Azathioprine and 6‐mercaptopurine for maintenance of surgically‐induced remission in Crohn's disease

Summary of findings for the main comparison. Azathioprine or 6‐mercaptopurine compared to placebo for maintenance of surgically‐induced remission in Crohn's disease

Azathioprine or 6‐mercaptopurine compared to placebo for maintenance of surgically‐induced remission in Crohn's disease
Patient or population: People with surgically‐induced remission in Crohn's disease Setting: Outpatient Intervention: Azathioprine (100‐150 mg/day) or 6‐mercaptopurine (1 mg/kg/day ‐ 50 mg/day) 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 Azathioprine or 6‐mercaptopurine
Clinical relapse Follow‐up: 12 to 36 months	Study population		RR 0.79 (0.67 to 0.92)	408 (3 RCTs)	⊕⊕⊕⊝ MODERATE¹	Clinical relapse defined as: a CDAI>250 (D'Haens 2008); a CDAI>150 and 100 point increase from baseline (Mowat 2016) or a grading score > 2 (Hanauer 2004).
	642 per 1,000	508 per 1,000 (430 to 591)
Endoscopic relapse Follow‐up:12 to 36 months	Study population		RR 0.85 (0.64 to 1.13)	321 (2 RCTs)	⊕⊕⊝⊝ LOW ^{2 3}	Endoscopic relapse defined as an endoscopic score i ≥2 (D'Haens 2008; Mowat 2016)
	752 per 1,000	639 per 1,000 (481 to 849)
Radiologic relapse	Outcome not reported					Not reported
Surgical relapse	Outcome not reported					Not reported
Adverse events Follow‐up: 12 to 24 months	Study population		RR 1.36 (0.57 to 3.27)	168 (2 RCTs)	⊕⊕⊝⊝ LOW ⁴	Reported adverse events include hair loss, leukopenia, diarrhoea, abdominal pain, hepatotoxicity and arthralgia
	99 per 1,000	134 per 1,000 (56 to 323)
Serious adverse events Follow‐up:12 to 36 months	Study population		RR 1.78 (0.39 to 8.18)	327 (2 RCTs)	⊕⊕⊝⊝ LOW ⁵	Reported serious adverse events include arthralgia, pancreatitis, leucopenia and bowel obstruction
	13 per 1,000	23 per 1,000 (5 to 108)
Withdrawal due to adverse events Follow‐up:12 to 36 months	Study population		RR 0.90 (0.63 to 1.29)	408 (3 RCTs)	⊕⊕⊕⊝ MODERATE⁶	Adverse events leading to withdrawal included abnormal blood results leading to temporary discontinuation of treatment in 28% of the participants. However, specific details on reasons for discontinuation were not clearly stated
	254 per 1,000	228 per 1,000 (160 to 328)
*The risk in the intervention group (and its 95% confidence interval) is based on the median risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio; OR: Odds ratio;
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
¹ Downgraded one level due serious imprecision (233 events). ² Downgraded one level due to serious inconsistency ( I² = 64%). ³ Downgraded one level due to serious imprecision (227 events). ⁴ Downgraded two levels due to very serious imprecision (20 events) and 95% CI which includes appreciable benefit and harm ⁵ Downgraded two levels due to very serious imprecision (7 events) and 95% CI which includes appreciable benefit and harm ⁶ Downgraded one level due to serious imprecision (100 events) and 95% CI which includes appreciable benefit and harm

Summary of findings 2. Azathioprine or 6‐mercaptopurine compared to 5‐aminosalicylic acid for maintenance of surgically‐induced remission in Crohn's disease

Azathioprine or 6‐mercaptopurine compared to 5‐aminosalicylic acid for maintenance of surgically‐induced remission in Crohn's disease
Patient or population: People with surgically‐induced remission in Crohn's disease Setting: Outpatient Intervention: Azathioprine (2 mg/kg/day) or 6‐mercaptopurine (50 mg/day) Comparison: 5‐aminosalicylic acid
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with 5‐aminosalicylic acid	Risk with Azathioprine or 6‐mercaptopurine
Clinical relapse Follow‐up: 12 to 24 months	Study population		RR 1.05 (0.89, 1.24)	347 (4 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	Clinical relapse defined as a clinical grading score ≥ 2 (Hanauer 2004;Savarino 2013) or CDAI ≥ 200 (Ardizzone 2004)
	556 per 1,000	595 per 1,000 (456 to 773)
Endoscopic relapse Follow‐up: 24 months	Study population		RR 0.78 (0.52 to 1.17)	35 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{1 3}	Endoscopic relapse defined as a Rugeerts score ≥ 2 (Savarino 2013)
	833 per 1,000	650 per 1,000 (433 to 975)
Radiologic relapse Follow‐up: 24 months	Study population		RR 0.92 (0.66 to 1.28)	35 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{1 4}	Radiologic relapse defined as a radiographic grading score ≥ 2 (Savarino 2013)
	833 per 1,000	767 per 1,000 (550 to 1,000)
Surgical relapse Follow‐up: 24 months	Study population		RR 0.81 (0.50 to 1.29)	142 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{1 5}	Surgical relapse defined as a need for another surgery (Ardizzone 2004)
	366 per 1,000	297 per 1,000 (183 to 472)
Adverse events Follow‐up: 12 to 24 months	Study population		RR 0.89 (0.74 to 1.07)	346 (4 RCTs)	⊕⊕⊝⊝ LOW ^{1 6}	Reported adverse events include leukopenia, abdominal pain, nausea, nasopharyngitis, diarrhoea. and headache
	476 per 1,000	424 per 1,000 (353 to 510)
Serious adverse events Follow‐up: 12 to 24 months	Study population		RR 3.39 (1.26 to 9.13)	311 (3 RCTs)	⊕⊝⊝⊝ VERY LOW ^{1 7}	Reported serious adverse events include postoperative bowel obstruction
	39 per 1,000	134 per 1,000 (50 to 360)
Withdrawal due to adverse events Follow‐up: 12 to 24 months	Study population		RR 2.21 (1.28, 3.81)	425 (4 RCTs)	⊕⊕⊝⊝ LOW ^{1 8}	Adverse events leading to withdrawal include Severe epigastric intolerance, increase in liver function test results, leukopenia, acute pancreatitis
	76 per 1,000	172 per 1,000 (93 to 317)
*The risk in the intervention group (and its 95% confidence interval) is based on the median risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio; OR: Odds ratio;
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
¹ Downgraded one level due to high risk of bias ² Downgraded one level due to serious imprecision (214 events) and 95% CI which includes no effect and appreciable harm ³ Downgraded two levels due to very serious imprecision (26 events) and 95% CI which includes no effect and appreciable benefit ⁴ Downgraded two levels due to very serious imprecision (28 events) and 95% CI which includes no effect and appreciable benefit ⁵ Downgraded two levels due to very serious imprecision (47 events) and 95% CI which includes appreciable benefit and harm ⁶ Downgraded one level due to serious imprecision (154 events) and 95% CI which includes no effect and appreciable benefit ⁷ Downgraded two levels due to serious imprecision (33 events) ⁸ Downgraded one level due to serious imprecision (58 events)

Summary of findings 3. Azathioprine or 6‐mercaptopurine compared to anti TNF‐α for maintenance of surgically‐induced remission in Crohn's disease

Azathioprine or 6‐mercaptopurine compared to anti‐TNF‐α for maintenance of surgically‐induced remission in Crohn's disease
Patient or population: People with surgically‐induced remission in Crohn's disease Setting: Outpatient Intervention: Azathioprine (2‐2.5 mg/kg/day) or 6‐mercaptopurine (1.5 mg/kg/day) Comparison:Anti‐TNF‐α
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with anti‐ TNF‐α	Risk with Azathioprine or 6‐mercaptopurine
Clinical relapse Follow‐up: 12 to 24 months	Study population		RR 2.89 (1.50 to 5.57)	139 (3 RCTs)	⊕⊕⊝⊝ VERY LOW ^{1 2}	Clinical relapse defined as: an HBI ≥2 (Armuzzi 2013), a clinical recurrence grading score ≥2 (Savarino 2013) or a CDAI score >200 (Lopez‐Sanroman 2017)
	139 per 1,000	401 per 1,000 (208 to 774)
Endoscopic relapse Follow‐up: 12 to 24 months	Study population		RR 3.67 (1.05 to 12.81)	157 (4 RCTs)	⊕⊝⊝⊝ VERY LOW ^{1 3 4}	Endoscopic relapse defined as a Rugeerts score ≥ 2 (Armuzzi 2013; Lopez‐Sanroman 2017; Savarino 2013; Scapa 2015)
	265 per 1,000	973 per 1,000 (278 to 1,000)
Radiologic relapse Follow‐up: 12 to 24 months	Study population		‐	117 (2 RCTs)	‐	Radiologic relapse defined as a radiographic grading score ≥ 2 (Savarino 2013) or magnetic resonance enterography score ≥2 (Lopez‐Sanroman 2017) *due to considerable heterogeneity (I² = 85%), data pooling was not feasible. It is uncertain whether azathioprine leads to a difference in radiologic relapse when compared to infliximab as the certainty of the evidence is very low (RR1.36, 95% CI 0.94 to 1.98; RR 12.24, 95% CI 1.8 to 83.12)
	see comment	see comment
Surgical relapse	Outcome not reported					Not reported
Adverse events Follow‐up: 12 to 24 months	Study population		RR 1.13 (0.83 to 1.53)	117 (2 RCTs)	⊕⊕⊝⊝ LOW ^{1 5}	Adverse events include bronchitis, nasopharyngitis, arthralgia, nausea, abscess were reported in Savarino 2013. Full details were not reported in Lopez‐Sanroman 2017
	508 per 1,000	574 per 1,000 (422 to 778)
Serious adverse events Follow‐up: 12 months	Study population		RR 0.51 (0.17 to 1.54)	84 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{1 6}	Serious adverse events not reported
	200 per 1,000	102 per 1,000 (34 to 308)
Withdrawal due to adverse events Follow‐up: 12 to 24 months	Study population		RR 3.97 (0.92 to 17.22)	139 (3 RCTs)	⊕⊕⊝⊝ LOW ^{1 6}	Adverse events leading to withdrawal included severe nausea, leukopenia, arthralgia, urothelial carcinoma, dyspepsia, dyspnoea, death, atopic dermatitis and abdominal pain with increase in pancreatic enzymes
	28 per 1,000	110 per 1,000 (26 to 478)
*The risk in the intervention group (and its 95% confidence interval) is based on the median risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio; OR: Odds ratio;
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
¹ Downgraded one level due to high risk of bias ² Downgraded two levels due to very serious imprecision (39 events) ³ Downgraded one level due to serious imprecision (65 events) ⁴ Downgraded one level due to substantial heterogeneity (I² = 64%) ⁵ Downgraded one level due to serious imprecision (63 events) ⁶ Downgraded two levels due to very serious imprecision (13 events)

Background

Description of the condition

Crohn’s disease is a chronic idiopathic inflammatory disorder of the gastrointestinal tract with an etiology that combines genetic predisposition, environmental factors and an inappropriate immune response to the gut microbiota which may involve the whole gastrointestinal tract (Abraham 2009). There is no cure for the disease, and management strategies are mainly focused on induction and maintenance of remission. Approximately 75% of patients with Crohn's disease will eventually undergo surgical resection (Bernell 2000), and this can induce remission. However, endoscopic recurrence of disease has been reported to be as high as 73% at one year post surgery (Rutgeerts 1990), and clinical relapse rates have been reported to range from 22 to 55% at five years post surgery (Williams 1990). There is no standard therapy for the prevention of postoperative recurrence in Crohn's disease (Hanauer 2001). A number of agents have been studied, but considerable uncertainty remains as to the efficacy of such treatments.

Description of the intervention

Corticosteroids, the mainstay of treatment of acute exacerbations, have been used extensively as Crohn's disease therapy. However, the chronic use of glucocorticosteroids is limited due to the multiple adverse reactions and the lack of effectiveness for maintaining remission in Crohn's disease (Steinhart 2003). 5‐aminosalicylic acid (5‐ASA) agents have been shown to be safe and may be effective for maintenance of post‐surgical remission, although the existing data suggests that the efficacy of these agents depends on the ability of these drugs to reach the terminal ileum and colon in therapeutic concentrations and may have limited clinical efficacy (Gordon 2011). Probiotics and budesonide do not appear to provide any benefit for maintenance of surgically‐induced remission in Crohn's disease (Benchimol 2009; Doherty 2009; Rolfe 2006). Nitroimidazole antibiotics may reduce the risk of relapse in surgically‐induced remission, however, these agents are not well tolerated and are associated with a higher risk of serious adverse events (Doherty 2009). Studies have demonstrated that tumour necrosis factor‐alpha (TNF‐α) antagonists such as infliximab (Regueiro 2009), or adalimumab (Savarino 2013) may provide a benefit for reducing the risk of relapse in surgically‐induced remission, but these agents are expensive. Purine analogues such as azathioprine (AZA) and 6‐mercaptopurine (6‐MP) have been used in clinical practice for over five decades with a demonstrated efficacy for the long‐term maintenance of remission in both Crohn's disease (Chande 2015), and ulcerative colitis (Timmer 2016), and are relatively inexpensive. Evidence suggests that the effect of thiopurine formulations seem to last for up to five years (Fraser 2002), significantly reducing the risk of perianal and intestinal surgery (Camus 2013).

How the intervention might work

6‐MP and its prodrug AZA which is non‐enzymatically degraded to 6‐MP are purine antimetabolites that reduce cell proliferation and have immune modulating properties. 6‐MP is metabolised to its active component 6‐thioguanine nucleotide which competitively interferes with nucleic acid metabolism by inhibiting the proliferation of T and B lymphocytes and reducing the numbers of cytotoxic T cells and plasma cells (Lennard 1992; Sahasranaman 2008). There are some trial data which suggest that neutrophil count is a predictor of induction and maintenance of remission (Colonna 1994), which may suggest the mechanism of action, although this is not well understood. The major limiting factor for long term use of AZA and 6‐MP agents has been the occurrence of adverse events in approximately 10% of patients leading to withdrawal of therapy (Hafraoui 2002), with dose‐dependent and idiosyncratic adverse events occurring. There is evidence which suggests that thiopurine methyltransferase deficiency accounts for some of the dose and metabolism‐related toxicity to purine analogues including leucopenia, thrombocytopenia and in the long‐term potentially lymphoma and non‐melanoma skin cancer (Axelrad 2016; Gomollon 2017; Lennard 1989; Weinshilboum 1980), while adverse reactions such as arthralgias, pancreatitis, hepatitis, nausea, non‐pancreatic abdominal pain, rush, fever and diarrhoea are attributed to hypersensitivity reactions (Sandborn 1996).

Why it is important to do this review

Maintenance of remission in Crohn's disease is a major issue as many patients fail to achieve remission with medical management and require surgical interventions. Purine analogues have been used to maintain surgically‐induced remission in Crohn's disease, but the effectiveness, tolerability and safety of these drugs remains controversial. Relatively few studies have been published that investigate the role of AZA or 6‐MP for maintenance of remission following surgery in patients with Crohn's disease. One multicentre randomised placebo controlled trial involving 81 patients found a significant reduction in endoscopic recurrence when AZA was used in conjunction with metronidazole in comparison to metronidazole alone (D'Haens 2008). In another multicentre randomised controlled trial, it was concluded that 6‐mercaptopurine was more effective than either mesalamine or placebo at preventing postoperative recurrence at 24 months following surgery (Hanauer 2001). However, a single‐centre randomised open‐label trial found no significant difference in clinical relapse rates between AZA and mesalamine (Ardizzone 2004). A previous review by this team in 2014 found evidence that purine analogues may be superior to placebo for maintenance of surgically‐induced remission in patients with Crohn's disease, although this was based on two small studies (Gordon 2014). The results for efficacy outcomes between purine analogues and 5‐ASA agents were uncertain. However, patients taking purine analogues were more likely than 5‐ASA patients to discontinue therapy due to adverse events. No firm conclusions could be drawn from the two small studies that compared AZA to infliximab or adalimumab. Adalimumab seemed superior to AZA but further research was needed to confirm these results. Hence, an up‐to‐date systematic review using the Cochrane Collaboration format was indicated to summarise the current evidence on the use of purine analogues for the maintenance of surgically‐induced remission in Crohn's disease.

Objectives

The primary objective was to evaluate the efficacy and safety of AZA and 6‐MP for maintenance of surgically‐induced remission in Crohn's disease.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials which compared AZA or 6‐MP agents to either a no treatment control, placebo or any other active intervention, with treatment durations of at least three months were considered for inclusion.

Types of participants

Participants of any age and sex with a diagnosis of Crohn's disease confirmed by any established method who were in remission following surgery were considered for inclusion. Remission could be defined by a recognized Crohn's disease activity index such as the Crohn's disease activity index (CDAI) or endoscopy, or by participants who have undergone a curative surgical resection, or as defined by the authors of the primary studies. were considered for inclusion. Eligible trials could be conducted in any setting (e.g. single centre or multi‐centre).

Types of interventions

The controlled interventions of interest included any randomised controlled trial that compares oral AZA or 6‐MP agents to an no treatment, placebo or another active intervention for maintenance of surgically‐induced remission. Studies that compare AZA or 6‐MP agents to an intervention that focuses on enteral nutrition, oral nutrient supplementation, medical foods, probiotics, parental nutrition or herbal medicines were excluded. We also excluded dose optimisation studies.

Types of outcome measures

Primary outcomes

The primary outcome measure was clinical relapse as defined by the primary studies.

Secondary outcomes

The secondary outcome measures included the proportion of participants who experienced:

Endoscopic relapse;
Radiologic relapse;
Surgical relapse;
Histologic relapse;
Adverse event (as defined by FDA 2018. We also noted where studies failed to provide sufficient information and simply report outcome as ‘adverse event’);
Serious adverse events (as defined by FDA 2018. We also noted where studies failed to provide sufficient information and simply report outcome as ‘serious adverse event’);
Withdrawal due to adverse events; and
Health‐related quality of life (HRQoL).

Adverse events and serious adverse events that are known to be associated with AZA or 6‐MP include:

Bone marrow suppression: pancytopenia, leucopenia, neutropenia, thrombocytopenia;
Hypersensitive reactions: malaise, vomiting, diarrhoea, rash, hypotension;
Malignancy;
Liver function impairment, jaundice;
Pancreatitis;
Pulmonary: pneumonitis; and
Renal: interstitial nephritis.

The outcome measures were reported at the last time point available (assumed to be at the end of follow‐up if not specified) and the time point specified in the methods as being of primary interest (if this was different from the latest time point available). However, it was also indicated when studies reported outcomes at other time points.

Search methods for identification of studies

Electronic searches

For the review update, we searched the following electronic databases from inception to 26 July 2018:

MEDLINE;
Embase;
PubMed;
CENTRAL; and
Cochrane IBD Group Specialized Register.

No restrictions were placed on language. Note that the searches were designed to include RCTs conducted on adults and children participants, but to exclude dose optimisation studies and trials that compare AZA or 6‐MP agents to oral nutrition supplements (enteral nutrition drinks, tube feeds), medical foods, probiotics, parenteral nutrition, herbal medicines or a combination of these modalities. The search strategy was more than one year old prior to publication of the updated review. Thus, we ran another search from inception to 31 July 2019 prior to publication. The search strategies are reported in Appendix 1.

Searching other resources

Reference searching

We searched reference lists from included articles and any existing relevant reviews. We also searched ongoing trials registered on ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform portal.

Abstracts of major gastroenterology meetings

A manual search of abstracts and proceedings submitted to recent major gastroenterology meetings was performed for the following journals to identify more trials:

Gastroenterology (American Gastroenterological Association);

Gut (British Society of Gastroenterology);

American Journal of Gastroenterology (American College of Gastroenterology); and

Journal of Pediatric Gastroenterology and Nutrition (European / North American Society of Paediatric Gastroenterology, Hepatology and Nutrition).

If a relevant abstract was identified, details of the full study methodology and results were requested from the authors in order to allow a thorough assessment of the quality of identified studies.

Data collection and analysis

This updated review was based upon the methods described in the published protocol (Gordon 2014), and in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

Selection of studies

The selection of studies included the following steps: title screening, abstract screening and full‐text review. Two authors (MG and TGH) independently reviewed each article at each stage of selection. Included and excluded studies were recorded. The two authors (MG and TGH) independently screened the titles using the titles of the papers that appeared to have even a minor possibility of inclusion. Adjudication did not occur at the title screening stage and studies that were ambiguous were included by default. The two authors (MG and TGH) then independently screened the abstracts of the articles that report studies with a reasonable possibility of inclusion. Differences in assessment for inclusion were resolved by discussion between the two independent investigators (MG and TGH). Adjudication did not occur at the abstract screening state. Lastly, the two authors (MG and TGH) independently screened the full text which involved selection of articles based on careful examination of the full report. Differences in assessment for inclusion were resolved by discussion between the two independent investigators. Adjudication was performed as needed by a third author (ZIE).

Data extraction and management

A data extraction form was developed to extract information on relevant features and results of included studies. Two authors (ZIE and TGH) independently extracted and recorded data on the predefined checklist. Extracted data included the following items:

Study design: type of RCT, setting, number of interventions, year, author's contact;
Population characteristics: age, sex, disease distribution, disease duration, site of disease, medication, type and time since operation, total number of participants originally assigned to each treatment group;
Intervention: type and dose of agent;
Control: no active treatment, placebo, other drugs;
Concurrent medications; and
Outcomes: time of assessment, length of follow up, type of Crohn's disease activity index used, definitions of remission and relapse, site of surgery, relapse rates, adverse events.

Assessment of risk of bias in included studies

Two authors (ZEI and TGH) independently assessed bias using the Cochrane risk of bias tool (Higgins 2011). Adjudication was performed as needed by a third author (MG). Each domain was assessed as having a low, high, or unclear risk of bias. Domains assessed included:

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

Allocation sequence concealment (i.e. was allocation adequately concealed?);

Blinding (i.e. was knowledge of the allocated intervention adequately prevented during the study?);

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

Selective outcome reporting (i.e. are reports of the study free of suggestion of selective outcome reporting?); and

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

Each domain followed standard definitions used for Cochrane systematic reviews (Higgins 2011). Study authors were contacted for further information when insufficient information was provided to determine the risk of bias.

The overall certainty of the evidence was assessed using the GRADE approach (Guyatt 2008; Schünemann 2011). The GRADE approach appraises the quality of a body of evidence based on the extent to which one can be confident that an estimate of effect or association reflects the item being assessed. Randomised trials start as high quality evidence, but may be downgraded due to: risk of bias (methodological quality), indirectness of evidence, unexplained heterogeneity, imprecision (sparse data) and publication bias. The overall quality of the evidence for each outcome was determined after considering each of these factors and graded as:

High ‐ we are very confident that the true effect lies close to that of the estimate of the effect;
Moderate ‐ 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 ‐ our confidence in the effect estimate is limited. The true effect may be substantially different from the estimate of the effect; or
Very low. We have very little confidence in the effect estimate. The true effect is likely to be substantially different from the estimate of effect.

Measures of treatment effect

For binary outcomes, risk ratio (RR) estimates and associated two‐sided 95% confidence intervals (CI) were calculated. For nominal or ordinal outcomes, we calculated the RR with corresponding 95% for each category relative to a reference category. For continuous outcomes, we calculated the mean difference (MD) and corresponding 95% CI. If studies in future updates report continuous outcomes that have been measured using different scales (e.g. IBDQ and SF‐36), we will calculate the standardised mean difference (SMD) and 95% CI.

Unit of analysis issues

The unit of analysis was the individual participant. We planned to include cross‐over trials if data were available from the first phase of the study (i.e. before cross‐over occurred). For outcomes where events recur (e.g. clinical relapses, adverse events), we calculated the proportion of participants who experienced at least one event, individual events were not counted separately. The studies were otherwise not anticipated to have repeated observations of outcomes or multiple treatment events. If studies had randomised subjects to more than one AZA or 6‐MP treatment arm, these groups would have been combined for the primary analysis.

Dealing with missing data

We collected information on how each trial handled missing data. When a study appeared to collect and not report all primary outcomes of interest, the original investigators were contacted to request missing data. If the original investigators did not provide the data, this would be noted in the systematic review. For studies with missing dichotomous data, an intention‐to‐treat analysis was performed where participants with missing data were assumed to have been treatment failures.

Assessment of heterogeneity

We assessed heterogeneity through visual inspection of the forest plots and by calculating the Chi² and I² statistics (Boreinstein 2009). For studies that had qualitative homogeneity, statistical heterogeneity was assessed using the Chi² test (a P value < 0.10 was considered statistically significant heterogeneity). The degree of heterogeneity across studies was estimated using the I² statistic. An I² of 25% or less was considered low heterogeneity, 26% to 50% was considered moderate heterogeneity, and 50% and greater was considered substantial heterogeneity. Where sufficient data are available, we planned to explore possible explanations for heterogeneity including factors such as participant characteristics (e.g. age, sex), condition severity, treatment type and dose, and healthcare system/country. Where appropriate, these factors were to be investigated further through sub‐group analyses and meta‐regression (Boreinstein 2009). Where sufficient data are available, we planned to use sensitivity analyses to explore possible causes of methodological heterogeneity (Sutton 2000).

Assessment of reporting biases

If there were an appropriate number of studies in a pooled analysis (i.e. > 10 studies), we planed to investigate potential publication bias using funnel plots (trial effects versus trial size) (Egger 2001). However, the number of studies in each comparison group was smaller than 10.

Data synthesis

Data from individual trials were combined for meta‐analysis if the interventions, patient groups and outcomes were sufficiently similar (determined by consensus). We calculated the pooled RR and corresponding 95% CI for dichotomous outcomes. Analyses were grouped by type of intervention treatment (e.g. AZA or 6‐MP versus placebo, AZA or 6‐MP versus 5‐ASA). Where there were multiple studies in an analysis we used a random‐effects model to obtain a more conservative interpretation otherwise, we used a fixed‐effect model.

Subgroup analysis and investigation of heterogeneity

We planned to assess the impact of potential effect modifiers such as age of participants (paediatric versus adult studies), drug type (azathioprine versus 6‐mercaptopurine) and length of follow‐up (12 months or less versus greater than 12 months). There were no studies on children.

Sensitivity analysis

Sensitivity analyses based on random‐effects versus fixed‐effect models were planned where appropriate data or numbers of studies were available. Sensitivity analysis was also planned to explore possible explanations for significant heterogeneity.

Results

Description of studies

Results of the search

The original search for this update was conducted on 26 July 2018. The search expired before the update could be published, so we ran another search on 31 July 2019. The electronic database search conducted on 31 July 2019 identified 1201 records, while 37 records were found through other sources (See Figure 1). Of the 52 full‐text records assessed for eligibility, 35 reports of 10 studies (928 participants) were included in this systematic review update. Two studies (NCT03185611; NL1344), were classified as ongoing Characteristics of ongoing studies. Eleven studies (15 reports) were excluded for different reasons as presented in the Characteristics of excluded studies table.

Figure 1

Study flow diagram.

Detailed information about all included studies are presented in the Characteristics of included studies table.

Included studies

Study design and setting

This systematic review includes reports of single centre and multicentre randomised controlled trials with a parallel design with a duration of 52 weeks (Lopez‐Sanroman 2017; Reinisch 2010), to 36 months (Mowat 2016). These studies were all published between 2004 and 2017. There were four single‐centre studies conducted in two different countries: Italy (Ardizzone 2004; Armuzzi 2013; Savarino 2013), and Israel (Scapa 2015). The multicentre studies were conducted across Belqium (D'Haens 2008), Spain (Lopez‐Sanroman 2017, the UK (Mowat 2016), or as a multinational collaboration of several countries across Europe and Israel (Reinisch 2010), and Europe and the USA (Hanauer 2004). Regarding the care setting, three studies were conducted either in gastroenterology hospitals and medical clinics/centres (Lopez‐Sanroman 2017; Reinisch 2010), and secondary and tertiary hospitals (Mowat 2016), or as a collaboration between university clinics and hospitals and medical centres in four studies (Ardizzone 2004; D'Haens 2008; Hanauer 2004; Savarino 2013). The country and care setting were not reported in Herfarth 2006.

Participants

The total number of participants included in nine studies was 928 and ranged from 22 participants (Armuzzi 2013), to 240 (Mowat 2016). Scapa 2015 did not clearly report the number of participants randomised. All participants were adults with Crohn's disease who had undergone a resective surgical procedure to remove macroscopic disease. The majority of these participants were recruited within three months of surgery or before hospital discharge, except the 78 participants in Reinisch 2010 who were enrolled between 6 and 24 months postoperatively. Interventions were conducted on participants with quiescent Crohn's disease and the disease activity prior to enrolment was established by generally accepted endoscopic, histological and radiological criteria. However, it is important to note that Reinisch 2010 included participants in subsequent postoperative clinical remission (CDAI < 200), but with signs of moderate to severe endoscopic recurrence. For this reason we only collected data on adverse events from Reinisch 2010.

The age of participants was reported nine studies and ranged between an average of 32.7 years (Scapa 2015), to 40 years (D'Haens 2008) in seven trials. Two trials reported age as median (Armuzzi 2013; Lopez‐Sanroman 2017). All studies appear to have been conducted on a male and female adult population. None of the studies included paediatric participants. Herfarth 2006 did not report any information about the age of participants.

Interventions

All included studies were parallel two arm trials except for two studies that had three intervention arms (Hanauer 2004; Savarino 2013). The duration of the intervention ranged from 52 weeks (Armuzzi 2013; Herfarth 2006; Lopez‐Sanroman 2017; Reinisch 2010; Scapa 2015), to 24 months (Ardizzone 2004; D'Haens 2008; Hanauer 2004; Mowat 2016; Savarino 2013).

The studies compared the efficacy of AZA or 6‐MP agents with placebo or another active treatment. Table 1 reports a summary of all interventions which are also summarized below:

Table 1. Key definitions and outcomes

Comparison	Study ID	Time from surgery till recruitment	Site of surgery % / exclusions	Clinical relapse definition	Endoscopic / surgical/ radiologic/ histological relapse definition/
AZA & 6‐MP versus Placebo
AZA versus Placebo (12 months) both arms 750 mg/day metronidazole (3 months)	D'Haens 2008	2 weeks	Perforating disease 48 *Macroscopic evidence for CD proximally or distally to the site of resection or the presence of frank pancolitis or an ileorectal anastomosis, patients with a stoma; operation for fibrostenosis only	CDAI > 250	Rutgeerts i ≥ 2
6‐MP 50 mg/day versus Placebo (24 months)	Hanauer 2004	Before postoperative hospital discharge	N/A * Active perianal disease or any active disease in other segments of the intestine	Clinical recurrence grading > 2 (Hanauer)	Rutgeerts i ≥ 2 Radiographic relapse: Radiographic recurrence grading > 2
6‐MP 1 mg/kg/day versus Placebo (3 years)	Mowat 2016	≤ 3 months	Ileal 39; Colonic 2; Ileocolonic 59 * Need for further surgery, stricturoplasty alone, formation of a stoma	CDAI > 150 and a 100‐point increase from baseline	Rutgeerts i ≥ 2 HRQOL: IBDQ scores
AZA & 6‐MP vs 5‐ASA
AZA 2 mg/kg versus Mesalamine 3 mg/kg (24 months)	Ardizzone 2004	Maximum 2 weeks	Small bowel only 25.3; Colon 5.6; Small bowel and colon 9.8; upper gastrointestinal tract 16.2 *Surgical procedures other than conservative surgery or for perianal disease only	CDAI > 200 Surgical relapse: need for another surgical procedure	n/a
6‐MP 50 mg/day versus Mesalamine 3 g/day (24 months)	Hanauer 2004	Before postoperative hospital discharge	N/A * Active perianal disease or any active disease in other segments of the intestine	clinical recurrence grading > 2	Rutgeerts i ≥ 2 Radiographic relapse: radiographic recurrence grading > 2
AZA 2 mg/kg/day versus Mesalamine 4 g/day (12 months)	Reinisch 2010	6‐24 months	N/A * Short bowel syndrome, an ileocolonic stoma	CDAI > 200	Rutgeerts i ≥ 2 HRQOL: IBDQ
AZA 2 mg/kg/day versus Mesalazine 3 g/day (24 months)	Savarino 2013	2‐4 weeks	Ileum 49, Ileocolonic 51. * Fibrostenotic stricture, macroscopically active disease not resected at the time of surgery, and presence of a stoma	1. ≥ 2 on the clinical recurrence grading scale by Hanauer 2. CDAI > 200	Rutgeerts i≥2 Radiologic relapse : ≥ 2 radiographic recurrence grading scale HRQl: IBDQ>170
AZA & 6‐MP vs anti‐TNF‐α
AZA 2.5 mg versus Infliximab 5 mg/kg (12 months)	Armuzzi 2013	2‐4 weeks	Not reported *Active perianal disease, presence of stoma	HBI ≥ 8	Rutgeerts' score ≥ i2
AZA 2 mg/kg/day versus Adalimumab (24 months)	Savarino 2013	2‐4 weeks	Ileum 49, Ileocolonic 51. * Fibrostenotic stricture, macroscopically active disease not resected at the time of surgery, and presence of a stoma	1. ≥ 2 on the clinical recurrence grading scale by Hanauer 2. CDAI > 200	Rutgeerts i ≥ 2 Radiologic relapse : ≥ 2 radiographic recurrence grading scale HRQl: IBDQ > 170
AZA 2.5 mg/kg/day versus Adalimumab (52 weeks) both arms 750mg/day metronidazole (3 months)	Lopez‐Sanroman 2017	2 weeks	Ileal 58, ileocolonic 41 * Postsurgical stoma, resection for short indolent stenosis, inaccessible anastomosis to endoscopy	CDAI > 200	Rutgeerts i ≥ 2
6‐MP 1.5 mg/kg/day versus Adalimumab (12 months)	Scapa 2015	< 45 days	6‐MP 1.5 mg/kg/day vs Placebo (12 months)	Scapa 2015	< 45 days

AZA: azathioprine; 6‐MP: 6‐mecarptopurine; mg: milligram; CD: Crohn's disease; CDAI: Crohn's disease activity index; NA: not applicable; kg: kilogram; g: gram; HRQOL: health related quality of life; IBDQ: inflammatory bowel disease questionnaire; TNF: tumour necrosis factor; HBI: Harvey Bradshaw index; 5‐ASA: 5‐aminosalicylic acid

AZA or 6‐MP versus placebo

AZA versus placebo (D'Haens 2008).

Both intervention arms also received concomitant metronidazole (750 mg/day) therapy for the first three months of the study (D'Haens 2008).

6‐Mercaptopurine versus placebo (Hanauer 2004; Mowat 2016).

AZA or 6‐MP versus oral 5‐ASA agents

AZA versus mesalamine (Ardizzone 2004; Herfarth 2006; Reinisch 2010; Savarino 2013).
6‐MP versus mesalamine (Hanauer 2004).

AZA or 6‐MP versus anti‐TNF‐α

AZA versus infliximab (Armuzzi 2013).
AZA versus adalimumab (Lopez‐Sanroman 2017; Savarino 2013).

In Lopez‐Sanroman 2017 both intervention arms also received concomitant metronidazole (750 mg/day) therapy for the first three months of the study.

6‐Mercaptopurine versus adalimumab (Scapa 2015).

No studies that compared AZA or 6‐MP agents to a no treatment control group were identified.

The use of concurrent treatment was discussed in all but two studies (Savarino 2013; Scapa 2015). In one study all participants were receiving oral metronidazole (500 mg/day) for two weeks after surgery (Armuzzi 2013), while in two studies both intervention arms were administered metronidazole (ornidazole) for the first three months after surgery (D'Haens 2008; Lopez‐Sanroman 2017). In Mowat 2016 any concomitant therapy had to be well documented and there was no reported use of active concomitant treatments. Corticosteroids were allowed to be tapered by standardized stepwise dose reductions in three studies (Ardizzone 2004; D'Haens 2008; Hanauer 2004). Symptomatic treatment with antacids, antidiarrhoeal or spasmolytic medication on demand was permitted in three studies but had to be scrupulously recorded (Ardizzone 2004; D'Haens 2008; Hanauer 2004). D'Haens 2008 permitted topical therapy for perianal disease and cholestyramine for the treatment of bile‐acid diarrhoea. Continous use of nonsteroidal anti‐inflammatory drugs was prohibited and only occasional use of paracetamol and tramadol was allowed in Savarino 2013.

Outcomes

Outcomes were reported at multiple time points in two studies (D'Haens 2008; Mowat 2016), and at a single time point in eight studies (Ardizzone 2004; Armuzzi 2013; Hanauer 2004; Herfarth 2006; Lopez‐Sanroman 2017; Reinisch 2010; Savarino 2013; Scapa 2015). Some studies had followed participants beyond the intervention period, however, outcome data from those time points were not reported in this review.

Outcomes of interest reported across studies included:

Primary outcomes

Clinical relapse (Ardizzone 2004; Armuzzi 2013; D'Haens 2008; Hanauer 2004; Herfarth 2006; Lopez‐Sanroman 2017; Mowat 2016; Reinisch 2010; Savarino 2013).

Secondary outcome

Endoscopic relapse (Armuzzi 2013; D'Haens 2008; Hanauer 2004; Mowat 2016; Reinisch 2010; Savarino 2013; Scapa 2015).
Radiological relapse (Hanauer 2004; Savarino 2013).
Histologic relapse (Armuzzi 2013).
Surgical relapse (Ardizzone 2004).
Adverse events (Ardizzone 2004; D'Haens 2008; Hanauer 2004; Lopez‐Sanroman 2017; Reinisch 2010; Savarino 2013).
Serious adverse events (Ardizzone 2004; Hanauer 2004; Lopez‐Sanroman 2017; Mowat 2016; Reinisch 2010).
Withrawal due to adverse events (Ardizzone 2004; Armuzzi 2013; D'Haens 2008; Hanauer 2004; Herfarth 2006; Lopez‐Sanroman 2017; Mowat 2016; Reinisch 2010; Savarino 2013).

A summary of interventions and outcomes is presented in additional Table 2.

See: Summary of findings for the main comparison Azathioprine or 6‐mercaptopurine compared to placebo for maintenance of surgically‐induced remission in Crohn's disease; Summary of findings 2 Azathioprine or 6‐mercaptopurine compared to 5‐aminosalicylic acid for maintenance of surgically‐induced remission in Crohn's disease; Summary of findings 3 Azathioprine or 6‐mercaptopurine compared to anti TNF‐α for maintenance of surgically‐induced remission in Crohn's disease

Table 2. Summary of interventions and outcomes

Study ID	Group 1	Group 2	Group 3	Relapse	Quailty of Life	Adverse Events/ Serious adverse/ Withdrawal due to adverse events
Ardizzone 2004	Azathioprine (2mg/kg/day)	Mesalamine (3g/day)		Clinical: 32/71 vs 35/71 Surgical: 26/71 vs 21/71	n/a	AE:18/71 vs 27/71 SAE:6/71 vs 15/71 Withdrwal due to AE: 6/71 vs 15/71
Armuzzi 2013	Azathioprine (2.5 mg/kg/day) Infliximab (5 mg/kg/day)	Infliximab (5 mg/kg/day)		Clinical: 2/11 vs 1/11 Endoscopic: 5/11 vs 1/11; Histologic: 9/11 vs 2/11	n/a	Withdrawal due to AE: 0/11 vs 1/11
D'Haens 2008	Metronidazole (750 mg/day) first 3 mo + Azathioprine (100‐150 mg/day)	Metronidazole (750mg/day) first 3 months + Placebo		Clinical: 11/40 vs 19/41 Endoscopic: 22/40 vs 32/41	n/a	AE: 3/40 vs 4/41 Withdrawal due to AE: 3/40 vs 4/41
Hanauer 2004	6‐Mercaptopurine (50 mg/day)	Mesalamine (3 g/day)	Placebo	Clinical: 32/47 vs 33/44 vs 35/40	n/a	AE: 9/47 vs 6/44 vs 4/40 SAE: 2/47 vs 0/44 vs 2/40 Withdrawal due to AE: 9/47 vs 6/44 vs 4/40
Lopez‐Sanroman 2017	Azathioprine (2.5 mg/kg/d) + Metronidazole (750 mg/day) first 3 mo	Adalimumab + Metronidazole (750 mg/day) first 3 mo		Clinical: 14/39 vs 7/45 Endoscopic: 23/39 vs 19/45 Radiologic: 26/39 vs 22/45	n.s. changes between groups	AE: 20/45 vs 18/39 SAE: 9/45 vs 4/39 Withdrawal due to AE: 1/39 vs 9/45
Mowat 2016	Mercaptopurine (1 mg/kg/day)	Placebo		Clinical: 66/128 vs 70/112 Endoscopic: 90/128 vs 83/112	n.s differences	SAE: 3/128 vs 2/112 Withdrawal due to AE: 39/128 vs 41/112
Reinisch 2010	Azathioprine (2.0‐2.5 mg/kg/d) + Placebo mesalazine	Mesalazine (4g/d) + Placebo azathioprine		Not included	Mean IBDQ change	AE: 34/37 vs 32/41 SAE: 0/37 vs 10/41 Withdrawal due to AE: 1/37 vs 10/41
Savarino 2013	Adalimumab (160‐80 mg 0‐2 weeks and 40 mg/week thereafter)	Azathioprine (2 mg/kg/day)	Mesalamine (3 g/day)	Clinical by Hanauer score: 2/16 vs 12/17 vs 9/18 Clinical by CDAI: 1/16 vs 12/17 vs 9/18 Endoscopic: 1/16 vs 11/17 vs 15/18 Radiologic: 1/16 vs 13/17 vs 15/18	HRQOL (IBDQ >170): 14/16 vs 2/17 vs 3/18	AE:11/16 vs 14/17 vs 14/18 Withdrawal due to AE:0/16 vs 1/17 vs 1/18
Scapa 2015	6‐mercaptopurine (1.5 mg/kg/day)	Adalimumab (160‐80‐40 mg/2 week intervals)		Endoscopic: 4/8 vs 1/11	n/a	n/a

mg: milligram; kg: kilogram; g: gram; CDAI: Crohn's disease activity index; NA: not applicable; AE: adverse events; SAE: serious adverse events; HRQOL: health related quality of life; IBDQ: inflammatory bowel disease questionnaire; ns: not significant

Funding and conflict of interest

Four studies were reportedly supported by pharmaceutical companies (Hanauer 2004; Herfarth 2006; Lopez‐Sanroman 2017; Reinisch 2010), but only two declared conflict of interest (Lopez‐Sanroman 2017; Reinisch 2010). The author of one study was contacted to clarify the role of the pharmaceutical company and he confirmed that the company had no role in the study design, data analysis or writing of the paper (Hanauer 2004), whereas the remaining authors did not respond. Two studies were not supported by any grant (Armuzzi 2013; Savarino 2013). Savarino 2013 reported no conflicts of interest. Armuzzi 2013 reported receiving educational grants, and consultancy and lecture fees from a pharmaceutical company. Mowat 2016 was funded by a governmental grant and adequately reported on conflicts of interest. Funding and conflict of interest was not reported in two studies (Ardizzone 2004, D'Haens 2008), but our attempt to clarify this by contacting the authors was unsuccessful.

Excluded studies

Eleven studies (15 reports) were excluded for different reasons. The reasons for exclusion for each study are presented in the Characteristics of excluded studies table and are summarised below:

Three studies were not RCTs (Nos 2000; Reinisch 2013; Robb 2015);

Six studies assessed the wrong intervention (Ferrante 2015; Mañosa 2013; NCT01876264; Wright 2014; Wright 2015; Zhu 2015);

One study was terminated due to slow recruitment (NCT02247258); and

One study assessed the wrong population (Vidigal 2014).

Risk of bias in included studies

The risk of bias was assessed as low in one study (Mowat 2016), high in six studies (Ardizzone 2004; Armuzzi 2013; Herfarth 2006; Lopez‐Sanroman 2017; Savarino 2013; Scapa 2015), and unclear in three (D'Haens 2008; Hanauer 2004; Reinisch 2010). Details of the risk of bias assessment are presented in the Characteristics of included studies tables, and in Figure 2 and Figure 3.

Figure 2

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

Figure 3

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

Allocation

Random sequence generation

In all of the included studies the allocation of participants to an active treatment or placebo was reported as random. Eight studies were judged as being at low risk of bias for random sequence generation as these studies employed computer‐generated randomisation (Ardizzone 2004; D'Haens 2008; Hanauer 2004; Herfarth 2006; Lopez‐Sanroman 2017; Mowat 2016; Reinisch 2010; Savarino 2013). Two studies were judged 'unclear' due to insufficient information on the method of randomisation (Armuzzi 2013; Scapa 2015).

Allocation concealment

In six studies the method of allocation concealment was considered adequate (Hanauer 2004; Herfarth 2006; D'Haens 2008; Lopez‐Sanroman 2017; Reinisch 2010; Savarino 2013). Four studies were judged as unclear risk of bias for allocation concealment as the methods were not adequately described (Ardizzone 2004; Armuzzi 2013; D'Haens 2008; Scapa 2015).

Blinding

Blinding of participants and personnel

Four of the studies included had an open‐label study design and were judged as being at high risk of bias (Ardizzone 2004; Armuzzi 2013; Lopez‐Sanroman 2017; Savarino 2013). All the remaining studies were described as double‐blind. The method of blinding was not adequately described in three studies (D'Haens 2008; Hanauer 2004; Scapa 2015), thus these studies were marked as 'unclear'. Two of these studies failed to describe whether the placebo was sufficiently identical to the intervention to blind study participants (D'Haens 2008; Hanauer 2004), and one study provided insufficient information to make a judgement (Scapa 2015). Due to an adequate description of blinding methods, three studies were assessed as having low risk of performance bias (Herfarth 2006; Mowat 2016; Reinisch 2010).

Blinding of outcome assessment

We assessed one study as having a high risk of detection bias (Armuzzi 2013). Four studies were marked as 'unclear' for detection bias, having failed to adequately describe blinding of outcome assessors (Ardizzone 2004; Reinisch 2010; Savarino 2013; Scapa 2015). The rest of the studies were judged as having a low risk of detection bias, for clearly describing the methods regarding blinding of outcome assessment (D'Haens 2008; Hanauer 2004; Herfarth 2006; Lopez‐Sanroman 2017; Mowat 2016).

Incomplete outcome data

All except two studies reported data fully and documented dropouts and reasons for withdrawals. Incomplete outcome data in these studies was due to study termination (Herfarth 2006), and failure to report the number of randomised and withdrawn participants and reasons for withdrawal (Scapa 2015). The authors of Scapa 2015 were contacted for clarification, however no additional information was provided, except that the study is under preparation for publication. An 'unclear' judgment for this domain was not made for any of the studies.

Selective reporting

Trial registration was available for four studies (Lopez‐Sanroman 2017; Mowat 2016; Reinisch 2010; Scapa 2015). Seven studies were judged as being at low risk of bias for reporting all outcomes prespecified in the trial registration or in the methods section of the study manuscript (Ardizzone 2004; Armuzzi 2013; D'Haens 2008; Hanauer 2004; Mowat 2016; Reinisch 2010; Savarino 2013). One study was judged to be 'unclear' (Herfarth 2006). Two studies were marked 'high' for reporting bias: Scapa 2015 failed to report outcomes prespecified in the trial registration and Lopez‐Sanroman 2017 did not adequately report on a prespecified outcome.

Other potential sources of bias

Eight studies were judged to be at low risk of bias for other apparent sources of potential bias. Two studies provided insufficient information to enable the reviewers make a judgement and were rated as 'unclear' (Herfarth 2006; Scapa 2015).

Effects of interventions

AZA or 6‐MP versus placebo

Three studies that compared AZA (100 to 150 mg/day) or 6‐MP in doses of 50 mg/day and 1 mg/kg/day to placebo were identified (D'Haens 2008; Hanauer 2004; Mowat 2016). In one of these studies all participants were also taking either metronidazole or ornidazole (750 mg/day) for the first three months of intervention (D'Haens 2008).

Primary outcome

Clinical relapse

Three studies reported on clinical relapse, with definitions for clinical relapse varying across studies. D'Haens 2008 defined clinical relapse as a CDAI > 250. Mowat 2016 defined clinical relapse as a CDAI > 150 and a 100 point increase in CDAI from baseline. Hanauer 2004 defined relapse as a clinical reoccurrence grading score > 2. There was moderate certainty evidence that AZA or 6‐MP are more efficient in preventing clinical relapse than placebo (Analysis 1.1; summary of findings Table for the main comparison). After a follow‐up of 12 to 36 months, 51% (109/215) of participants in the AZA/6‐MP group relapsed compared to 64% (124/193) of the placebo group (RR 0.79; 95% CI 0.67 to 0.92; 408 participants; 3 studies; I² = 0%; GRADE moderate certainty evidence). The subgroup analysis found no evidence of a difference in clinical relapse between AZA and 6‐MP (P = 0.34). A subgroup analysis based on length of follow‐up found no evidence of a difference in clinical relapse when measured at 12 months or less or at over 12 months (P = 0.34).

Secondary outcomes

The effect on endoscopic relapse rate as well as the tolerability and safety of AZA/6‐MP agents compared to placebo was uncertain, due to very low to low certainty evidence (summary of findings Table for the main comparison).

Endoscopic relapse

Endoscopic relapse (Analysis 1.3), defined as endoscopic score i ≥ 2, was reported in two studies (D'Haens 2008; Mowat 2016). During a follow‐up period of 12 to 36 months, 67% (112/168) of AZA/6‐MP participants relapsed endoscopically, compared to 75% (115/153) of placebo participants (RR 0.85; 95% CI 0.64 to 1.13; 321 participants; 2 studies; I² = 62%; GRADE low certainty evidence). We found no evidence of a difference in endoscopic relapse rates between AZA and 6‐MP (P = 0.11). A subgroup analysis based on length of follow‐up found no evidence of a difference in endoscopic relapse rates measured at different times (P = 0.11). Hanauer 2004 compared endoscopic relapse rates (defined as i ≥ 2) between the 6‐MP (16%; 95% CI 7% to 35%) and placebo (42%; 95% CI 21% to 70%) groups at 24 months (HR 0.48; reported P = 0.13; 87 participants). However, these data were insufficiently reported to be included in the meta‐analysis.

Radiologic relapse

Radiologic relapse rate defined as a radiographic recurrence grading score ≥ 2 was reported in Hanauer 2004. After 24 months, 33% (95% CI, 19% to 54%) of participants treated with 6‐MP had radiologic relapse compared to 49% (95% CI, 30% to 72%) in the placebo group (HR, 0.61; reported P = 0.19; 84 participants).

Adverse events, serious adverse events and withdrawal due to adverse events

There was no clear difference in the number of participants who experienced adverse events, serious adverse events or withdrawal due to adverse events when AZA/6‐MP drugs are compared to placebo. Adverse events (Analysis 1.5) were reported in two studies ( D'Haens 2008; Hanauer 2004). For the follow‐up period of 12 to 24 months, 14% (12/87) of AZA/6‐MP participants experienced at least one adverse event that was possibly related to treatment compared to 10% (8/81) of the placebo participants (RR 1.36; 95% CI 0.57 to 3.27; 168 participants; 2 studies; I² = 0%; GRADE low certainty evidence). The subgroup analysis found no evidence of a difference in adverse events when AZA and 6‐MP were compared (P = 0.32). A subgroup analysis based on length of follow‐up found no evidence of a difference in adverse events when measured at 12 months or less and over 12 months (P = 0.32). Commonly reported adverse events included hair loss, leukopenia, diarrhoea, abdominal pain, hepatotoxicity and arthralgia.

Two studies (Hanauer 2004; Mowat 2016), reported on serious adverse events (Analysis 1.7), and three studies reported on the number of withdrawn participants as a result of adverse reactions to the treatment (D'Haens 2008; Hanauer 2004; Mowat 2016). About 3% (5/175) of AZA/6‐MP participants and 1% (2/152) of placebo participants experienced serious adverse events during 12 to 36 months of intervention follow‐up (RR 1.78; 95% CI 0.39 to 8.18; 327 participants; 2 studies; I² = 0%; GRADE low certainty evidence). During the same period, almost an equal proportion of AZA/6‐MP (24%; 51/215) and placebo (25%; 49/193) participants were withdrawn from the study due to adverse events (RR 0.90; 95% CI 0.63 to 1.29; 408 participants; 3 studies; I² = 3%; GRADE moderate certainty evidence). A subgroup analysis found no evidence of a difference in withdrawal due to adverse events when AZA and 6‐MP were compared (P = 0.69). A subgroup analysis based on length of follow‐up found no evidence of a difference in withdrawals due to adverse events when measured at 12 months or less or at over 12 months (P = 0.69). Commonly reported serious adverse events included arthralgia, pancreatitis, leucopenia and bowel obstruction. Adverse events leading to withdrawal from the study included abnormal blood results leading to led to temporary discontinuation of treatment in 28% of the participants. However, specific details on reasons for discontinuation were not clearly reported.

Health related quality of life

HRQoL was reported in one study with no difference in IBDQ scores between the two treatments group. However these data were insufficiently reported for inclusion in the analysis (Mowat 2016).

AZA or 6‐MP versus 5‐ASA

A total of five studies compared the efficacy of AZA (2 to 2.5 mg/day) or 6‐MP (50 mg/day) to mesalamine (dose 3 to 4 g/day) (Ardizzone 2004; Hanauer 2004; Herfarth 2006; Reinisch 2010; Savarino 2013). Due to the specific inclusion/exclusion criteria in Reinisch 2010 (Characteristics of included studies), the clinical and endoscopic relapse data from this study were not included in meta‐analyses.

Primary outcome

Clinical relapse

Clinical relapse defined as a clinical recurrence grading score ≥ 2 (Hanauer 2004; Savarino 2013), or a CDAI ≥ 200 (Ardizzone 2004), was reported in four studies. There was low certainty evidence on the efficacy of AZA or 6‐MP for maintaining postoperative clinical remission in comparison to 5‐ASA compounds (Analysis 2.1; summary of findings Table 2). At the end of the 24 month follow‐up, 64% (113/177) of AZA treated participants clinically relapsed compared to 59% (101/170) of 5‐ASA treated ones (RR 1.05, 95% CI 0.89 to 1.24, 347 participants, 4 studies, I² = 8%; GRADE low certainty evidence). We carried out a subgroup analysis and found no evidence of a difference in clinical relapse rates between AZA and 6‐MP (P = 0.18). A subgroup analysis based on length of follow‐up found no evidence of a difference in clinical relapse rates when measured at 12 months or less or at over 12 months (P = 0.97).

Secondary outcomes

Endoscopic relapse

Endoscopic relapse was reported in one study (Savarino 2013), and was defined as a Rugeerts score ≥ 2. The efficacy of AZA in comparison to 5‐ASA formulations in preventing endoscopic relapse was uncertain, as the certainty of evidence was very low (Analysis 2.3; Summary of findings table 2). After 24 months, 65% (11/17) of AZA participants relapsed endoscopically compared to 83% (15/18) of 5‐ASA participants (RR 0.78; 95% CI 0.52 to 1.17; 35 participants; 1 study; GRADE very low certainty evidence).

Radiologic relapse

Radiologic relapse (follow‐up 24 months), defined as radiographic recurrence grading score ≥ 2, was reported in one study (Savarino 2013) and meta‐analysis was not performed. There is a very low certainty evidence regarding the effect of purine analogues on radiologic relapse rate compared to 5‐ASA drugs (Analysis 2.4). Sixty‐four percent (13/17) of AZA participants experienced radiologic relapse compared to 83% (15/18) of 5‐ASA participants (RR 0.92; 95% CI 0.66 to 1.28; 35 participants; 1 study; GRADE very low certainty evidence).

Endoscopic and radiologic relapse rates were reported by Hanauer 2004, however the results were insufficient to be included in the meta‐analysis. At 24 months (participants 91), reported endoscopic and radiologic relapse rates were 16%; 95% CI 7% to 35% and 33%; 95% CI 19% to 54% in the purine antimetabolites compared to 48%; 95% CI 30% to 70% and 46%; 95% CI 29% to 66% in the mesalamine intervention respectively.

Surgical relapse

Surgical relapse (follow‐up 24 months), defined as the need for another surgery was reported in Ardizzone 2004. The effect of purine analogues compared to 5‐ASA for the maintenance of surgical remission was uncertain, because the quality of evidence was judged as very low (Analysis 2.5; summary of findings Table 2). During the follow‐up period of two years, the proportion of participants with surgical relapse was 37% (26/71) in the 5‐ASA group versus 30% (21/71) in the purine analogues group (RR 0.81; 95% CI 0.50 to 1.29; 142 participants; 1 study; GRADE very low certainty evidence).

Adverse events, serious adverse events and withdrawal due to adverse events

Adverse events and withdrawal due to adverse events were reported in four studies (Ardizzone 2004; Hanauer 2004; Reinisch 2010; Savarino 2013), and five studies respectively (Ardizzone 2004; Hanauer 2004; Herfarth 2006; Reinisch 2010; Savarino 2013), while serious adverse events were reported in three studies (Ardizzone 2004; Hanauer 2004; Reinisch 2010). The effect of purine analogues when compared to 5‐ASA drugs on adverse events was uncertain, as the quality of evidence was low (summary of findings Table 2). During a follow‐up of 12 to 24 months, the proportion of participants who experienced at least one adverse event (Analysis 2.6) was 41% (73/176) and 48% (81/170) in the AZA/6‐MP and 5‐ASA groups respectively (RR 0.89; 95% CI 0.74 to 1.07; 346 participants; 4 studies; I² = 15%; GRADE low certainty evidence). We found no evidence of a difference in adverse events when we carried out a subgroup analysis comparing AZA to 6‐MP (P = 0.34). A subgroup analysis based on length of follow‐up found no evidence of a difference in adverse events when measured at 12 months or less or at over 12 months (P = 0.66). Commonly reported adverse events included leukopenia, abdominal pain, nausea, nasopharyngitis, diarrhoea. and headache. During a 12 to 24 months follow‐up, serious adverse events (Analysis 2.8) were experienced by 17% (27/159) of purine analogue participants compared to 4% (6/152) of 5‐ASA participants (RR 3.39; 95% CI 1.26 to 9.13; 311 participants; 3 studies; I² = 9%; GRADE very low certainty evidence). We found no evidence of a difference in serious adverse events when we carried out a subgroup comparing AZA to 6‐MP (P = 1.0). A subgroup analysis based on length of follow‐up found no evidence of a difference in serious adverse events when measured at 12 months or less or at over 12 months (P = 0.19). Commonly reported serious adverse events include postoperative bowel obstruction. The proportion of participants that withdrew from the trial due to an adverse event (Analysis 2.10) during 12 to 24 months follow‐up were 19% (42/218) versus 8% (16/207) in the AZA/6‐MP and 5‐ASA groups respectively (RR 2.21, 95% CI 1.28 to 3.81; participants = 425; studies = 5; I² = 0%; GRADE low certainty evidence). We found no evidence of a difference in withdrawals due to adverse event when AZA and 6‐MP were compared (P = 0.25). A subgroup analysis based on length of follow‐up found no evidence of a difference in withdrawal due to adverse events when measured at 12 months or less or at over 12 months (P = 0.46). Adverse events leading to withdrawal included severe epigastric intolerance, increase in liver function test results, leukopenia and acute pancreatitis.

HRQoL

Two studies with a follow‐up of 12 to 24 months reported on HRQoL based on the IBDQ score (Reinisch 2010; Savarino 2013). The effect of 5‐ASA agents on HRQoL was uncertain as a result of serious limitations due to unclear risk of performance and outcome assessment bias and high risk of performance bias and outcome assessment bias in each study respectively and very serious limitations due to sparse data in both studies. Savarino 2013 reported on the proportion of participants with an IBDQ score > 170 (ranging from 32 to 224), which is regarded as a symptomatic remission score (Analysis 2.12). After 24 months of follow‐up, 12% (2/17) of AZA treated participants reported an IBDQ score >170 compared with 17% (3/18) of 5‐ASA participants (RR 0.71; 95% CI 0.13 to 3.72; 35 participants; 1 study; GRADE very low certainty evidence). Reinisch 2010, assessed HRQoL based on the mean change of IBDQ scores compared to baseline (Analysis 2.13.) At 12 months, the mean IBDQ difference compared to baseline was 9 (SD 17.7) in the AZA group versus 5 (SD 27.4) in the 5‐ASA treated group (MD 4; CI 14.36 to ‐6.36; 78 participants; 1 study; GRADE very low certainty evidence).

AZA or 6‐MP versus anti‐TNF‐α

Three studies comparing AZA to either infliximab (Armuzzi 2013), or adalimumab (Lopez‐Sanroman 2017; Savarino 2013), and one study comparing 6‐mercaptopurine to adalimumab (Scapa 2015), were identified. For all of the studies, the intervention started within 45 days of surgery. In the Lopez‐Sanroman 2017 study all participants were also administered metronidazole (750 mg/day) for the first three months of the study. The total number of assessed participants was 157.

Primary outcome

Clinical relapse

Clinical relapse (Analysis 3.1) was reported in three studies (Armuzzi 2013; Lopez‐Sanroman 2017; Savarino 2013), and was defined as HBI ≥ 2 (Armuzzi 2013), clinical recurrence grading score ≥2 (Savarino 2013) or CDAI score >200 (Lopez‐Sanroman 2017). The certainty of evidence that AZA is inferior in preventing postsurgical clinical relapse compared to anti‐TNF‐α agents was very low (summary of findings Table 3). During a follow‐up of 12 to 24 months, 43% (29/67) of participants treated with AZA clinically relapsed compared to 14% (10/72) of participants in the anti‐TNF‐α group (RR 2.89, 95% CI 1.50 to 5.57, 139 participants, 3 studies, I² = 0%; GRADE very low certainty evidence). The subgroup analysis based on length of follow‐up found no evidence of a difference in clinical relapse when measured at 12 months or less or at over 12 months (P = 0.2).

Secondary outcomes

Endoscopic relapse

Endoscopic relapse was reported in four studies and was defined as Rutgeerts score ≥2. The evidence that purine antimetabolites are less efficient than anti‐TNF‐α for maintaining endoscopic remission was very low (Analysis 3.3; summary of findings Table 3). The proportion of purine analogue participants with endoscopic relapse was 58% (43/74) compared to 26% (22/83) of anti‐TNF‐α participants (RR 3.67; 95% CI 1.05 to 12.81; 157 participants; 4 studies; I² = 64%; GRADE very low certainty evidence). We found no evidence of a difference in endoscopic relapse rates when we carried out a subgroup analysis to compare AZA to 6‐MP (P = 0.72). A subgroup analysis based on length of follow‐up found a quantitative difference in endoscopic relapse when measured at 12 months or less or at over 12 months (P = 0.2).

Radiologic relapse

Radiologic relapse was reported in two studies, defined either as radiographic recurrence grading score ≥ 2 (Savarino 2013), or based on magnetic resonance enterography scores MR2 and MR3 (Lopez‐Sanroman 2017). The certainty of evidence that purine analogues are less efficient to prevent endoscopic relapse than adalimumab or infliximab was very low (Analysis 3.5; summary of findings Table 3). The proportion of participants with endoscopic relapse after 12 to 24 months follow‐up was 69% (39/56) and 38% (23/61) among the purine analogues and anti‐TNF‐α treated groups respectively. Data pooling for Savarino 2013 (RR 1.36, 95% CI 0.94 to 1.98) and Lopez‐Sanroman 2017 (RR 12.24, 95% CI 1.8 to 83.12) was not feasible due to considerable heterogeneity (I² = 85%). The certainty of evidence was rated very low due to high risk of bias and very serious imprecision.

Histologic relapse

Histologic relapse (Analysis 3.6), based on the scoring system modified by Regueiro 2009 was reported only in Armuzzi 2013 and a meta‐analysis was not performed. Histologic relapse during 12 months of follow‐up was detected in 82% (9/11) and 18% (2/11) participants of the AZA and infliximab groups respectively (RR 4.50; 95% CI 1.25 to 16.25).

Adverse events, serious adverse events and withdrawal due to adverse events

Adverse events were reported in two studies (Lopez‐Sanroman 2017; Savarino 2013). Withdrawals due to adverse events were reported in three studies ( Armuzzi 2013; Lopez‐Sanroman 2017; Savarino 2013), and serious adverse events were reported in one study (Lopez‐Sanroman 2017). The certainty of evidence regarding tolerability and safety of purine analogues in comparison to anti‐TNF‐α agents ranged from very low to low (summary of findings Table 3). There was no clear difference in the number of participants who experienced adverse events when AZA/6‐MP was compared to anti‐TNF‐α during the follow‐up of 12 to 24 months (Analysis 3.7). Fifty‐seven per cent (32/56) of AZA/6‐MP participants experienced at least one adverse event compared to 51% (31/61) of the anti‐TNF‐α group (RR 1.13; 95% CI 0.83 to 1.53; 117 participants; 2 studies; I² = 0%; GRADE low certainty evidence). A subgroup analysis based on length of follow‐up found no evidence of a difference in adverse events when measured at 12 months or less or at over 12 months (P = 0.65). Commonly reported adverse events included bronchitis, nasopharyngitis, arthralgia, nausea and abscess (Savarino 2013). Full details concerning adverse events were not reported in Lopez‐Sanroman 2017. The evidence regarding serious adverse events comes from a single study and a meta‐analysis was not performed (Analysis 3.9). The proportion of participants with serious adverse events was 10% (4/39) and 20% (9/45) among the AZA and adalimumab groups respectively (RR 0.51; 95% CI 0.17 to 1.54; 84 participants; 1 study; GRADE very low certainty evidence). The types of serious adverse events were not described. Over a follow‐up that ranged from 12 to 24 months, 16% (11/67) of AZA/6‐MP treated participants were withdrawn from the study due to an adverse event compared to 3% (2/72) of the anti‐TNF‐α treated participants (RR 3.97; 95% CI 0.92 to 17.22; 139 participants; 3 studies; I² = 4%; GRADE low certainty evidence). A subgroup analysis based on length of follow‐up found no evidence of a difference in withdrawals due to adverse events when measured at 12 months or less or at over 12 months (P = 0.24). Adverse events leading to withdrawal included severe nausea, leukopenia, arthralgia, urothelial carcinoma, dyspepsia, dyspnoea, death, atopic dermatitis and abdominal pain with increase in pancreatic enzymes.

Health related quality of life

A single study assessed health‐related quality of life (Savarino 2013). At 24 months, an IBDQ > 170 was recorded in 12% (2/17) of participants treated with AZA compared to 88% (14/16) treated with adalimumab (RR 0.13; 95% CI 0.04 to 0.50; 34 participants; 1 study). This outcome was also evaluated in Lopez‐Sanroman 2017 using the EuroQol‐5 dimensions questionnaire, however reported data were insufficient to be included in meta‐analysis.

Special safety note

It is worth specifically noting that two adverse events leading to therapy cessation occurred in more than 1% of participants treated with purine analogues. Pancreatitis was a serious adverse event that led to withdrawal across several of the studies. However, the rate of occurrence was almost exclusively in participants receiving purine analogues (11 cases in 354 participants taking purine analogues, 3.1%) compared to no occurrences in any of the comparison groups (i.e. placebo, 5‐ASA or anti‐TNF‐α) in this review. Leucopenia occurred also almost exclusively in participants receiving purine analogues (9 cases in 354 participants, 2.5%) compared to no occurrences in all other groups.

Discussion

Summary of main results

Ten RCTs assessing the effectiveness of AZA and 6‐MP for maintaining surgically‐induced remission in Crohn's disease were studied. The studies recruited people with Crohn's disease over the age of 16 years who were in surgically‐induced remission. The participants were randomised after surgery to received AZA or 6‐MP, placebo, 5‐ASA agents or anti‐TNF‐α agents.

We found moderate certainty evidence that AZA and 6‐MP are superior to placebo for preventing post‐surgical clinical relapse in Crohn's disease. There was low certainty evidence on the safety of AZA and 6‐MP when adverse events and serious adverse events were considered. The certainty of the evidence for withdrawal due to adverse events was moderate. This was due to imprecision resulting from very sparse data.

There was no clear difference clinical relapse when AZA or 6‐MP was compared with 5‐ASA. There are considerable concerns raised with the safety profile of AZA and 6‐MP. For the safety outcome serious adverse events and withdrawals due to adverse events, 5‐ASA was found to be superior to AZA and 6‐MP. However, the certainty of evidence for the efficacy and safety outcomes were low due to high risk of performance bias and sparse data.

There was very low certainty evidence that AZA and 6‐MP may be inferior to anti‐TNF‐α agents for preventing clinical relapse. We also found no clear difference in adverse events, serious adverse events or withdrawals due to adverse events when both outcomes were compared. This was also based on low certainty evidence. The evidence was downgraded due to high risk of performance bias and sparse data, therefore, the results should be interpreted with caution.

Overall completeness and applicability of evidence

We consider the evidence from this review to be applicable to most patients with post‐surgical remission of Crohn's disease. The evidence only assesses AZA or 6‐MP compared with placebo, 5‐ASA or anti‐TNF‐α agents. All outcomes which we had aimed to analyse were reported in the studies. However, other measures of relapse (endoscopic, histologic and surgical) and health related quality of life were sparsely reported. This meant that very little data were available for our analysis. The review found moderate certainty evidence on the efficacy of AZA and 6‐MP compared with placebo. This can be considered complete and not requiring new studies, however, the evidence comparing it with 5‐ASA and anti‐TNF‐α were of very low or low certainty. Therefore, additional studies may change the results. The use of concomitant treatments such as antidiarrhoeal agents, corticosteroids and antibiotics were reported in some studies. The circumstances in which these treatments were given were noted in the studies and judged to be reflective of clinical practice.

Quality of the evidence

The certainty of the evidence was rated very low to moderate. Overall, six studies were at high risk of bias and three were at unclear risk of bias. Downgrading for limitations was mostly due to lack of blinding and this was particularly common with the head to head comparisons of active drugs. For most of the outcomes, there was imprecision due to sparse data as the number of events ranged between seven and 154. For the clinical relapse outcome, the optimal information size was obtained from power calculations in the largest and most recent study (Mowat 2016). Most of the results were consistent except in two instances where there was substantial statistical heterogeneity (I² between 62% and 64%). There was no indirectness as all the studies in the review met the criteria proposed in the scope of the review. We were unable to assess for publication bias due to insufficient data.

Potential biases in the review process

We attempted to reduce potential biases in the review process. A comprehensive literature search was performed to identify all eligible studies. Two review authors independently assessed studies for inclusion, extracted data and assessed study quality.

All analyses were completed using the intention‐to‐treat principle, whereby participants with final missing outcomes were assumed to have relapsed. Given the high attrition rate in the purine analogue groups compared to the 5‐ASA groups, this may have affected the difference in clinical relapse rates between purine analogues and 5‐ASA. However, it is arguably a moot point given that even if purine analogues did have superior efficacy, it is difficult to rationalise the use on the basis of the poor adverse event profile in the published evidence.

One of the included studies administered active concomitant treatments in both intervention arms. This was considered a source of clinical heterogeneity in analyses that included studies which had no concomitant treatments. We did not remove this study from the analyses as a sensitivity analysis showed no difference in the results.

Agreements and disagreements with other studies or reviews

A recent Cochrane review assessing the use of AZA or 6‐MP for maintenance of medically‐induced remission in Crohn's disease revealed that the purine analogues are more effective than placebo, with higher response rates for AZA than 6‐MP (Chande 2015). These findings are mirrored in the four studies comparing purine analogues to placebo. No difference in efficacy was found between AZA or 6‐MP and 5‐ASA. This could be due to lower disease activity following resection of the gut than is achieved in medically‐induced remission of Crohn's disease, so that a milder anti‐inflammatory agent such as 5‐ASA, gives a better risk versus benefit ratio when compared to AZA and 6‐MP. It is also possible that the methodology of the included studies supports this hypothesis, with all but one study recruiting participants in the immediate post‐surgical setting. As such, the participants are potentially at their lowest period of disease activity clinically and microscopically. The findings regarding safety are also consistent with the two most common adverse events noted as pancreatitis and leucopenia.

A Cochrane review looking at the use of 5‐ASA for the maintenance of surgically‐induced remission in Crohn's disease suggests that 5‐ASA may be superior to placebo (Gjuladin‐Hellon 2019). It also suggested that 5‐ASA is a safe and well‐tolerated drug, as the incidence of adverse events was not different in participants receiving 5‐ASA compared to those receiving placebo. The results of this systematic review question the risk versus benefit balance of starting a purine analogue over 5‐ASA in postoperative Crohn's disease.

The international guidance from the European Crohn's and colitis Organisation (ECCO) updated in 2016 is consistent with the findings of this review in that they do indicate a role for purine analogues (Gionchetti 2016), although they do not comment on the limitations of the evidence as identified in this review. Most importantly, the significant safety questions raised in the synthesis performed in this review are not discussed within the ECCO documents in this context (section 8G), although leucopenia is noted in medical induced remission maintenance (6E). The National Insitutue for Clinical and Care Excellence (NICE) in the UK suggests AZA as first line therapy in this context (NICE 2016), citing the previous version of this review in their 2016 guidance (Gordon 2014). Unlike ECCO, the NICE guidance makes no specific mention of safety concerns. Given one of the specific adverse events noted frequently (leucopenia) is recognised in the context of purine analogues, it is important to highlight that the risk of pancreatitis has not been highlighted in either guideline, despite being recognised as the most common serious adverse events in both Cochrane reviews on these medications.

Figure 1

Study flow diagram.

Figure 2

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

Figure 3

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

Analysis 1.1

Comparison 1 Azathioprine or 6‐mercaptopurine versus placebo, Outcome 1 Clinical relapse at 12 to 36 months (subgroup by drug type).

Analysis 1.2

Comparison 1 Azathioprine or 6‐mercaptopurine versus placebo, Outcome 2 Clinical relapse (subgroup by length of follow‐up).

Analysis 1.3

Comparison 1 Azathioprine or 6‐mercaptopurine versus placebo, Outcome 3 Endoscopic relapse at 12 to 36 months (subgroup by drug type).

Analysis 1.4

Comparison 1 Azathioprine or 6‐mercaptopurine versus placebo, Outcome 4 Endoscopic relapse (subgroup by length of follow‐up).

Analysis 1.5

Comparison 1 Azathioprine or 6‐mercaptopurine versus placebo, Outcome 5 Adverse events at 12 to 24 months (subgroup by drug type).

Analysis 1.6

Comparison 1 Azathioprine or 6‐mercaptopurine versus placebo, Outcome 6 Adverse events (subgroup by length of follow‐up).

Analysis 1.7

Comparison 1 Azathioprine or 6‐mercaptopurine versus placebo, Outcome 7 Serious adverse events at 24 to 36 months.

Analysis 1.8

Comparison 1 Azathioprine or 6‐mercaptopurine versus placebo, Outcome 8 Withdrawal due to adverse events at 12 to 36 months (subgroup by drug type).

Analysis 1.9

Comparison 1 Azathioprine or 6‐mercaptopurine versus placebo, Outcome 9 Withdrawal due to adverse events (subgroup by length of follow‐up).

Analysis 2.1

Comparison 2 Azathioprine or 6‐mercaptopurine versus 5‐aminosalicylic acid, Outcome 1 Clinical relapse at 12 to 24 months (subgroup by drug type).

Analysis 2.2

Comparison 2 Azathioprine or 6‐mercaptopurine versus 5‐aminosalicylic acid, Outcome 2 Clinical relapse (subgroup by length of follow‐up).

Analysis 2.3

Comparison 2 Azathioprine or 6‐mercaptopurine versus 5‐aminosalicylic acid, Outcome 3 Endoscopic relapse at 24 months.

Analysis 2.4

Comparison 2 Azathioprine or 6‐mercaptopurine versus 5‐aminosalicylic acid, Outcome 4 Radiologic relapse at 24 months.

Analysis 2.5

Comparison 2 Azathioprine or 6‐mercaptopurine versus 5‐aminosalicylic acid, Outcome 5 Surgical relapse at 24 months.

Analysis 2.6

Comparison 2 Azathioprine or 6‐mercaptopurine versus 5‐aminosalicylic acid, Outcome 6 Adverse events at 12 to 24 months (subgroup by drug type).

Analysis 2.7

Comparison 2 Azathioprine or 6‐mercaptopurine versus 5‐aminosalicylic acid, Outcome 7 Adverse events (subgroup by length of follow‐up).

Analysis 2.8

Comparison 2 Azathioprine or 6‐mercaptopurine versus 5‐aminosalicylic acid, Outcome 8 Serious adverse events at 12 to 24 months (subgroup by drug type).

Analysis 2.9

Comparison 2 Azathioprine or 6‐mercaptopurine versus 5‐aminosalicylic acid, Outcome 9 Serious adverse events (subgroup by length of follow‐up).

Analysis 2.10

Comparison 2 Azathioprine or 6‐mercaptopurine versus 5‐aminosalicylic acid, Outcome 10 Withdrawal due to adverse events at 12 to 24 months (subgroup by drug type).

Analysis 2.11

Comparison 2 Azathioprine or 6‐mercaptopurine versus 5‐aminosalicylic acid, Outcome 11 Withdrawal due to adverse events (subgroup by length of follow‐up).

Analysis 2.12

Comparison 2 Azathioprine or 6‐mercaptopurine versus 5‐aminosalicylic acid, Outcome 12 HRQoL ‐ IBDQ > 170 at 24 months.

Analysis 2.13

Comparison 2 Azathioprine or 6‐mercaptopurine versus 5‐aminosalicylic acid, Outcome 13 HRQoL ‐ IBDQ at 12 months.

Analysis 3.1

Comparison 3 Azathioprine or 6‐mercaptopurine versus anti TNF‐α, Outcome 1 Clinical relapse at 12 to 24 months.

Analysis 3.2

Comparison 3 Azathioprine or 6‐mercaptopurine versus anti TNF‐α, Outcome 2 Clinical relapse (subgroup by length of follow‐up).

Analysis 3.3

Comparison 3 Azathioprine or 6‐mercaptopurine versus anti TNF‐α, Outcome 3 Endoscopic relapse at 12 to 24 months (subgroup by drug type).

Analysis 3.4

Comparison 3 Azathioprine or 6‐mercaptopurine versus anti TNF‐α, Outcome 4 Endoscopic relapse (subgroup by length of follow‐up).

Analysis 3.5

Comparison 3 Azathioprine or 6‐mercaptopurine versus anti TNF‐α, Outcome 5 Radiologic relapse at 12 to 24 months.

Analysis 3.6

Comparison 3 Azathioprine or 6‐mercaptopurine versus anti TNF‐α, Outcome 6 Histologic relapse at 12 months.

Analysis 3.7

Comparison 3 Azathioprine or 6‐mercaptopurine versus anti TNF‐α, Outcome 7 Adverse events at 12 to 24 months.

Analysis 3.8

Comparison 3 Azathioprine or 6‐mercaptopurine versus anti TNF‐α, Outcome 8 Adverse events (subgroup by length of follow‐up).

Analysis 3.9

Comparison 3 Azathioprine or 6‐mercaptopurine versus anti TNF‐α, Outcome 9 Serious adverse events at 12 months.

Analysis 3.10

Comparison 3 Azathioprine or 6‐mercaptopurine versus anti TNF‐α, Outcome 10 Withdrawal due to adverse events at 12 to 24 months.

Analysis 3.11

Comparison 3 Azathioprine or 6‐mercaptopurine versus anti TNF‐α, Outcome 11 Withdrawal due to adverse events (subgroup by length of follow‐up).

Analysis 3.12

Comparison 3 Azathioprine or 6‐mercaptopurine versus anti TNF‐α, Outcome 12 HRQoL ‐ IBDQ > 170 at 24 months.

Summary of findings for the main comparison. Azathioprine or 6‐mercaptopurine compared to placebo for maintenance of surgically‐induced remission in Crohn's disease

Azathioprine or 6‐mercaptopurine compared to placebo for maintenance of surgically‐induced remission in Crohn's disease
Patient or population: People with surgically‐induced remission in Crohn's disease Setting: Outpatient Intervention: Azathioprine (100‐150 mg/day) or 6‐mercaptopurine (1 mg/kg/day ‐ 50 mg/day) 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 Azathioprine or 6‐mercaptopurine
Clinical relapse Follow‐up: 12 to 36 months	Study population		RR 0.79 (0.67 to 0.92)	408 (3 RCTs)	⊕⊕⊕⊝ MODERATE¹	Clinical relapse defined as: a CDAI>250 (D'Haens 2008); a CDAI>150 and 100 point increase from baseline (Mowat 2016) or a grading score > 2 (Hanauer 2004).
	642 per 1,000	508 per 1,000 (430 to 591)
Endoscopic relapse Follow‐up:12 to 36 months	Study population		RR 0.85 (0.64 to 1.13)	321 (2 RCTs)	⊕⊕⊝⊝ LOW ^{2 3}	Endoscopic relapse defined as an endoscopic score i ≥2 (D'Haens 2008; Mowat 2016)
	752 per 1,000	639 per 1,000 (481 to 849)
Radiologic relapse	Outcome not reported					Not reported
Surgical relapse	Outcome not reported					Not reported
Adverse events Follow‐up: 12 to 24 months	Study population		RR 1.36 (0.57 to 3.27)	168 (2 RCTs)	⊕⊕⊝⊝ LOW ⁴	Reported adverse events include hair loss, leukopenia, diarrhoea, abdominal pain, hepatotoxicity and arthralgia
	99 per 1,000	134 per 1,000 (56 to 323)
Serious adverse events Follow‐up:12 to 36 months	Study population		RR 1.78 (0.39 to 8.18)	327 (2 RCTs)	⊕⊕⊝⊝ LOW ⁵	Reported serious adverse events include arthralgia, pancreatitis, leucopenia and bowel obstruction
	13 per 1,000	23 per 1,000 (5 to 108)
Withdrawal due to adverse events Follow‐up:12 to 36 months	Study population		RR 0.90 (0.63 to 1.29)	408 (3 RCTs)	⊕⊕⊕⊝ MODERATE⁶	Adverse events leading to withdrawal included abnormal blood results leading to temporary discontinuation of treatment in 28% of the participants. However, specific details on reasons for discontinuation were not clearly stated
	254 per 1,000	228 per 1,000 (160 to 328)
*The risk in the intervention group (and its 95% confidence interval) is based on the median risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio; OR: Odds ratio;
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
¹ Downgraded one level due serious imprecision (233 events). ² Downgraded one level due to serious inconsistency ( I² = 64%). ³ Downgraded one level due to serious imprecision (227 events). ⁴ Downgraded two levels due to very serious imprecision (20 events) and 95% CI which includes appreciable benefit and harm ⁵ Downgraded two levels due to very serious imprecision (7 events) and 95% CI which includes appreciable benefit and harm ⁶ Downgraded one level due to serious imprecision (100 events) and 95% CI which includes appreciable benefit and harm

Summary of findings for the main comparison. Azathioprine or 6‐mercaptopurine compared to placebo for maintenance of surgically‐induced remission in Crohn's disease

Summary of findings 2. Azathioprine or 6‐mercaptopurine compared to 5‐aminosalicylic acid for maintenance of surgically‐induced remission in Crohn's disease

Azathioprine or 6‐mercaptopurine compared to 5‐aminosalicylic acid for maintenance of surgically‐induced remission in Crohn's disease
Patient or population: People with surgically‐induced remission in Crohn's disease Setting: Outpatient Intervention: Azathioprine (2 mg/kg/day) or 6‐mercaptopurine (50 mg/day) Comparison: 5‐aminosalicylic acid
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with 5‐aminosalicylic acid	Risk with Azathioprine or 6‐mercaptopurine
Clinical relapse Follow‐up: 12 to 24 months	Study population		RR 1.05 (0.89, 1.24)	347 (4 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	Clinical relapse defined as a clinical grading score ≥ 2 (Hanauer 2004;Savarino 2013) or CDAI ≥ 200 (Ardizzone 2004)
	556 per 1,000	595 per 1,000 (456 to 773)
Endoscopic relapse Follow‐up: 24 months	Study population		RR 0.78 (0.52 to 1.17)	35 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{1 3}	Endoscopic relapse defined as a Rugeerts score ≥ 2 (Savarino 2013)
	833 per 1,000	650 per 1,000 (433 to 975)
Radiologic relapse Follow‐up: 24 months	Study population		RR 0.92 (0.66 to 1.28)	35 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{1 4}	Radiologic relapse defined as a radiographic grading score ≥ 2 (Savarino 2013)
	833 per 1,000	767 per 1,000 (550 to 1,000)
Surgical relapse Follow‐up: 24 months	Study population		RR 0.81 (0.50 to 1.29)	142 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{1 5}	Surgical relapse defined as a need for another surgery (Ardizzone 2004)
	366 per 1,000	297 per 1,000 (183 to 472)
Adverse events Follow‐up: 12 to 24 months	Study population		RR 0.89 (0.74 to 1.07)	346 (4 RCTs)	⊕⊕⊝⊝ LOW ^{1 6}	Reported adverse events include leukopenia, abdominal pain, nausea, nasopharyngitis, diarrhoea. and headache
	476 per 1,000	424 per 1,000 (353 to 510)
Serious adverse events Follow‐up: 12 to 24 months	Study population		RR 3.39 (1.26 to 9.13)	311 (3 RCTs)	⊕⊝⊝⊝ VERY LOW ^{1 7}	Reported serious adverse events include postoperative bowel obstruction
	39 per 1,000	134 per 1,000 (50 to 360)
Withdrawal due to adverse events Follow‐up: 12 to 24 months	Study population		RR 2.21 (1.28, 3.81)	425 (4 RCTs)	⊕⊕⊝⊝ LOW ^{1 8}	Adverse events leading to withdrawal include Severe epigastric intolerance, increase in liver function test results, leukopenia, acute pancreatitis
	76 per 1,000	172 per 1,000 (93 to 317)
*The risk in the intervention group (and its 95% confidence interval) is based on the median risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio; OR: Odds ratio;
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
¹ Downgraded one level due to high risk of bias ² Downgraded one level due to serious imprecision (214 events) and 95% CI which includes no effect and appreciable harm ³ Downgraded two levels due to very serious imprecision (26 events) and 95% CI which includes no effect and appreciable benefit ⁴ Downgraded two levels due to very serious imprecision (28 events) and 95% CI which includes no effect and appreciable benefit ⁵ Downgraded two levels due to very serious imprecision (47 events) and 95% CI which includes appreciable benefit and harm ⁶ Downgraded one level due to serious imprecision (154 events) and 95% CI which includes no effect and appreciable benefit ⁷ Downgraded two levels due to serious imprecision (33 events) ⁸ Downgraded one level due to serious imprecision (58 events)

Summary of findings 2. Azathioprine or 6‐mercaptopurine compared to 5‐aminosalicylic acid for maintenance of surgically‐induced remission in Crohn's disease

Summary of findings 3. Azathioprine or 6‐mercaptopurine compared to anti TNF‐α for maintenance of surgically‐induced remission in Crohn's disease

Azathioprine or 6‐mercaptopurine compared to anti‐TNF‐α for maintenance of surgically‐induced remission in Crohn's disease
Patient or population: People with surgically‐induced remission in Crohn's disease Setting: Outpatient Intervention: Azathioprine (2‐2.5 mg/kg/day) or 6‐mercaptopurine (1.5 mg/kg/day) Comparison:Anti‐TNF‐α
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with anti‐ TNF‐α	Risk with Azathioprine or 6‐mercaptopurine
Clinical relapse Follow‐up: 12 to 24 months	Study population		RR 2.89 (1.50 to 5.57)	139 (3 RCTs)	⊕⊕⊝⊝ VERY LOW ^{1 2}	Clinical relapse defined as: an HBI ≥2 (Armuzzi 2013), a clinical recurrence grading score ≥2 (Savarino 2013) or a CDAI score >200 (Lopez‐Sanroman 2017)
	139 per 1,000	401 per 1,000 (208 to 774)
Endoscopic relapse Follow‐up: 12 to 24 months	Study population		RR 3.67 (1.05 to 12.81)	157 (4 RCTs)	⊕⊝⊝⊝ VERY LOW ^{1 3 4}	Endoscopic relapse defined as a Rugeerts score ≥ 2 (Armuzzi 2013; Lopez‐Sanroman 2017; Savarino 2013; Scapa 2015)
	265 per 1,000	973 per 1,000 (278 to 1,000)
Radiologic relapse Follow‐up: 12 to 24 months	Study population		‐	117 (2 RCTs)	‐	Radiologic relapse defined as a radiographic grading score ≥ 2 (Savarino 2013) or magnetic resonance enterography score ≥2 (Lopez‐Sanroman 2017) *due to considerable heterogeneity (I² = 85%), data pooling was not feasible. It is uncertain whether azathioprine leads to a difference in radiologic relapse when compared to infliximab as the certainty of the evidence is very low (RR1.36, 95% CI 0.94 to 1.98; RR 12.24, 95% CI 1.8 to 83.12)
	see comment	see comment
Surgical relapse	Outcome not reported					Not reported
Adverse events Follow‐up: 12 to 24 months	Study population		RR 1.13 (0.83 to 1.53)	117 (2 RCTs)	⊕⊕⊝⊝ LOW ^{1 5}	Adverse events include bronchitis, nasopharyngitis, arthralgia, nausea, abscess were reported in Savarino 2013. Full details were not reported in Lopez‐Sanroman 2017
	508 per 1,000	574 per 1,000 (422 to 778)
Serious adverse events Follow‐up: 12 months	Study population		RR 0.51 (0.17 to 1.54)	84 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{1 6}	Serious adverse events not reported
	200 per 1,000	102 per 1,000 (34 to 308)
Withdrawal due to adverse events Follow‐up: 12 to 24 months	Study population		RR 3.97 (0.92 to 17.22)	139 (3 RCTs)	⊕⊕⊝⊝ LOW ^{1 6}	Adverse events leading to withdrawal included severe nausea, leukopenia, arthralgia, urothelial carcinoma, dyspepsia, dyspnoea, death, atopic dermatitis and abdominal pain with increase in pancreatic enzymes
	28 per 1,000	110 per 1,000 (26 to 478)
*The risk in the intervention group (and its 95% confidence interval) is based on the median risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio; OR: Odds ratio;
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
¹ Downgraded one level due to high risk of bias ² Downgraded two levels due to very serious imprecision (39 events) ³ Downgraded one level due to serious imprecision (65 events) ⁴ Downgraded one level due to substantial heterogeneity (I² = 64%) ⁵ Downgraded one level due to serious imprecision (63 events) ⁶ Downgraded two levels due to very serious imprecision (13 events)

Summary of findings 3. Azathioprine or 6‐mercaptopurine compared to anti TNF‐α for maintenance of surgically‐induced remission in Crohn's disease

Table 1. Key definitions and outcomes

Comparison	Study ID	Time from surgery till recruitment	Site of surgery % / exclusions	Clinical relapse definition	Endoscopic / surgical/ radiologic/ histological relapse definition/
AZA & 6‐MP versus Placebo
AZA versus Placebo (12 months) both arms 750 mg/day metronidazole (3 months)	D'Haens 2008	2 weeks	Perforating disease 48 *Macroscopic evidence for CD proximally or distally to the site of resection or the presence of frank pancolitis or an ileorectal anastomosis, patients with a stoma; operation for fibrostenosis only	CDAI > 250	Rutgeerts i ≥ 2
6‐MP 50 mg/day versus Placebo (24 months)	Hanauer 2004	Before postoperative hospital discharge	N/A * Active perianal disease or any active disease in other segments of the intestine	Clinical recurrence grading > 2 (Hanauer)	Rutgeerts i ≥ 2 Radiographic relapse: Radiographic recurrence grading > 2
6‐MP 1 mg/kg/day versus Placebo (3 years)	Mowat 2016	≤ 3 months	Ileal 39; Colonic 2; Ileocolonic 59 * Need for further surgery, stricturoplasty alone, formation of a stoma	CDAI > 150 and a 100‐point increase from baseline	Rutgeerts i ≥ 2 HRQOL: IBDQ scores
AZA & 6‐MP vs 5‐ASA
AZA 2 mg/kg versus Mesalamine 3 mg/kg (24 months)	Ardizzone 2004	Maximum 2 weeks	Small bowel only 25.3; Colon 5.6; Small bowel and colon 9.8; upper gastrointestinal tract 16.2 *Surgical procedures other than conservative surgery or for perianal disease only	CDAI > 200 Surgical relapse: need for another surgical procedure	n/a
6‐MP 50 mg/day versus Mesalamine 3 g/day (24 months)	Hanauer 2004	Before postoperative hospital discharge	N/A * Active perianal disease or any active disease in other segments of the intestine	clinical recurrence grading > 2	Rutgeerts i ≥ 2 Radiographic relapse: radiographic recurrence grading > 2
AZA 2 mg/kg/day versus Mesalamine 4 g/day (12 months)	Reinisch 2010	6‐24 months	N/A * Short bowel syndrome, an ileocolonic stoma	CDAI > 200	Rutgeerts i ≥ 2 HRQOL: IBDQ
AZA 2 mg/kg/day versus Mesalazine 3 g/day (24 months)	Savarino 2013	2‐4 weeks	Ileum 49, Ileocolonic 51. * Fibrostenotic stricture, macroscopically active disease not resected at the time of surgery, and presence of a stoma	1. ≥ 2 on the clinical recurrence grading scale by Hanauer 2. CDAI > 200	Rutgeerts i≥2 Radiologic relapse : ≥ 2 radiographic recurrence grading scale HRQl: IBDQ>170
AZA & 6‐MP vs anti‐TNF‐α
AZA 2.5 mg versus Infliximab 5 mg/kg (12 months)	Armuzzi 2013	2‐4 weeks	Not reported *Active perianal disease, presence of stoma	HBI ≥ 8	Rutgeerts' score ≥ i2
AZA 2 mg/kg/day versus Adalimumab (24 months)	Savarino 2013	2‐4 weeks	Ileum 49, Ileocolonic 51. * Fibrostenotic stricture, macroscopically active disease not resected at the time of surgery, and presence of a stoma	1. ≥ 2 on the clinical recurrence grading scale by Hanauer 2. CDAI > 200	Rutgeerts i ≥ 2 Radiologic relapse : ≥ 2 radiographic recurrence grading scale HRQl: IBDQ > 170
AZA 2.5 mg/kg/day versus Adalimumab (52 weeks) both arms 750mg/day metronidazole (3 months)	Lopez‐Sanroman 2017	2 weeks	Ileal 58, ileocolonic 41 * Postsurgical stoma, resection for short indolent stenosis, inaccessible anastomosis to endoscopy	CDAI > 200	Rutgeerts i ≥ 2
6‐MP 1.5 mg/kg/day versus Adalimumab (12 months)	Scapa 2015	< 45 days	6‐MP 1.5 mg/kg/day vs Placebo (12 months)	Scapa 2015	< 45 days
AZA: azathioprine; 6‐MP: 6‐mecarptopurine; mg: milligram; CD: Crohn's disease; CDAI: Crohn's disease activity index; NA: not applicable; kg: kilogram; g: gram; HRQOL: health related quality of life; IBDQ: inflammatory bowel disease questionnaire; TNF: tumour necrosis factor; HBI: Harvey Bradshaw index; 5‐ASA: 5‐aminosalicylic acid

Table 1. Key definitions and outcomes

Table 2. Summary of interventions and outcomes

Study ID	Group 1	Group 2	Group 3	Relapse	Quailty of Life	Adverse Events/ Serious adverse/ Withdrawal due to adverse events
Ardizzone 2004	Azathioprine (2mg/kg/day)	Mesalamine (3g/day)		Clinical: 32/71 vs 35/71 Surgical: 26/71 vs 21/71	n/a	AE:18/71 vs 27/71 SAE:6/71 vs 15/71 Withdrwal due to AE: 6/71 vs 15/71
Armuzzi 2013	Azathioprine (2.5 mg/kg/day) Infliximab (5 mg/kg/day)	Infliximab (5 mg/kg/day)		Clinical: 2/11 vs 1/11 Endoscopic: 5/11 vs 1/11; Histologic: 9/11 vs 2/11	n/a	Withdrawal due to AE: 0/11 vs 1/11
D'Haens 2008	Metronidazole (750 mg/day) first 3 mo + Azathioprine (100‐150 mg/day)	Metronidazole (750mg/day) first 3 months + Placebo		Clinical: 11/40 vs 19/41 Endoscopic: 22/40 vs 32/41	n/a	AE: 3/40 vs 4/41 Withdrawal due to AE: 3/40 vs 4/41
Hanauer 2004	6‐Mercaptopurine (50 mg/day)	Mesalamine (3 g/day)	Placebo	Clinical: 32/47 vs 33/44 vs 35/40	n/a	AE: 9/47 vs 6/44 vs 4/40 SAE: 2/47 vs 0/44 vs 2/40 Withdrawal due to AE: 9/47 vs 6/44 vs 4/40
Lopez‐Sanroman 2017	Azathioprine (2.5 mg/kg/d) + Metronidazole (750 mg/day) first 3 mo	Adalimumab + Metronidazole (750 mg/day) first 3 mo		Clinical: 14/39 vs 7/45 Endoscopic: 23/39 vs 19/45 Radiologic: 26/39 vs 22/45	n.s. changes between groups	AE: 20/45 vs 18/39 SAE: 9/45 vs 4/39 Withdrawal due to AE: 1/39 vs 9/45
Mowat 2016	Mercaptopurine (1 mg/kg/day)	Placebo		Clinical: 66/128 vs 70/112 Endoscopic: 90/128 vs 83/112	n.s differences	SAE: 3/128 vs 2/112 Withdrawal due to AE: 39/128 vs 41/112
Reinisch 2010	Azathioprine (2.0‐2.5 mg/kg/d) + Placebo mesalazine	Mesalazine (4g/d) + Placebo azathioprine		Not included	Mean IBDQ change	AE: 34/37 vs 32/41 SAE: 0/37 vs 10/41 Withdrawal due to AE: 1/37 vs 10/41
Savarino 2013	Adalimumab (160‐80 mg 0‐2 weeks and 40 mg/week thereafter)	Azathioprine (2 mg/kg/day)	Mesalamine (3 g/day)	Clinical by Hanauer score: 2/16 vs 12/17 vs 9/18 Clinical by CDAI: 1/16 vs 12/17 vs 9/18 Endoscopic: 1/16 vs 11/17 vs 15/18 Radiologic: 1/16 vs 13/17 vs 15/18	HRQOL (IBDQ >170): 14/16 vs 2/17 vs 3/18	AE:11/16 vs 14/17 vs 14/18 Withdrawal due to AE:0/16 vs 1/17 vs 1/18
Scapa 2015	6‐mercaptopurine (1.5 mg/kg/day)	Adalimumab (160‐80‐40 mg/2 week intervals)		Endoscopic: 4/8 vs 1/11	n/a	n/a
mg: milligram; kg: kilogram; g: gram; CDAI: Crohn's disease activity index; NA: not applicable; AE: adverse events; SAE: serious adverse events; HRQOL: health related quality of life; IBDQ: inflammatory bowel disease questionnaire; ns: not significant

Table 2. Summary of interventions and outcomes

Comparison 1. Azathioprine or 6‐mercaptopurine versus placebo

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Clinical relapse at 12 to 36 months (subgroup by drug type) Show forest plot	3	408	Risk Ratio (M‐H, Random, 95% CI)	0.79 [0.67, 0.92]

1.1 Azathioprine	1	81	Risk Ratio (M‐H, Random, 95% CI)	0.59 [0.33, 1.08]
1.2 6‐mercaptopurine	2	327	Risk Ratio (M‐H, Random, 95% CI)	0.80 [0.68, 0.94]
2 Clinical relapse (subgroup by length of follow‐up) Show forest plot	3	408	Risk Ratio (M‐H, Random, 95% CI)	0.79 [0.67, 0.92]

2.1 12 months or less	1	81	Risk Ratio (M‐H, Random, 95% CI)	0.59 [0.33, 1.08]
2.2 Over 12 months	2	327	Risk Ratio (M‐H, Random, 95% CI)	0.80 [0.68, 0.94]
3 Endoscopic relapse at 12 to 36 months (subgroup by drug type) Show forest plot	2	321	Risk Ratio (M‐H, Random, 95% CI)	0.85 [0.64, 1.13]

3.1 Azathioprine	1	81	Risk Ratio (M‐H, Random, 95% CI)	0.70 [0.51, 0.97]
3.2 6‐mercaptopurine	1	240	Risk Ratio (M‐H, Random, 95% CI)	0.95 [0.81, 1.11]
4 Endoscopic relapse (subgroup by length of follow‐up) Show forest plot	2	321	Risk Ratio (M‐H, Random, 95% CI)	0.85 [0.64, 1.13]

4.1 12 months or less	1	81	Risk Ratio (M‐H, Random, 95% CI)	0.70 [0.51, 0.97]
4.2 Over 12 months	1	240	Risk Ratio (M‐H, Random, 95% CI)	0.95 [0.81, 1.11]
5 Adverse events at 12 to 24 months (subgroup by drug type) Show forest plot	2	168	Risk Ratio (M‐H, Random, 95% CI)	1.36 [0.57, 3.27]

5.1 Azathioprine	1	81	Risk Ratio (M‐H, Random, 95% CI)	0.77 [0.18, 3.22]
5.2 6‐mercaptopurine	1	87	Risk Ratio (M‐H, Random, 95% CI)	1.91 [0.64, 5.75]
6 Adverse events (subgroup by length of follow‐up) Show forest plot	2	168	Risk Ratio (M‐H, Random, 95% CI)	1.36 [0.57, 3.27]

6.1 12 months or less	1	81	Risk Ratio (M‐H, Random, 95% CI)	0.77 [0.18, 3.22]
6.2 Over 12 months	1	87	Risk Ratio (M‐H, Random, 95% CI)	1.91 [0.64, 5.75]
7 Serious adverse events at 24 to 36 months Show forest plot	2	327	Risk Ratio (M‐H, Random, 95% CI)	1.78 [0.39, 8.18]

8 Withdrawal due to adverse events at 12 to 36 months (subgroup by drug type) Show forest plot	3	408	Risk Ratio (M‐H, Random, 95% CI)	0.90 [0.63, 1.29]

8.1 Azathioprine	1	81	Risk Ratio (M‐H, Random, 95% CI)	0.77 [0.18, 3.22]
8.2 6‐mercaptopurine	2	327	Risk Ratio (M‐H, Random, 95% CI)	1.07 [0.50, 2.27]
9 Withdrawal due to adverse events (subgroup by length of follow‐up) Show forest plot	3	408	Risk Ratio (M‐H, Random, 95% CI)	0.90 [0.63, 1.29]

9.1 12 months or less	1	81	Risk Ratio (M‐H, Random, 95% CI)	0.77 [0.18, 3.22]
9.2 Over 12 months	2	327	Risk Ratio (M‐H, Random, 95% CI)	1.07 [0.50, 2.27]

Comparison 1. Azathioprine or 6‐mercaptopurine versus placebo

Comparison 2. Azathioprine or 6‐mercaptopurine versus 5‐aminosalicylic acid

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Clinical relapse at 12 to 24 months (subgroup by drug type) Show forest plot	4	347	Risk Ratio (M‐H, Random, 95% CI)	1.05 [0.89, 1.24]

1.1 Azathioprine	3	256	Risk Ratio (M‐H, Random, 95% CI)	1.13 [0.94, 1.37]
1.2 6‐mercaptopurine	1	91	Risk Ratio (M‐H, Random, 95% CI)	0.91 [0.70, 1.18]
2 Clinical relapse (subgroup by length of follow‐up) Show forest plot	4	347	Risk Ratio (M‐H, Random, 95% CI)	1.05 [0.89, 1.24]

2.1 12 months or less	1	79	Risk Ratio (M‐H, Random, 95% CI)	1.08 [0.84, 1.39]
2.2 Over 12 months	3	268	Risk Ratio (M‐H, Random, 95% CI)	1.07 [0.82, 1.39]
3 Endoscopic relapse at 24 months Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

4 Radiologic relapse at 24 months Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

5 Surgical relapse at 24 months Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

6 Adverse events at 12 to 24 months (subgroup by drug type) Show forest plot	4	346	Risk Ratio (M‐H, Random, 95% CI)	0.89 [0.74, 1.07]

6.1 Azathioprine	3	255	Risk Ratio (M‐H, Random, 95% CI)	0.87 [0.70, 1.09]
6.2 6‐mercaptopurine	1	91	Risk Ratio (M‐H, Random, 95% CI)	1.40 [0.54, 3.62]
7 Adverse events (subgroup by length of follow‐up) Show forest plot	4	346	Risk Ratio (M‐H, Random, 95% CI)	0.89 [0.74, 1.07]

7.1 12 months or less	1	78	Risk Ratio (M‐H, Random, 95% CI)	0.85 [0.70, 1.03]
7.2 Over 12 months	3	268	Risk Ratio (M‐H, Random, 95% CI)	0.94 [0.64, 1.38]
8 Serious adverse events at 12 to 24 months (subgroup by drug type) Show forest plot	3	311	Risk Ratio (M‐H, Random, 95% CI)	3.39 [1.26, 9.13]

8.1 Azathioprine	2	220	Risk Ratio (M‐H, Random, 95% CI)	4.68 [0.64, 34.12]
8.2 6‐mercaptopurine	1	91	Risk Ratio (M‐H, Random, 95% CI)	4.69 [0.23, 95.00]
9 Serious adverse events (subgroup by length of follow‐up) Show forest plot	3	311	Risk Ratio (M‐H, Random, 95% CI)	3.39 [1.26, 9.13]

9.1 12 months or less	1	78	Risk Ratio (M‐H, Random, 95% CI)	19.00 [1.15, 313.35]
9.2 Over 12 months	2	233	Risk Ratio (M‐H, Random, 95% CI)	2.63 [1.12, 6.16]
10 Withdrawal due to adverse events at 12 to 24 months (subgroup by drug type) Show forest plot	5	425	Risk Ratio (M‐H, Random, 95% CI)	2.21 [1.28, 3.81]

10.1 Azathioprine	4	334	Risk Ratio (M‐H, Random, 95% CI)	2.76 [1.41, 5.40]
10.2 6‐mercaptopurine	1	91	Risk Ratio (M‐H, Random, 95% CI)	1.40 [0.54, 3.62]
11 Withdrawal due to adverse events (subgroup by length of follow‐up) Show forest plot	5	425	Risk Ratio (M‐H, Random, 95% CI)	2.21 [1.28, 3.81]

11.1 12 months or less	2	157	Risk Ratio (M‐H, Random, 95% CI)	3.54 [0.83, 15.08]
11.2 Over 12 months	3	268	Risk Ratio (M‐H, Random, 95% CI)	1.95 [1.03, 3.68]
12 HRQoL ‐ IBDQ > 170 at 24 months Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

13 HRQoL ‐ IBDQ at 12 months Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected

Comparison 2. Azathioprine or 6‐mercaptopurine versus 5‐aminosalicylic acid

Comparison 3. Azathioprine or 6‐mercaptopurine versus anti TNF‐α

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Clinical relapse at 12 to 24 months Show forest plot	3	139	Risk Ratio (M‐H, Random, 95% CI)	2.89 [1.50, 5.57]

2 Clinical relapse (subgroup by length of follow‐up) Show forest plot	3	139	Risk Ratio (M‐H, Random, 95% CI)	2.89 [1.50, 5.57]

2.1 12 months or less	2	106	Risk Ratio (M‐H, Random, 95% CI)	2.27 [1.07, 4.82]
2.2 Over 12 months	1	33	Risk Ratio (M‐H, Random, 95% CI)	6.12 [1.63, 22.97]
3 Endoscopic relapse at 12 to 24 months (subgroup by drug type) Show forest plot	4	157	Risk Ratio (M‐H, Random, 95% CI)	3.67 [1.05, 12.81]

3.1 Azathioprine	3	138	Risk Ratio (M‐H, Random, 95% CI)	3.47 [0.75, 16.03]
3.2 6‐mercaptopurine	1	19	Risk Ratio (M‐H, Random, 95% CI)	5.5 [0.75, 40.36]
4 Endoscopic relapse (subgroup by length of follow‐up) Show forest plot	4	157	Risk Ratio (M‐H, Random, 95% CI)	3.67 [1.05, 12.81]

4.1 12 months or less	3	124	Risk Ratio (M‐H, Random, 95% CI)	2.44 [0.83, 7.18]
4.2 Over 12 months	1	33	Risk Ratio (M‐H, Random, 95% CI)	10.35 [1.50, 71.32]
5 Radiologic relapse at 12 to 24 months Show forest plot	2		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

6 Histologic relapse at 12 months Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

7 Adverse events at 12 to 24 months Show forest plot	2	117	Risk Ratio (M‐H, Random, 95% CI)	1.13 [0.83, 1.53]

8 Adverse events (subgroup by length of follow‐up) Show forest plot	2	117	Risk Ratio (M‐H, Random, 95% CI)	1.13 [0.83, 1.53]

8.1 12 months or less	1	84	Risk Ratio (M‐H, Random, 95% CI)	1.04 [0.65, 1.66]
8.2 Over 12 months	1	33	Risk Ratio (M‐H, Random, 95% CI)	1.20 [0.81, 1.78]
9 Serious adverse events at 12 months Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

10 Withdrawal due to adverse events at 12 to 24 months Show forest plot	3	139	Risk Ratio (M‐H, Random, 95% CI)	3.97 [0.92, 17.22]

11 Withdrawal due to adverse events (subgroup by length of follow‐up) Show forest plot	3	141	Risk Ratio (M‐H, Random, 95% CI)	4.17 [0.99, 17.47]

11.1 12 months or less	2	106	Risk Ratio (M‐H, Random, 95% CI)	7.17 [1.32, 38.96]
11.2 Over 12 months	1	35	Risk Ratio (M‐H, Random, 95% CI)	1.06 [0.07, 15.62]
12 HRQoL ‐ IBDQ > 170 at 24 months Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

Comparison 3. Azathioprine or 6‐mercaptopurine versus anti TNF‐α