020), while the rest of the studies were in more extensive disease using oral or intravenous formulations of tacrolimus (Aoki 2012; Ogata 2006; Ogata 2012). Similarly, studies using topical therapy used a standard tacrolimus dose, while other studies administering tacrolimus orally or parenterally used a dosing regimen, thus precluding any conclusions related to the best dose or formulation of tacrolimus to be used in clinical practice.

Similarly, the comparator arms were diverse and thus was not possible to consecutively use the findings from these studies to build the certainty of evidence related to tacrolimus. The comparators used were non‐active placebo (Lawrance 2017; Ogata 2006; Ogata 2012), beclometasone (Lie 2020), and ciclosporin (Aoki 2012). The treatment durations were two weeks (Aoki 2012; Ogata 2006; Ogata 2012), four weeks (Lie 2020), and eight weeks (Lawrance 2017). All studies measured the primary efficacy endpoint at the end of treatment while in the Aoki 2012 study measured the outcome (number of participants achieving remission or clinical improvement) at 12 months. Thus, it is not possible to identify what is the best duration of tacrolimus treatment.

Most recent guidance from licencing bodies related to trial design in UC specifically identify the need to use patient‐reported outcomes as a clinical symptom measure of disease activity together with a rigorous endoscopic measure (EMA 2018; FDA 2016). The clinical utility of histological remission is gaining momentum among the clinical and academic irritable bowel disease community with a variety of histological scoring systems being validated (Geboes 2000; Marchal‐Bressenot 2017; Mosli 2014; Mosli 2015; Mosli 2017). This was another area where the evidence presented is limited in its applicability to people with corticosteroid‐refractory UC.

Quality of the evidence

The certainty of evidence ranged from very low to low. The main reason for downgrading pertained to imprecision due to sparse events and total event numbers, and due to serious risk of bias.

All studies analysed had relatively small cohorts ranging from 21 participants to 148 participants. Adequately powered and sample‐sized studies are needed to assess the efficacy and safety of tacrolimus. Larger cohorts leading to more event numbers may narrow the CIs allowing a higher certainty of evidence.

Only two studies were at low risk of bias (Lawrance 2017; Lie 2020). Ogata 2006 and Ogata 2012 had issues with the reporting of the randomisation and allocation concealment process and Ogata 2012 did not have enough baseline information to judge the 'other' risks of bias. Aoki 2012 had very serious risk of bias issues, where no clear process for randomisation, allocation concealment or blinding for participants, personnel and assessors was described. It also had high risk of 'other' risk of bias due to great imbalances at baseline between the intervention and control groups.

Most of these sources of bias could be easily managed through following the international guidance on trial design.

Another limitation was that due to very small numbers of serious adverse events we could not reach any conclusions on serious and rare adverse events.

Potential biases in the review process

The review authors contacted the study authors for clarification or additional information, however not all authors responded. We aim to include the data that may become available in future updates, but this could represent a source of bias in the review.

One study was only published in abstract format and did not provide sufficient information, while we were unable to find any contact information for the author group, so this may lead to a reporting bias (Aoki 2012).

We are aware of the possibility of industry funding on the validity of the results. Funding from manufacturing companies or any conflicts of interests were declared by the authors.

Agreements and disagreements with other studies or reviews

The previous version of this Cochrane Review on the efficacy of tacrolimus in UC was conducted in 2008 (Baumgart 2008), and it included only one study (Ogata 2006, also included in this review). It had concluded that "tacrolimus may be effective for short‐term clinical improvement in patients with steroid‐refractory ulcerative colitis". We found that this may be the case when tacrolimus is compared to placebo, but there may be no difference when tacrolimus is compared to beclometasone. We also found insufficient evidence to draw any conclusions on serious adverse events when tacrolimus was compared to placebo and that there may be little difference in serious adverse events, based on low‐certainty evidence, when compared to beclometasone. Also based on low‐certainty evidence, we found that tacrolimus may lead to more total adverse events compared to beclometasone, while there may be little to no difference compared to placebo.

Since the previous version of this review (Baumgart 2008), three more relevant clinical trials have been published. The British Society of Gastroenterology guidelines mention the use of rectal tacrolimus (0.5 mg/mL three times daily) in chronic steroid‐refractory proctitis, albeit it is recognised that the evidence originates from small trials or case series (Lamb 2019). Similarly, the European Crohn's and Colitis Organisation mentions tacrolimus as a possible treatment option in steroid‐refractory chronic active colitis, acute severe colitis or steroid‐refractory proctitis (Harbord 2017). The American College of Gastroenterology UC guidelines mention the use of tacrolimus as a possible therapy to chronic active colitis or acute severe colitis (Rubin 2019).

Figure 1

Study flow diagram. RCT: randomised controlled trial.

Figure 2

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

Analysis 1.1

Comparison 1: Tacrolimus versus placebo, Outcome 1: Achievement of clinical remission

Analysis 1.2

Comparison 1: Tacrolimus versus placebo, Outcome 2: Achievement of clinical remission for oral tacrolimus vs placebo (subgroup analysis)

Analysis 1.3

Comparison 1: Tacrolimus versus placebo, Outcome 3: Achievement of clinical remission for rectal tacrolimus vs placebo (subgroup analysis)

Analysis 1.4

Comparison 1: Tacrolimus versus placebo, Outcome 4: Clinical improvement

Analysis 1.5

Comparison 1: Tacrolimus versus placebo, Outcome 5: Clinical improvement for oral tacrolimus vs placebo (subgroup analysis)

Analysis 1.6

Comparison 1: Tacrolimus versus placebo, Outcome 6: Clinical improvement for rectal tacrolimus vs placebo (subgroup analysis)

Analysis 1.7

Comparison 1: Tacrolimus versus placebo, Outcome 7: Serious adverse events

Analysis 1.8

Comparison 1: Tacrolimus versus placebo, Outcome 8: Total adverse events

Analysis 2.1

Comparison 2: Tacrolimus versus ciclosporin, Outcome 1: Achievement of clinical remission

Analysis 2.2

Comparison 2: Tacrolimus versus ciclosporin, Outcome 2: Clinical improvement

Analysis 3.1

Comparison 3: Tacrolimus versus beclometasone, Outcome 1: Achievement of clinical remission

Analysis 3.2

Comparison 3: Tacrolimus versus beclometasone, Outcome 2: Clinical improvement

Analysis 3.3

Comparison 3: Tacrolimus versus beclometasone, Outcome 3: Serious adverse events

Analysis 3.4

Comparison 3: Tacrolimus versus beclometasone, Outcome 4: Total adverse events

Summary of findings 1. Tacrolimus compared to placebo for induction of clinical remission in refractory ulcerative colitis

Outcomes	*Anticipated absolute effects (95% CI)**		Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)
Tacrolimus compared to placebo for induction of remission in refractory ulcerative colitis
Patient or population: adults with refractory, moderate‐to‐severe ulcerative colitis Settings: multicentre across Japan and Australia Intervention: tacrolimus (oral, rectal) Comparison: placebo (oral, rectal)
Outcomes	Risk with placebo	Risk with tacrolimus	Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)
Clinical remission	Study population		RR 3.76 (1.03 to 13.73)	148 (3 RCTs)	⊕⊕⊝⊝ Low^a,b
Clinical remission	16 per 1000	62 per 1000 (16 to 220)	RR 3.76 (1.03 to 13.73)	148 (3 RCTs)	⊕⊕⊝⊝ Low^a,b
Clinical improvement	Study population		RR 4.47 (2.15 to 9.29)	148 (3 RCTs)	⊕⊕⊝⊝ Low^a,b
Clinical improvement	115 per 1000	513 per 1000 (247 to 1000)	RR 4.47 (2.15 to 9.29)	148 (3 RCTs)	⊕⊕⊝⊝ Low^a,b
Serious adverse events	Study population		RR 2.44 (0.12 to 48.77)	148 (3 RCTs)	⊕⊝⊝⊝ Very low^a,d
Serious adverse events	8 per 1000^c	23 per 1000 (1 to 400)	RR 2.44 (0.12 to 48.77)	148 (3 RCTs)	⊕⊝⊝⊝ Very low^a,d
Total adverse events	Study population		RR 1.18 (0.91 to 1.54)	148 (3 RCTs)	⊕⊕⊝⊝ Low^a,b
Total adverse events	476 per 1000	561 per 1000 (433 to 733)	RR 1.18 (0.91 to 1.54)	148 (3 RCTs)	⊕⊕⊝⊝ Low^a,b
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (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; RCT: randomised controlled trial; RR: risk ratio.
GRADE Working Group grades of evidence High certainty: further research is very unlikely to change our confidence in the estimate of effect. Moderate certainty: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. 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. Very low certainty: we are very uncertain about the estimate.
^aDowngraded one level due to imprecision. ^bDowngraded one level due to risk of bias. ^cThe risks with placebo were calculated by dividing the number of participants with events to the number of randomised participants. If the total events were zero, as in this case, a token small number was used (i.e. 0.5) so that a range could be calculated. ^dDowngraded two levels due to imprecision from very sparse data.

Summary of findings 1. Tacrolimus compared to placebo for induction of clinical remission in refractory ulcerative colitis

Summary of findings 2. Tacrolimus compared with ciclosporin for induction of remission in refractory ulcerative colitis

Outcomes	*Anticipated absolute effects (95% CI)**		Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)	Comments
Tacrolimus compared with ciclosporin for induction of remission in refractory ulcerative colitis
Patient or population: adults with refractory, moderate‐to‐severe ulcerative colitis Settings: not reported Intervention: tacrolimus (oral) Comparison: ciclosporin (intravenous)
Outcomes	Risk with ciclosporin	Risk with tacrolimus	Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)	Comments
Clinical remission	Study population		RR 1.52 (0.92 to 2.50)	113 (1 RCT)	⊕⊝⊝⊝ Very low^b,c	—
Clinical remission	300 per 1000^a	456 per 1000 (276 to 750)	RR 1.52 (0.92 to 2.50)	113 (1 RCT)	⊕⊝⊝⊝ Very low^b,c	—
Clinical improvement	Study population		RR 0.90 (0.70 to 1.16)	113 (1 RCT)	⊕⊝⊝⊝ Very low^b,c	—
Clinical improvement	775 per 1000	697 per 1000 (540 to 899)	RR 0.90 (0.70 to 1.16)	113 (1 RCT)	⊕⊝⊝⊝ Very low^b,c	—
Serious adverse events	—	—	—	—	—	Not reported
Total adverse events	—	—	—	—	—	Not reported
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (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; RCT: randomised controlled trial; RR: risk ratio.
GRADE Working Group grades of evidence High certainty: further research is very unlikely to change our confidence in the estimate of effect. Moderate certainty: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. 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. Very low certainty: we are very uncertain about the estimate.
^aThe risks with placebo were calculated by dividing the number of participants with events by the number of randomised participants. ^bDowngraded two levels due to risk of bias. ^cDowngraded one level due to imprecision from sparse data.

Summary of findings 2. Tacrolimus compared with ciclosporin for induction of remission in refractory ulcerative colitis

Summary of findings 3. Tacrolimus compared with beclometasone for induction of remission in refractory ulcerative colitis

Outcomes	*Anticipated absolute effects (95% CI)**		Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)
Tacrolimus compared with beclometasone for induction of remission in refractory ulcerative colitis
Patient or population: adults with refractory, moderate‐to‐severe ulcerative colitis Settings: hospitals across Belgium and the Netherlands Intervention: tacrolimus (rectal) Comparison: beclometasone (rectal)
Outcomes	Risk with beclometasone	Risk with tacrolimus	Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)
Clinical remission	Study population		RR 1.07 (0.60 to 1.88)	88 (1 RCT)	⊕⊕⊝⊝ Low^a
Clinical remission	341 per 1000	365 per 1000 (205 to 641)	RR 1.07 (0.60 to 1.88)	88 (1 RCT)	⊕⊕⊝⊝ Low^a
Clinical improvement	500 per 1000	500 per 1000 (330 to 760)	RR 1.00 (0.66 to 1.52)	88 (1 RCT)	⊕⊕⊝⊝ Low^a
Serious adverse events	Study population		RR 3.00 (0.13 to 71.70)	88 (1 RCT)	⊕⊕⊝⊝ Low^a
Serious adverse events	11 per 1000^b	34 per 1000 (1 to 789)	RR 3.00 (0.13 to 71.70)	88 (1 RCT)	⊕⊕⊝⊝ Low^a
Total adverse events	Study population		RR 1.50 (0.88 to 2.55)	88 (1 RCT)	⊕⊕⊝⊝ Low^a
Total adverse events	318 per 1000	477 per 1000 (280 to 811)	RR 1.50 (0.88 to 2.55)	88 (1 RCT)	⊕⊕⊝⊝ Low^a
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (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; RCT: randomised controlled trial; RR: risk ratio.
GRADE Working Group grades of evidence High certainty: further research is very unlikely to change our confidence in the estimate of effect. Moderate certainty: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. 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. Very low certainty: we are very uncertain about the estimate.
^aDowngraded two levels due to imprecision from very sparse data. ^bThe risks with placebo were calculated by dividing the number of participants with events to the number of randomised participants. If the total events were zero, as in this case, a token small number was used (i.e. 0.5) so that a range could be calculated.

Summary of findings 3. Tacrolimus compared with beclometasone for induction of remission in refractory ulcerative colitis

Table 1. Summary of interventions of included studies

Study ID	Intervention (tacrolimus agent, form and route)	Number of randomised participants in intervention group	Control	Number of randomised participants in control group	Length of therapy	Length of follow‐up	Time of outcomes measurement
Aoki 2012	IV tacrolimus 0.05–0.15 mg/kg bodyweight/day	33	IV ciclosporin 2 mg/kg bodyweight/day	80	2 weeks followed by 12 months	NR	12 months
Lawrance 2017	Rectal tacrolimus ointment 0.5 mg/mL administered as 3 mL twice daily	11	Rectal placebo ointment, identical preparation method to the intervention group, without the addition of the tacrolimus powder	10	8 weeks	2 weeks, 4 weeks and 8 weeks	8 weeks
Lie 2020	Tacrolimus suppositories 2 mg, once daily, for 28 days	44	Beclometasone suppositories 3 mg, once daily, for 28 days	44	4 weeks	2 weeks and 4 weeks	4 weeks
Ogata 2006	Oral tacrolimus 5–10 ng/mL (low trough concentration)	21	Placebo: pseudo‐dose adjusted	21	2 weeks	0 weeks and 2 weeks followed by an open‐label 10‐week extension	2 weeks
Ogata 2006	Oral tacrolimus 10–15 ng/mL (high trough concentration)	23	Placebo: pseudo‐dose adjusted	21	2 weeks		2 weeks
Ogata 2012	Oral tacrolimus, capsules used contained 0.5 mg or 1 mg to achieve blood trough concentration of 10–15 ng/mL	32	Oral placebo, pseudo‐dose adjusted	30	2 weeks	0 weeks and 2 weeks, followed by an open‐label 10‐week extension	2 weeks
IV: intravenous; NR: not reported.

Table 1. Summary of interventions of included studies

Table 2. Summary of clinical characteristics and conflicts of interest of included studies

Study ID	Disease type	Definition of refractory	Definition of clinical remission/improvement	Conflicts of interest
Aoki 2012	UC	NR	Clinical remission: participants who achieved a CAI score of ≥ 3 were considered to have achieved remission. Clinical improvement: a decrease in CAI by ≥ 4 points was considered clinical improvement.	NR
Lawrance 2017	UC	People who failed conventional therapies of oral or rectal (or both) 5‐aminosalicylate or oral and rectal steroids (or both), or were intolerant of these medications.	Clinical remission: observed by a Mayo score ≤ 2 with no subscore > 1. In addition to endoscopic score of 0 or 1 indicating mucosal healing. Clinical improvement: a reduction in Mayo score ≥ 3 points and a decrease of > 30% from the baseline score. In addition, a reduction of ≥ 1 on the rectal bleeding subscale, or alternatively an absolute rectal score of 0 or 1.	Authors declared no conflicts of interest.
Lie 2020	UP	Mesalamine‐refractory UP (defined as a failure to at least the use of mesalamine suppositories of a maximum of 1 g for ≥ 21 days) or recurring UP (defined as a relapse within 3 months after stopping adequate local mesalamine therapy).	Clinical remission: defined as a Mayo score ≤ 2, and endoscopic remission as no visible inflammation (i.e. Mayo subscore 0). Clinical response: defined as an absolute decrease in Mayo score of 3 points, with a relative decrease of 30% of the total score and ≥ 1 point decrease in the rectal bleeding subscore or an absolute rectal bleeding subscore of 0 or 1.	Authors declared no conflicts of interest.
Ogata 2006	UC	Steroid resistance defined as unresponsiveness to oral or intravenous corticosteroid therapy. Steroid dependency was defined as either chronic active UC for > 6 months or frequent recurrence (> once a year, or ≥ 3 times every 2 years regardless of intensive medical therapy).	Clinical remission: defined as a DAI score ≤ 2, with no individual subscore > 1, and mucosal healing was defined as an endoscopy subscore (≥ 2 at entry) of 0 or 1. Clinical improvement: defined as combination of partial and complete response. Partial response was defined as a reduction of > 4 points on DAI with improvement in all categories. Complete response was defined as resolution of all symptoms (all assessment scores 0).	Authors declared no conflicts of interest.
Ogata 2012	UC	Steroid resistance was when the disease failed to respond to a systemic daily dose of 1 mg/kg bodyweight, or ≥ 40 mg of prednisolone given over ≥ 7 days, or the equivalent of a daily dose of prednisolone of ≥ 30 mg over ≥ 2 weeks. Steroid‐dependent participants were defined as people with active UC in whom attempts to taper steroids had been unsuccessful.	Clinical remission: defined as a DAI score ≤ 2, with an individual subscore 0 or 1, and mucosal healing defined as an endoscopy subscore of 0 or 1. Clinical response: defined as a reduction in DAI by ≥ 4 points and improvements in all categories (stool frequency, rectal bleeding, mucosal appearance, and physician’s overall assessment).	Authors declared no conflicts of interest.
CAI: Clinical Activity Index; DAI: Disease Activity Index; NR: not reported; UC: ulcerative colitis; UP: ulcerative proctitis.

Table 2. Summary of clinical characteristics and conflicts of interest of included studies

Comparison 1. Tacrolimus versus placebo

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Achievement of clinical remission Show forest plot	3	148	Risk Ratio (M‐H, Random, 95% CI)	3.76 [1.03, 13.73]

1.1.1 Oral tacrolimus – low target serum concentration of tacrolimus: 5–10 ng/mL	1	33	Risk Ratio (M‐H, Random, 95% CI)	2.29 [0.12, 43.84]
1.1.2 Oral tacrolimus – high target serum concentration of tacrolimus: 10–15 ng/mL	2	94	Risk Ratio (M‐H, Random, 95% CI)	3.07 [0.57, 16.58]
1.1.3 Rectal tacrolimus vs placebo	1	21	Risk Ratio (M‐H, Random, 95% CI)	10.08 [0.63, 162.06]
1.2 Achievement of clinical remission for oral tacrolimus vs placebo (subgroup analysis) Show forest plot	2	127	Risk Ratio (M‐H, Random, 95% CI)	2.85 [0.66, 12.35]

1.2.1 Oral tacrolimus – low target serum concentration of tacrolimus: 5–10 ng/mL	1	33	Risk Ratio (M‐H, Random, 95% CI)	2.29 [0.12, 43.84]
1.2.2 Oral tacrolimus – high target serum concentration of tacrolimus: 10–15 ng/mL	2	94	Risk Ratio (M‐H, Random, 95% CI)	3.07 [0.57, 16.58]
1.3 Achievement of clinical remission for rectal tacrolimus vs placebo (subgroup analysis) Show forest plot	1	21	Risk Ratio (M‐H, Random, 95% CI)	10.08 [0.63, 162.06]

1.3.1 Rectal tacrolimus versus placebo	1	21	Risk Ratio (M‐H, Random, 95% CI)	10.08 [0.63, 162.06]
1.4 Clinical improvement Show forest plot	3	148	Risk Ratio (M‐H, Random, 95% CI)	4.47 [2.15, 9.29]

1.4.1 Low target serum concentration of tacrolimus: 5–10 ng/mL	1	33	Risk Ratio (M‐H, Random, 95% CI)	3.48 [0.50, 24.25]
1.4.2 High target serum concentration of tacrolimus: 10–15 ng/mL	2	94	Risk Ratio (M‐H, Random, 95% CI)	4.25 [1.78, 10.12]
1.4.3 Rectal tacrolimus vs placebo	1	21	Risk Ratio (M‐H, Random, 95% CI)	7.27 [1.09, 48.35]
1.5 Clinical improvement for oral tacrolimus vs placebo (subgroup analysis) Show forest plot	2	127	Risk Ratio (M‐H, Random, 95% CI)	4.11 [1.86, 9.08]

1.5.1 Low target serum concentration of tacrolimus: 5–10 ng/mL	1	33	Risk Ratio (M‐H, Random, 95% CI)	3.48 [0.50, 24.25]
1.5.2 High target serum concentration of tacrolimus: 10–15 ng/mL	2	94	Risk Ratio (M‐H, Random, 95% CI)	4.25 [1.78, 10.12]
1.6 Clinical improvement for rectal tacrolimus vs placebo (subgroup analysis) Show forest plot	1	21	Risk Ratio (M‐H, Random, 95% CI)	7.27 [1.09, 48.35]

1.6.1 Rectal tacrolimus versus placebo	1	21	Risk Ratio (M‐H, Random, 95% CI)	7.27 [1.09, 48.35]
1.7 Serious adverse events Show forest plot	3	148	Risk Ratio (M‐H, Random, 95% CI)	2.44 [0.12, 48.77]

1.8 Total adverse events Show forest plot	3	148	Risk Ratio (M‐H, Random, 95% CI)	1.18 [0.91, 1.54]

Comparison 1. Tacrolimus versus placebo

Comparison 2. Tacrolimus versus ciclosporin

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Achievement of clinical remission Show forest plot	1	113	Risk Ratio (M‐H, Random, 95% CI)	1.52 [0.92, 2.50]

2.2 Clinical improvement Show forest plot	1	113	Risk Ratio (M‐H, Random, 95% CI)	0.90 [0.70, 1.16]

Comparison 2. Tacrolimus versus ciclosporin

Comparison 3. Tacrolimus versus beclometasone

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Achievement of clinical remission Show forest plot	1	88	Risk Ratio (M‐H, Random, 95% CI)	1.07 [0.60, 1.88]

3.2 Clinical improvement Show forest plot	1	88	Risk Ratio (M‐H, Random, 95% CI)	1.00 [0.66, 1.52]

3.3 Serious adverse events Show forest plot	1	88	Risk Ratio (M‐H, Random, 95% CI)	3.00 [0.13, 71.70]

3.4 Total adverse events Show forest plot	1	88	Risk Ratio (M‐H, Random, 95% CI)	1.50 [0.88, 2.55]

Comparison 3. Tacrolimus versus beclometasone