A randomised, double-blind, parallel-group trial to assess mercaptopurine versus placebo to prevent or delay recurrence of Crohn's disease following surgical resection (TOPPIC)

Article history

The research reported in this issue of the journal was funded by the EME programme as project number 09/800/04. The contractual start date was in October 2007. The final report began editorial review in February 2016 and was accepted for publication in March 2017. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The EME editors and production house have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the final report document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Jack Satsangi reports personal fees from Takeda Pharmaceutical Company Ltd, AbbVie Inc. and Dr Falk Pharma UK Ltd, and research funding from the European Commission. Nicholas A Kennedy reports grants from The Wellcome Trust; personal fees from Merck & Co., Inc., Takeda Pharmaceutical Company Ltd, Dr Falk Pharma UK Ltd, Allergan, plc, and Pharmacosmos A/S; non-financial support from Norgine BV, AbbVie Inc., Shire, plc, and Janssen Global Services, LLC; and other support from Merck & Co., Inc., AbbVie Inc. and Takeda Pharmaceutical Company Ltd outside the submitted work. Ian Arnott reports personal fees from Vifor Pharma Management Ltd, Takeda Pharmaceutical Company Ltd and Dr Falk Pharma UK Ltd. Steff Lewis reports membership of the Health Technology Assessment Efficient Study Designs Board and University of Edinburgh grant funding.

Copyright statement

© Queen’s Printer and Controller of HMSO 2017. This work was produced by Satsangi et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Crohn’s disease treatment

Treatment algorithms for the medical management of active CD have evolved very rapidly in recent years, and the relative importance of thiopurines in induction of remission, and in maintenance, has been re-evaluated since this study was first designed. Although steroids, biological therapies and combination therapies (biologics used together with thiopurine agents) have an evidence basis for induction of remission, a recent Cochrane review suggested that there was no place for monotherapy with thiopurine in induction of remission. 3

However, for patients entering medically induced remission, immunosuppression with azathioprine or mercaptopurine (MP) is an effective maintenance strategy (odds ratio 3.17, number needed to treat 3.3 favouring 2 mg/kg azathioprine over placebo), offering an acceptable balance of efficacy, tolerability and cost. 4 The relative merits of monotherapy with thiopurines alone, anti-tumour necrosis factor alpha (TNF-α) agents or combination regimes of thiopurines and biological agents remain contentious in this setting of maintenance therapy.

Despite current medical therapy, up to 50% of patients have aggressive disease and require surgery within 5 years of diagnosis. 5 Unfortunately, disease relapse rates are high within 2 years of surgery for both endoscopic (72–98%) and clinical recurrence (37–70%). 6 Re-operation rates cumulate at 5% of patients per year. 7

Evidence for postoperative use of thiopurines

The development of strategies to prevent or delay postoperative recurrence of CD is therefore of major clinical importance; however, there is a paucity of evidence to support any particular drug strategy. 8,9 The thiopurines, azathioprine and MP have an established role in inducing remission in CD and in the maintenance of medically induced remission. These agents have historically been suggested in treatment algorithms for patients at ‘high risk’ of postoperative relapse, but the evidence basis for their use in this context, and indeed the evidence that clinical parameters may predict patients at ‘high risk’ of relapse, is weak. 7 A meta-analysis conducted by Jones et al. 10 found that the efficacy data for thiopurines in the postoperative setting were inconclusive and, aside from smoking, there were no consistent predictors of postoperative relapse. A 2014 Cochrane review also concluded that, although there was evidence to suggest that thiopurines may reduce endoscopic and clinical recurrence, the quality of evidence was degraded by small numbers and flawed study designs. 3 The value of thiopurine metabolites in this situation is unknown. 11 More recently, the role of biological therapies in postoperative prophylaxis has received considerable attention. Following smaller randomised studies of infliximab in postoperative disease, the recent Post-Operative Crohn’s Endoscopic Recurrence (POCER) study demonstrated that targeted escalation of immune-modulatory therapy (thiopurines followed by adalimumab) in patients with early endoscopic evidence of recurrence may delay subsequent endoscopic, although not clinical, recurrence. 11–13

Current practice

Currently no medical therapy is licensed to prevent postoperative recurrence in CD. As well as thiopurine agents and anti-TNF-α agents, 5-aminosalicylates, corticosteroids and metronidazole have all been suggested, but the evidence in favour of these drugs is weak and none is widely used in the UK. 6 Postoperative maintenance therapy with azathioprine or MP is more widely used and is included in a number of clinical algorithms, but trial data from controlled trials available thus far have failed to demonstrate whether or not this is an effective therapy. Existing studies have been underpowered, poorly designed or used an inappropriate drug dose. 6,14–19

For these reasons, we consider that evidence-based strategies to prevent postoperative recurrence are both an unmet clinical need and a research priority.

Controversies in use of thiopurines in postoperative use

As we discussed in our recent review (see Jones et al. 10), there are many unresolved controversies associated with the use of thiopurines in the prevention of postoperative recurrence in CD, and this remains contentious. The definition of recurrence itself, specifically the distinction between endoscopic, clinical and surgical recurrence, the identification of high-risk individuals, the influence of pre-operative therapies on responsiveness, and the relative merits of thiopurines compared with anti-TNF-α therapies are key issues; however, no real evidence-based consensus has emerged. 10 The European Crohn’s and Colitis Organisation (ECCO) 2010 guidelines recommend their use postoperatively in ‘high-risk’ patients (defined as smokers and those with previous intestinal surgery, penetrating disease, perianal disease or extensive small bowel section). 20 The British Society of Gastroenterology’s guidelines are more cautious in emphasising the weak evidence base for risk factors in predicting disease recurrence (with the exception of smoking) and for the use of thiopurines in this setting. 8

The use of thiopurines by women, or the male partners of women, trying to conceive or in pregnancy

With respect to safety, use in patients of childbearing age has been a key issue. When the trial commenced, the published guidelines of both ECCO and the British Society of Gastroenterology supported the use of thiopurines in inflammatory bowel disease (IBD) in women of childbearing potential and during pregnancy on the grounds that the risk to the unborn child from disease activity appeared greater than continued therapy. As part of the initial trial assessment by the Medicines and Healthcare products Regulatory Agency, supporting documentation was reviewed and the inclusion of women of childbearing potential was approved without inclusion criteria of agreement to the use of contraceptives. No comment was made for inclusion of males who could potentially father a child while receiving MP. However, the sponsor, Trial Management Group (TMG) and Trial Steering Committee (TSC) examined this position in more detail in 2010.

The TMG and TSC considered the evidence and decided not to exclude female patients who either were planning to become pregnant or became pregnant, or male patients whose partners were planning to become pregnant. The existing patient information sheets, however, were revised to detail the potential risks to male patients. All enrolled male patients were also reconsented to the new information sheets and consent forms. Pregnancy data were also included in the TSC and Data Monitoring Committee (DMC) standard reports.

Since 2010, several large studies have been published examining thiopurine use for IBD during pregnancy, which provided further support for the initial guidelines. The Cancers Et Surrisque Associé aux Maladies inflammatoires intestinales En France (CESAME) study group examined 215 pregnancies in 204 women and concluded that thiopurine use during pregnancy was not associated with increased risks, including an increased risk of congenital abnormalities. 21 This has been supported by the findings of two further studies published in 2013 and 2014: Casanova et al. 22 included a cohort of 187 pregnancies in participants using thiopurines in IBD and found no evidence of an increased risk of complications, whereas Ban et al. 23 included a cohort of 115 pregnancies in participants using thiopurines for CD and found no evidence of an increased risk of major birth defects.

Potential health-care impact

The present study has an important potential impact on health care in terms of defining, with accuracy, the clinical effectiveness of thiopurine therapy in the prevention of postoperative relapse. The data from the study overall suggest that routine postoperative use of MP in all patients undergoing surgery is not justified as a preventative measure. The subgroup analyses have helped identify smokers as the subgroup of individuals at highest risk of relapse, and helped to demonstrate that this subgroup is most likely to benefit from thiopurine therapy, thereby helping to target immunomodulatory therapy towards patients most likely to benefit. The important secondary objectives allowed comparison of clinical and endoscopic recurrence rates, and analysis of clinical as well as biomarker predictors of disease recurrence. Importantly, we have assessed age, the need for previous surgery, previous thiopurine use and smoking as potential determinants of outcome.

Rationale for research

At present, there is no widely accepted algorithm or management plan in preventing postoperative relapse of CD. Non-evidence-based algorithms have been widely used and the present study will allow clear evidence to be defined in the largest study carried out in the use of thiopurine therapy.

The secondary objectives allowed comparison of clinical and endoscopic recurrence rates, and the effects of thiopurine therapy on each of these aspects. The study design has also allowed examination of drug safety, tolerability and adherence to therapy. The biomarkers in the study allow the relative predictive value of faecal calprotectin levels in assessing recurrence to be defined, as well as the relationship between thiopurine use and calprotectin levels. Quality of life was also measured in the present study. The primary objective of this study was to assess if MP can prevent or delay postoperative recurrence in CD.

Study aims

The primary objective of the Trial Of Prevention of Post operative Crohn’s disease (TOPPIC) was to assess if MP can prevent or delay postoperative recurrence of CD.

Secondary objectives included if MP can prevent or delay endoscopic evidence of recurrence and, finally, whether or not endoscopic recurrence can predict clinical recurrence, whether or not faecal calprotectin levels can be used as a non-invasive marker of disease recurrence that may remove the need for colonoscopy in some patients, whether or not drug metabolite levels relate to clinical efficacy of MP, and whether or not we can predict postoperative recurrence using clinical, genetic or serological data.

Development of TOPPIC

Development of TOPPIC was initiated in 2006 by a team of collaborators based in Scotland, with ethics and clinical trial authorisation sought in 2007. The trial was extended throughout the UK in 2010 to include collaborators from across England and Wales.

Trial design

This was a multicentre, parallel-group, double-blind, randomised controlled study conducted in the UK (29 sites). Participants were allocated 1 : 1 to active treatment with MP or a matching placebo.

Following slower than projected recruitment in the first 24 months, the protocol was modified to reduce the number of colonoscopies performed as part of the trial and to allow the inclusion of patients with a previous history of malignancy, providing that they had been in remission for 5 years.

Participants

Participants were adults aged ≥ 16 years in Scotland (≥ 18 years in England and Wales) with a histologically confirmed diagnosis of CD (according to the Lennard-Jones criteria) undergoing ileocolonic or small bowel resection. 24 Patients were approached ≤ 3 months after surgery during the perioperative period and, following consent for screening, underwent a screening assessment consisting of a blood sample to check thiopurine S-methyltransferase (TPMT) activity and a postoperative stool sample to exclude enteric infection.

One week prior to randomisation, eligible participants underwent standard safety blood tests (including checks of liver function, haemoglobin levels, and white blood cell, neutrophil and platelet counts) in addition to a urinary pregnancy test.

Inclusion and exclusion criteria are described in the following sections.

Identifying participants

The study aimed to recruit and randomise from participating sites. Potentially eligible patients were identified by clinicians and research/specialist nurses at sites following ileocolonic or small bowel resection.

Settings and locations where the data were collected

The study took place in the following tertiary referral hospitals in the UK: Western General Hospital, Edinburgh; Aberdeen Royal Infirmary, Aberdeen; Ninewells Hospital, Dundee; Glasgow Royal Infirmary and Stobhill Hospital, Glasgow; Royal Devon and Exeter Hospital, Exeter; University Hospital Coventry, Coventry; Royal Liverpool University Hospital, Liverpool; Manchester Royal Infirmary, Manchester; Bristol Royal Infirmary, Bristol; University College Hospital, Royal Free Hospital and Royal London Hospital, London; John Radcliffe Hospital, Oxford; Salford Royal Hospital, Salford; Singleton Hospital, Swansea; Torbay Hospital, Torquay; Southampton General Hospital, Southampton; County Durham and Darlington Memorial Hospitals, Durham and Darlington; Royal Stoke University Hospital, Stoke-on-Trent; Queen’s Medical Centre, Nottingham; St Mark’s Hospital, Harrow; Rotherham General Hospital, Rotherham; Queen Elizabeth Hospital, Birmingham; and Raigmore Hospital, Inverness. Randomisation was open from May 2008 until June 2012 (a total of 49 months).

Interventions

The intervention was a random allocation to MP in 50-mg tablets or placebo in identical 50-mg tablets and matched packaging supplied to sites by the manufacturer.

Outcomes

The primary study outcome was clinical recurrence of CD defined by a Crohn’s Disease Activity Index (CDAI) score of ≥ 150 points together with a 100-point rise in the CDAI score from baseline, with the need for anti-inflammatory rescue therapy or primary surgical intervention. 25

The secondary study outcomes were as follows:

• clinical recurrence defined by a CDAI score of ≥ 150 points together with a 100-point rise in the CDAI score from baseline, or anti-inflammatory rescue therapy or primary surgical intervention

• the need for a second operation to remove recurrent CD from the anastomotic site

• endoscopic recurrence using the Rutgeerts and Crohn’s Disease Endoscopic Index of Severity (CDEIS) scoring systems26,27

• faecal calprotectin levels

• thioguanine (TGN) levels

• changes in self-rated quality-of-life scores.

Measurements

Following patient consent for screening, the trial screening process consisted of analysis of TPMT activity and a postoperative stool sample to exclude enteric infection. One week prior to randomisation, eligible participants underwent blood tests (including liver function tests, haemoglobin concentration tests, and white blood cell, neutrophil and platelet counts) and a urinary pregnancy test. Patients returned to designated clinics run by clinical teams for a baseline assessment including the CDAI, patient-reported outcome measures including the Inflammatory Bowel Disease Questionnaire (IBDQ), a physical examination and a blood sample for assay of drug metabolite level. 28 Additional blood samples for genetic and serological analysis were also taken, and results are reported separately.

Patients were randomised at baseline to receive either MP or placebo. MP was given at a dose of 1–1.5 mg/kg of body weight, rounded to the nearest 25 mg as a single daily dose in the morning. Patients with low TPMT activity were prescribed MP at half the normal dose. Treatment was on a maintenance basis for 3 years, with the dose adjusted up or down for weight changes during the study.

Follow-up data were collected at 10 time points, at weeks 6, 13, 31, 49, 67, 85, 103, 121, 139 and 157 post randomisation. At each visit, the following data were collected: CDAI score, physical characteristics (as measured by examination), concomitant medications taken and patient-reported outcomes including the IBDQ. At weeks 0, 13, 49, 103 and 157 post randomisation, patients were also asked to provide additional stool and blood samples for the central assessment of faecal calprotectin and TGN metabolite levels, with endoscopic assessment at weeks 49 and 157 post randomisation.

Sample processing

Stool samples were collected at participating centres and frozen on site at –80 °C. Samples were then shipped on dry ice in two batches during 2014 and 2015 to a central biochemistry laboratory at NHS Lothian Western General Hospital in Edinburgh. Faecal calprotectin levels were all assessed using the CALPRO^® calprotectin enzyme-linked immunosorbent assay (alkaline phosphatase) (NovaTec Immundiagnostica GmbH, Dietzenbach, Germany) and processed by technicians blinded to the treatment allocation. Blood samples for TGN testing were sent ambient to Viapath (formerly GSTS Pathology; London, UK) for analysis in the Purine Research Laboratory. An independent analyst blinded to the treatment allocation then assessed each level.

Therapeutic monitoring

All patients underwent regular safety blood monitoring every week for the initial 6 weeks and thereafter at 6-weekly intervals as long as the patient remained on the study drug. Blood samples were processed at participating site laboratories. An independent central clinician who was blinded to the treatment allocation reviewed results. Prespecified dose reduction or cessation then occurred in the event of abnormal monitoring parameters. In the event of patient intolerance (profound nausea or persistent influenza-like symptoms), protocol-driven dose reduction was also undertaken. If abnormal parameters improved after a temporary stop, treatment was started again at a lower level. The central clinicians, blinded to the treatment allocation, made all of these decisions.

To protect blinding, a programme of sham dose reductions was planned for patients on placebo. On the advice of the DMC these were not undertaken, although the investigators were not informed of this decision, hence protecting the study blinding.

Dose changes

All randomised patients allocated to the treatment group were given 1–1.5 mg/kg body weight, rounded to the nearest 25 mg, to be taken as single daily dose in the morning. Treatment was on a maintenance basis for 3 years or until the study drug was discontinued. Patients heterozygous for TPMT mutations were prescribed MP at half the normal dose. The dose was adjusted up or down for weight changes according to Table 1.

The blinded treatment of either MP or placebo could also be reduced or temporarily stopped following patient intolerance (profound nausea or persistent influenza-like symptoms) or abnormal blood safety results. In these cases, the dose was reduced in accordance with Table 2 and not increased again during the course of the trial. If symptoms or blood safety test result abnormalities persisted and were, in the view of the investigator, of sufficient severity, the drug could be stopped. One of the blinded clinicians assessing safety blood test results could also stop the drug.

Sample size

It was estimated that a study population of 182 evaluable patients would provide the trial with 80% power to detect a reduction in the frequency of recurrence from 50% in the placebo group to 30% in the treatment arm by 3 years at the 5% level of significance. Preliminary data from surgical and pathology databases at the five Scottish recruiting sites indicated that this was an achievable target. In 2006–7, approximately 130 ileocolonic resections were performed annually at the five Scottish recruiting sites initially proposed. We estimated that 60% of the potential patients could be recruited, giving a sample population of 234 patients and allowing for a 15% dropout rate. This would leave an evaluable population of 200 patients over 3 years.

These figures were based on an intention-to-treat analysis, with the number needed to treat in order to prevent one recurrence predicted to be five. It is notable that the treatment effect of 20% was lower than in previously conducted studies (40% reduction in mild CD lesions and 75% reduction in more severe lesions in Hanauer et al. ;18 25% in Ardizzone et al. 19). We also judged a treatment effect of 20% to be appropriate because, given the side-effect profile of MP, it was arguable that a treatment effect of significantly < 20% would be of limited clinical significance.

The sample size chosen had 80% power to detect a reduction in the frequency of recurrence from 50% in the placebo group to 30% in the treatment arm by 3 years at the 5% level of significance.

Interim analyses and stopping guidelines

An independent DMC reviewed trial data throughout the course of the trial, and there were no formal stopping rules. The study recruitment period was extended and additional sites in England and Wales were opened up to address concerns regarding the recruitment rate. Aside from this, the trial recruited according to plan.

Randomisation

Trial data management

Trial data were manually entered into a web-based data collection system by researchers at participating sites. In this system, active server pages supported web-based electronic case report forms and queried a Microsoft SQL Server 2000 database (Microsoft Corporation, Redmond, WA, USA). Data were automatically encrypted (using Secure Sockets Layer protocol V329) when entered into the forms, transferred via the internet, decrypted and stored in a secure maintained network at the University of Edinburgh. The level of encryption was dependent on the client’s web browser and operating system, but University of Edinburgh servers could offer up to 256-bit encryption. In the same network, data were then queried using statistical software (SAS; version 9.4, SAS Institute Inc., Cary, NC, USA) for the purposes of data analysis. All changes to electronic data by site staff were traceable by date and user. A copy of the trial electronic case report form is given in Report Supplementary Material 1.

Statistical methods

An analysis was undertaken using SAS. The full statistical analysis plan is given in Report Supplementary Material 2.

Primary outcomes

The primary outcome variable was postoperative recurrence of CD and its timing if it recurred. Analysis was intention to treat and was based on the application of a Cox proportional hazards model. The primary analysis included terms for treatment, the variables on which the randomisation was stratified (smoking status and recruitment site) and adjusted for baseline values of previous treatment with MP and previous treatment with azathioprine. Adjusted and unadjusted Cox proportional hazard ratios (HRs) are presented as the comparison of MP versus placebo (reference), with a HR of < 1 indicating a treatment effect in favour of MP. The adjusted analysis was considered to be the primary analysis of the primary outcome.

Secondary outcomes

The secondary outcome variable of clinical recurrence of CD (defined by a CDAI score of ≥ 150 points together with a 100-point rise in the CDAI score from baseline) or the need for anti-inflammatory rescue therapy or primary surgical intervention was analysed in the same manner as the primary outcome. For this secondary outcome, as for the primary outcome, the adjusted analysis was considered the primary analysis.

Endoscopic recurrence using both the Rutgeerts and CDEIS scoring systems was summarised by time and treatment group. Colonoscopy results at week 157 post randomisation (study visit 12) were compared between the treatment groups using a chi-squared test to compare the incidence of positive colonoscopies in the MP group and the placebo group. Both adjusted and unadjusted analyses were performed, with the adjusted analysis incorporating the same covariates as for the primary outcome, with odds ratios and 95% confidence intervals (CIs) presented. CDEIS scores at week 157 post randomisation were compared between treatment groups using a t-test. Adjusted and unadjusted analyses were performed, with treatment effects and 95% CIs presented. Faecal calprotectin results were summarised by time and treatment group, both as a continuous measure and categorically.

The use of faecal calprotectin levels as a non-invasive marker of disease recurrence was examined in two ways: first, it was considered as a time-dependent covariate in Cox proportional hazards model and, second, levels were compared descriptively between those with negative and positive colonoscopies (defined as negative if the Rutgeerts score was < i2 and positive if the Rutgeerts score was ≥ i2) at weeks 49 and 157 post randomisation. Similarly, TGN levels of the MP drug metabolite were also considered as a time-dependent covariate in the Cox proportional hazards model.

The quality-of-life variables were analysed using repeated measures analysis of covariance to evaluate treatment and treatment by time interactions. Quality of life, as measured by the IBDQ, was summarised at each visit based on observed scores and change from baseline scores for each of the four IBDQ subscales (bowel symptoms, emotional health, systemic systems and social function). Averages and totals across all subscales were summarised similarly. In addition, the overall average and overall total scores were analysed using a change from baseline repeated measures analysis of covariance to evaluate the effect of treatment over time. Quality of life, as measured by the EuroQol-5 Dimensions (EQ-5D), was summarised by treatment group across study visits.

Subgroup analyses

Subgroup analyses of the primary and secondary outcomes were carried out to assess for a treatment effect in terms of thiopurine naivety, previous treatment with infliximab or methotrexate, previous surgery, smoking status, duration of disease and age at diagnosis. The interaction between subgroup and treatment was included in the Cox regression model to determine if the treatment effect differed by subgroup. The same subgroups analysed for the primary and secondary outcomes were also analysed with respect to colonoscopy results and CDEIS scores.

Publication policy

To safeguard the integrity of the trial, the primary results of the trial were published by the group as a whole in collaboration with local investigators, and local investigators were acknowledged. 30 The success of the trial was dependent on the collaboration of many people and, particularly, the local investigators. The results were, therefore, presented to the trial local investigators first.

Organisation

A TSC and a DMC were established. No formal charter was put in place, but both committees met formally every 6 months throughout the trial in accordance with Medical Research Council (MRC)/Efficacy and Mechanism Evaluation (EME) programme requirements following circulation of a standardised report. Day-to-day management of the trial was overseen by a TMG comprising the chief investigator, the principal investigator at the lead site in Edinburgh, the lead research nurse and the trial manager. Support from the Chief Scientist Officer enabled a dedicated specialist nurse to support the trial at the initial five Scottish sites. The local Comprehensive Local Research Networks supported research nursing time and employed or re-allocated a research nurse to support all aspects of the trial at sites across England and Wales.

Confidentiality

Patients were identified by their trial number to ensure confidentiality. Stringent precautions were taken to ensure confidentiality of names and addresses at ECTU and the sites. The chief investigator and local investigators ensured conservation of records in areas to which access is restricted.

Audit

A risk-based monitoring strategy was implemented for all participating sites with onsite monitoring conducted by either the sponsor’s monitoring team or a member of the TMG at a selected number of sites depending on the issues identified at each site. Site visits were conducted during the course of the trial at University College Hospital, London, on 5 December 2012, Torbay Hospital, Torquay, on 6 November 2013, Ninewells Hospital, Dundee, on 12 March 2014, Aberdeen Royal Infirmary, Aberdeen, on 11 September 2015, Bristol Royal Infirmary, Bristol, on 30 September 2015 and St Marks Hospital, Harrow, on 7 October 2015. A central audit of trial management was conducted on 12 December 2012 at the central co-ordinating site at ECTU. In addition, an investigation was conducted by the sponsor following the lodging of a formal complaint to the Medicines and Healthcare products Regulatory Agency in July 2013 regarding the wording on an information newsletter sent to a site regarding withdrawal procedures. Recommendations were made following the investigation, but no evidence was found of improper conduct regarding withdrawal procedures.

Termination of the study

Before termination of recruitment, ECTU contacted all sites by telephone or e-mail in order to inform sites of the final date for recruitment. Once the recruitment period had expired, the internet-based randomisation database was disabled to prevent further recruitment. All recruited patients received trial medication for a maximum of 3 years (36 months) post randomisation. The final patient’s final visit took place in April 2015 and the database was locked in June 2015. A declaration of the end of trial form was sent to the Multicentre Research Ethics Committee following the formal trial end date of 30 September 2015. The following documents were archived in each site file to be kept for at least 20 years in accordance with MRC policy: original consent forms, data forms, trial-related documents and correspondence. The trial master files at ECTU will be archived for at least 20 years.

Funding

The costs for the study itself were covered by a grant from the MRC. See the MRC/EME programme website for further project information. 31 Additional support was provided by the Chief Scientist Officer.

Indemnity

If there was negligent harm during the clinical trial, then the NHS body owes a duty of care to the person harmed and NHS indemnity covered NHS staff, medical academic staff with honorary contracts and those conducting the trial. NHS indemnity did not offer no-fault compensation. The co-sponsors were responsible for ensuring that proper provision was made for insurance or indemnity to cover their liability and the liability of the chief investigator and staff.

Ethics approval and research governance

Ethics approval for the study was given by Scotland A Research Ethics Committee in August 2007 (reference number 07/MRE00/74). Local NHS management approval and appropriate site-specific assessments were obtained at each participating NHS trust. The trial was registered with the International Standard Randomised Controlled Trial Register under the reference number ISRCTN89489788 along with the European Clinical Trials Database under the reference number 2006-005800-15, and the UK Clinical Research Network Portfolio Database under the reference number 5813. A summary of the changes made to the original protocol is given in Report Supplementary Material 3 and an overview of all participant information and consent forms is given in Report Supplementary Material 4.

Recruitment

Patients were recruited from the following dates: June 2008 to April 2012 (46 months). The addition of 24 sites in England and Wales was in response to slower than anticipated recruitment over the initial 24 months. Initial study recruitment projection figures were based on surgical data collected from the five original Scottish hospitals in the 5 years prior to the grant application. These data indicated that 60% of patients would proceed from surgery into the trial. In practice, the proportion of patients recruited into the trial following surgery at the initial five sites was 30% as a result of higher than expected contraindications to the trial entry and patients’ greater than anticipated reluctance to receive placebo. An extension to the study recruitment period was granted in 2010, along with approval for the trial to be extended to additional centres across England and Wales. The trial completed recruitment on target according to the revised schedule on 23 April 2012, with the final patient’s final visit completed on schedule in April 2015.

Participant flow

A Consolidated Standards of Reporting Trials (CONSORT) diagram for recruitment is provided in Figure 1. In total, 328 patients were screened and 240 were randomised, thereby meeting and slightly exceeding the recruitment target of 234. Seventy-eight per cent of eligible patients were randomised to the study following screening, with 22% of patients declining to take part.

FIGURE 1.

The CONSORT diagram for TOPPIC.

All 240 randomised patients were analysed. A total of 128 (53%) participants were randomised to receive MP and 112 (47%) to placebo. All randomised patients received at least one dose of study drug. A protocol violation was recorded involving a participant who was prescribed a study drug from the wrong treatment arm 6 weeks post randomisation; the error was reported and the correct study drug issued.

Summary of ineligible and non-recruited patients

Table 3 summarises the number of ineligible and non-recruited patients at the participating centres. Five centres did not randomise any patients during the recruitment period and were closed early.

Characteristics used for stratification of randomisation

Randomised patients were stratified according to centre and smoking status (31 blocks in total). Table 4 shows the numbers of patients randomised per recruiting centre, with the Edinburgh site recruiting the highest proportion of patients (32.5%). Smoking status (defined as those patients smoking more than one cigarette per day) is shown in Table 5.

Characteristics at pre-assessment

Characteristics of randomised patients at pre-assessment are shown in Table 6. The group randomised to MP had slightly more patients who had previously been treated with MP, azathioprine and immunosuppressant agents, although all other demographics and characteristics were similar.

Patient characteristics at randomisation

Table 7 shows the baseline characteristics of participants at randomisation (week 0). The two treatment groups were similar across the variables assessed, which included weight, faecal calprotectin levels, and CDAI or IBDQ scores.

Adherence to trial protocol

Adherence to trial medication

A summary of the duration of trial medication is shown in Table 8, with Figure 2 showing a Kaplan–Meier plot of the duration of medication in years. The reasons for non-completion of the full treatment period are detailed in Table 9. One hundred and thirty-six patients (56.7%) did not complete 3 years of trial medication, with 80 (58.8%) patients not completing as the result of adverse events, 21 (15.4%) because of early withdrawal, 18 (13.2%) because of abnormal safety blood results, 16 (11.8%) being lost to follow-up and one (0.7%) patient dying.

TABLE 8 - Duration of trial medication

Duration of trial medication (years)	Treatment, n (%)		Overall (N = 240), n (%)
	MP (N = 128)	Placebo (N = 112)
≥ 3	62 (48.4)	42 (37.5)	104 (43.3)
< 3	66 (51.6)	70 (62.5)	136 (56.7)

FIGURE 2.

Kaplan–Meier plot showing duration of trial medication in years.

TABLE 9 - Reasons for randomised patients not completing 3 years of trial medication

Reason	Treatment, n (%)		Overall (N = 240), n (%)
	MP (N = 128)	Placebo (N = 112)
Adverse event	39 (59.1)	41 (58.6)	80 (58.8)
Blood test result	12 (18.2)	6 (8.6)	18 (13.2)
Early withdrawal	8 (12.1)	13 (18.6)	21 (15.4)
Deceased	0 (0.0)	1 (1.4)	1 (0.7)
Lost to follow-up	7 (10.6)	9 (12.9)	16 (11.8)

Dosage was adjusted throughout the trial in accordance with the protocol (see Tables 1 and 2). Table 10 provides a summary of doses at first and final visits. Table 11 summarises doses at final visit compared with dose first allocated. Only a small proportion of patients continued on the same dosage (n = 63, 26.3%) or at an increased dosage (n = 13, 5.4%); in the majority either the dose was decreased (n = 68, 28.3%) or the trial medication was stopped altogether (n = 96, 40.0%). The dosage was decreased in a higher proportion of patients in the MP group (39.1%) than in the placebo group (16.1%), whereas a higher proportion in the placebo group than in the MP group remained on the same dosage (37.5% compared with 16.4%, respectively) or stopped trial medication (42.9% compared with 37.5, respectively). The dose was increased in a small number of patients, but the proportion was marginally higher in the MP group than in the placebo group (7.0% compared with 3.6%).

TABLE 10 - Summary of doses at first and final visit

Dosage	Treatment, n (%)		Overall (N = 240), n (%)
	MP (N = 128)	Placebo (N = 112)
Dose (mg) allocated at first visit
50 alternate days	0 (0.0)	1 (0.9)	1 (0.4)
50 daily	17 (13.3)	17 (15.2)	34 (14.2)
50/100 alternate days	73 (57.0)	62 (55.4)	135 (56.3)
100 daily	34 (26.6)	27 (24.1)	61 (25.4)
150 daily	4 (3.1)	5 (4.5)	9 (3.8)
Dose allocated at final visit
Stop	48 (37.5)	48 (42.9)	96 (40.0)
50 alternate days	17 (13.3)	5 (4.5)	22 (9.2)
50 daily	33 (25.8)	20 (17.9)	53 (22.1)
50/100 alternate days	17 (13.3)	25 (22.3)	42 (17.5)
100 daily	13 (10.2)	13 (11.6)	26 (10.8)
150 daily	0 (0.0)	1 (0.9)	1 (0.4)

TABLE 11 - Summary of doses at final visit, compared with dose first allocated

Final dose change	Treatment, n (%)		Overall (N = 240), n (%)
	MP (N = 128)	Placebo (N = 112)
Dose decreased	50 (39.1)	18 (16.1)	68 (28.3)
Dose stayed the same	21 (16.4)	42 (37.5)	63 (26.3)
Dose increased	9 (7.0)	4 (3.6)	13 (5.4)
Trial medication stopped	48 (37.5)	48 (42.9)	96 (40.0)

Concomitant medication

Concomitant medications are shown in Table 16.

Outcomes and estimation

Secondary outcome: clinical recurrence

The secondary outcome of clinical recurrence was postoperative clinical recurrence of CD, defined as a CDAI score of > 150 points with an increase from baseline of 100 points, or the need for anti-inflammatory rescue therapy or primary surgical intervention.

The secondary outcome events and statistical analysis are summarised in Tables 20 and 21 respectively. The secondary outcome was reached by 74 (30.8%) patients: 34 out of 128 (26.6%) on MP versus 40 out of 112 (35.7%) on placebo (HR 0.737, 95% CI 0.44 to 1.23; adjusted p = 0.243; HR 0.746, 95% CI 0.47 to 1.18; unadjusted p = 0.211). Similar to the primary outcome, the HRs are in favour of MP, but the associated CIs indicate a lack of statistical significance.

TABLE 20 - Summary of secondary outcome events

Secondary outcome	Treatment, n (%)		Overall (N = 240), n (%)
	MP (N = 128)	Placebo (N = 112)
Secondary outcome met?
Yes	34 (26.6)	40 (35.7)	74 (30.8)
No	94 (73.4)	72 (64.3)	166 (69.2)

TABLE 21 - Secondary statistical analysis (Cox proportional hazards model)

Analysis type	HR	95% CI limit		p > χ2
		Lower	Upper
Adjusted	0.737	0.44	1.23	0.243
Unadjusted	0.746	0.47	1.18	0.211

Crohn’s Disease Activity Index score

The mean score change from baseline in the CDAI for patients is shown in Figure 3 from randomisation at visit 2 (week 0) to the final study visit at visit 12 (104 weeks post randomisation).

FIGURE 3.

Mean change from baseline in CDAI scores per study visit.

Secondary outcome: endoscopic recurrence

Tables 24 and 25 present a summary of endoscopic recurrence, as defined by Rutgeerts score, at visits 6 (week 49) and 12 (week 157), categorised by overall score and as either negative or positive. Figure 4 presents the number of negative and positive scores as bar charts.

TABLE 24 - Summary of colonoscopy results defined by Rutgeerts score and categorised by score

Visit 6 was week 49 post randomisation and visit 12 was week 157 post randomisation.

Time point	Treatment, n (%)		Overall (N = 240), n (%)
	MP (N = 128)	Placebo (N = 112)
Visit 6 (week 49, n = 208)
i0: no lesions	33 (29.7)	14 (14.4)	47 (22.6)
i1: ≤ 5 aphthous ulcers	24 (21.6)	28 (28.9)	52 (25.0)
i2: > 5 aphthous ulcers	18 (16.2)	19 (19.6)	37 (17.8)
i3: diffuse aphthous ileitis	11 (9.9)	13 (13.4)	24 (11.5)
i4: diffuse ileal inflammation	5 (4.5)	3 (3.1)	8 (3.8)
Other	4 (3.6)	0 (0.0)	4 (1.9)
Missing	16 (14.4)	20 (20.6)	36 (17.3)
Visit 12 (week 157, n = 161)
i0: no lesions	20 (22.5)	9 (12.5)	29 (18.0)
i1: ≤ 5 aphthous ulcers	18 (20.2)	20 (27.8)	38 (23.6)
i2: > 5 aphthous ulcers	18 (20.2)	11 (15.3)	29 (18.0)
i3: diffuse aphthous ileitis	4 (4.5)	5 (6.9)	9 (5.6)
i4: diffuse ileal inflammation	7 (7.9)	12 (16.7)	19 (11.8)
Other	2 (2.2)	2 (2.8)	4 (2.5)
Missing	20 (22.5)	13 (18.1)	33 (20.5)

TABLE 25 - Summary of colonoscopy results defined by Rutgeerts score as either negative or positive

Visit 6 was week 49 post randomisation and visit 12 was week 157 post randomisation.

Time point	Treatment, n (%)		Overall (N = 240), n (%)
	MP (N = 128)	Placebo (N = 112)
Visit 6 (week 49, n = 208)
Negative (< i2)	57 (51.4)	42 (43.3)	99 (47.6)
Positive (≥ i2)	34 (30.6)	35 (36.1)	69 (33.2)
Other	4 (3.6)	0 (0.0)	4 (1.9)
Missing	16 (14.4)	20 (20.6)	36 (17.3)
Visit 12 (week 157, n = 161)
Negative (< i2)	38 (42.7)	29 (40.3)	67 (41.6)
Positive (≥ i2)	29 (32.6)	28 (38.9)	57 (35.4)
Other	2 (2.2)	2 (2.8)	4 (2.5)
Missing	20 (22.5)	13 (18.1)	33 (20.5)

FIGURE 4.

Colonoscopy by study visit (visit 6 was week 49 post randomisation and visit 12 was week 157 post randomisation). (a) Visit 6 and (b) visit 12.

Colonoscopy results (defined as negative or positive using the Rutgeerts score) are also summarised and the results of the statistical analysis of visit 12 colonoscopy results are presented in Table 26.

TABLE 26 - Visit 12 colonoscopy analysis results

Analysis type	Visit 12 odds ratio	95% CI limit		p-value
		Lower	Upper
Adjusted	0.66	0.26	1.67	0.382
Unadjusted	0.79	0.39	1.61	0.516

Comparing the incidence of positive colonoscopies at visit 12 for the MP group with the placebo group gave an adjusted odds ratio of 0.66 (95% CI 0.26 to 1.67; p = 0.382) and an unadjusted odds ratio of 0.79 (95% CI 0.39 to 1.61; p = 0.516), indicating no significant difference between treatment arms.

Endoscopic recurrence was also assessed using CDEIS score. Figure 5 shows CDEIS scores in a box plot, with a higher CDEIS score indicating greater severity of disease. The results of the statistical analysis are shown in Table 27, with an adjusted treatment effect of –0.612 (95% CI –1.92 to 0.69; p = 0.354), where an effect < 0 indicates a treatment effect in favour of MP. Higher CDEIS scores were recorded for the placebo group than for those patients allocated to MP at visits 6 and 12, although the difference was not statistically significant.

FIGURE 5.

Crohn’s Disease Endoscopic Index of severity score by study visit (visit 6 was 49 weeks post randomisation and visit 12 was 157 weeks post randomisation). (a) Visit 6 and (b) visit 12. Mean CDEIS scores are indicated by a diamond in the box plot. The minimum is indicated by the bottom of the lower whisker, the maximum is indicated by the top of the upper whisker. The lower quartile is indicated by the horizontal line at the bottom of the box and the upper quartile is indicated by the horizontal line at the top of the box. Outliers are shown as small circles, identified as being 1.5 times outside the interquartile range. A higher CDEIS score indicates greater severity of the disease.

TABLE 27 - Visit 12 treatment effect for CDEIS score

The adjusted analysis takes into account the stratification variables centre and smoking status, and also adjusts for previous treatment with MP or azathioprine. The visit 12 treatment effect compares MP with placebo (reference). An effect of < 0 indicates a treatment effect in favour of MP.

Analysis type	Visit 12 treatment effect	95% CI limit		p-value
		Lower	Upper
Adjusted	–0.612	–1.92	0.69	0.354
Unadjusted	–0.661	–1.89	0.57	0.289

Secondary outcome: faecal calprotectin levels

Faecal calprotectin results were summarised by time and treatment group both as a continuous measure and categorically. The continuous change from baseline to each time point was also summarised and is presented in Figure 6, indicating an increase from baseline across all visits in the placebo group, in contrast to a decrease from baseline across all visits in the MP group. Figure 7 shows a categorical summary of faecal calprotectin levels, indicating the proportions in both MP and placebo groups.

FIGURE 6.

Mean change from baseline in faecal calprotectin level. Faecal calprotectin results of < 20µg/g faeces have been included in the summary with a value of 0.

FIGURE 7.

Categorical summary of faecal calprotectin levels (µg/g) in faeces. (a) MP group and (b) placebo group. Percentages are determined on non-missing data only.

Faecal calprotectin levels were also assessed as a non-invasive marker of clinical recurrence of CD by incorporating it as a time-varying covariate in the Cox proportional hazards model. Tables 28 and 29 summarise faecal calprotectin levels by colonoscopy result at visits 6 and 12. The results of the statistical analysis are presented in Table 30 and indicate that faecal calprotectin levels are a significant predictor of the primary outcome. For each 100 µg/g increase in faecal calprotectin, the probability of reaching the primary outcome increased by 17.7%. These results should, however, be treated with caution as only 31 out of 42 events have been included in this analysis as a result of missing data.

TABLE 28 - Faecal calprotectin levels by colonoscopy result at visit 6

max., maximum; min., minimum; Q1, quartile 1; Q3, quartile 3; SD, standard deviation.

Colonoscopy result	Treatment		Overall (N = 240)
	MP (N = 128)	Placebo (N = 112)
Negative (< i2) (n = 99, 47.6%)
Mean	98	181	133
Median	45	110	60
SD	148	243	197
Q1, Q3	20, 90	30, 180	20, 130
Min., max.	20, 810	20, 1080	20, 1080
n (missing)	44 (13)	33 (9)	77 (22)
Positive (≥ i2) (n = 57, 35.4%)
Mean	240	353	291
Median	170	220	170
SD	305	308	308
Q1, Q3	50, 260	120, 590	90, 350
Min., max.	20, 1200	20, 1130	20, 1200
n (missing)	27 (7)	22 (13)	49 (20)

TABLE 29 - Faecal calprotectin levels by colonoscopy result at visit 12

max., maximum; min., minimum; Q1, quartile 1; Q3, quartile 3; SD, standard deviation.

i2: ≥ 5 aphthous lesions with normal mucosa between lesions, or skip areas of larger lesions or lesions confined to the ileocolonic anastomosis.

Colonoscopy result	Treatment		Overall (N = 240)
	MP (N = 128)	Placebo (N = 112)
Negative (< i2) (n = 99, 47.6%)
Mean	141	133	137
Median	45	70	60
SD	174	170	170
Q1, Q3	20, 260	20,200	20, 200
Min., max.	20, 660	20,750	20, 750
n (missing)	24 (14)	23 (6)	47 (20)
Positive (≥ i2) (n = 57, 35.4%)
Mean	146	262	200
Median	110	200	140
SD	143	203	181
Q1, Q3	50, 180	130, 270	80, 240
Min., max.	20, 530	30, 750	20, 750
n (missing)	22 (7)	19 (9)	41 (16)

TABLE 30 - Analysis of faecal calprotectin as a time-varying covariate (Cox proportional hazards model)

For each 100 µg/g increase in faecal calprotectin, the probability of reaching the primary outcome increases by 17.7%.

Model	HR	95% CI limit		p > χ2	Number of primary events
		Lower	Upper
100-unit change in calprotectin	1.177	1.082	1.282	0.0002	31

Secondary outcome: thioguanine nucleotide concentration levels

Thioguanine levels were summarised by time and treatment group both as a continuous measure and categorically following independent blind review. Figure 8 shows the mean change from baseline in TGN levels, with a slight and unexplained rise in TGN levels in the placebo arm. Figure 9 presents a categorical summary of TGN levels in the MP group at each of the five time points measured throughout the trial.

FIGURE 8.

Mean change from baseline in TGN levels. RBC, red blood cell.

FIGURE 9.

Categorical summary of TGN levels (pmol/8 × 10⁸ red blood cells). Percentages are determined on non-missing data only. RBC, red blood cell.

In a similar manner to faecal calprotectin levels, TGN was assessed as a non-invasive marker of clinical recurrence of CD by incorporating results from MP patients only as a time-varying covariate in the Cox proportional hazards model. As shown in Table 31, the corresponding time-varying covariate analysis of TGN concentrations (data available for 14 of the 16 patients receiving MP only) indicated that, for every 100-pmol/8 × 10⁸ red blood cell increase in TGN, the hazard of reaching the primary out come decreased by 20% (HR 0.800, 95% CI 0.565 to 1.132; p = 0.207).

TABLE 31 - Analysis of TGN as a time-varying covariate (patients receiving MP only) (Cox proportional hazards model)

For each 100-pmol/8 × 10⁸ red blood cells increase in TGN, the probability of reaching the primary outcome decreased by 20.0%. Fourteen out of 16 events have been included in this analysis, based on those patients receiving MP only.

Model	HR	95% CI limit		p > χ2	Number of primary events
		Lower	Upper
100-pmol/8 × 108 red blood cells change in TGN	0.800	0.565	1.132	0.207	14

Secondary outcome: changes in self-rated quality-of-life scores

Patient-reported outcome measures (as measured by IBDQ) were summarised at each visit based on observed scores and change from baseline scores for each of the four IBDQ subscales (bowel symptoms, emotional health, systemic systems and social function). Averages and totals across all subscales were summarised similarly.

In addition, the overall average IBDQ score and overall total IBDQ score were statistically analysed using a repeated measures analysis of covariance, modelling change from baseline, fitting terms for treatment, visit and the interaction between treatment and time.

For both the overall average and overall total, there were no significant differences between treatment and placebo groups across all study visits and the study overall. However, both treatment groups showed an improvement in quality of life across all study visits in comparison with baseline, with the MP group demonstrating the greater change across all visits. Figures 10 and 11 depict the results of the statistical analysis for the overall average IBDQ and overall total IBDQ, respectively.

FIGURE 10.

Change from baseline final model treatment effects and 95% CIs for overall average IBDQ.

FIGURE 11.

Change from baseline final model treatment effects and 95% CIs for overall total IBDQ.

Safety outcomes

Prespecified subgroup analyses of primary and secondary outcomes

Subgroup analyses of the primary and secondary outcomes (clinical recurrence of CD) were analysed similarly to the main analysis of primary and secondary outcomes with the addition of an interaction term for treatment and subgroup in the Cox proportional hazards model.

Primary outcome

The results show that, in the entire study cohort, smokers had a higher chance of reaching the primary outcome than non-smokers [15/55 (27.3%) vs. 27/185 (14.6%); p = 0.018]. Among smokers, 3 out of 29 on MP (10.3%) had a clinical recurrence, compared with 12 out of 26 (46.2%) on placebo, demonstrating that MP was effective at preventing postoperative recurrence in smokers (unadjusted HR 0.127, 95% CI 0.04 to 0.46) but not in non-smokers (unadjusted HR 0.898, 95% CI 0.42 to 1.94). The number needed to treat was calculated as 3 in smokers and 31 in non-smokers. Other subgroup analyses assessing previous TPMT exposure, prior methotrexate use, prior infliximab use, previous surgery and age at diagnosis did not identify any differences between the groups.

Figure 12 shows the results of the overall unadjusted analysis of the primary outcome together with the results of the individual subgroup analyses.

FIGURE 12.

Forest plot of primary outcome (unadjusted). The subgroup p-values are from a test for an interaction between the treatment and subgroup variables.

Secondary outcome

Clinical recurrence

Similar to the subgroup analyses of the primary outcome, subgroup analyses for the secondary outcome demonstrated that smokers had a higher chance of reaching the secondary outcome than non-smokers [19/55 (34.5%) vs. 55/185 (29.7%); p = 0.033]. Among smokers, 6 out of 29 on MP (20.7%) had a secondary outcome, compared with 13 out of 26 (50.0%) on placebo, demonstrating that MP was effective at preventing postoperative recurrence in smokers (unadjusted HR 0.270, 95% CI 0.10 to 0.72) but not in non-smokers (unadjusted HR 0.995, 95% CI 0.58 to 1.70). There were no significant differences in the analyses of the other subgroups. Figure 13 shows the results of the overall unadjusted analysis of the secondary outcome together with the results of the individual subgroup analyses.

FIGURE 13.

Forest plot of secondary outcome (unadjusted). The subgroup p-values are from a test for an interaction between the treatment and subgroup variables.

Endoscopic recurrence

There were no significant differences in endoscopic recurrence, as defined by the Rutgeerts score, by treatment group on subgroup analyses except in the thiopurine-naive group (38 in the MP group vs. 29 in the placebo group; p = 0.516).

Figures 14 and 15 show the results of the overall unadjusted analysis of visit 12 colonoscopy results and CDEIS scores, together with the results of the individual subgroup analyses.

FIGURE 14.

Forest plot of colonoscopy results at visit 12 (unadjusted). The percentage of patients with a positive colonoscopy is based on positive (≥ i2) and negative (< i2) colonoscopy results only. Missing and other colonoscopy results have been excluded. The subgroup p-values are from a test for an interaction between the treatment and subgroup variables. i2: ≥ 5 aphthous lesions with normal mucosa between lesions, or skip areas of larger lesions or lesions confined to the ileocolonic anastomosis.

FIGURE 15.

Forest plot of CDEIS score at visit 12 (unadjusted). The subgroup p-values are from a test for an interaction between the treatment and subgroup variables.

Context

This study is the largest randomised, double-blind study in postoperative CD reported to date. In the study cohort, analysed on an intention-to-treat basis, there is evidence of a reduction of clinical recurrence in patients on MP compared with placebo, although this finding was not statistically significant. It is notable that the treatment effect of 20% is lower than in previously conducted studies (Hanauer et al. :18 40% reduction in mild CD lesions and 75% reduction in more severe lesions; Ardizzone et al. :19 25% reduction). We judged a treatment effect of 20% appropriate because, given the side-effect profile of MP, it is arguable that a treatment effect of significantly < 20% is of limited clinical significance.

Endoscopic recurrence is an important secondary outcome that other studies have prioritised over clinical outcomes. No overall benefit of MP therapy on endoscopic outcome was identified in the present study using the prespecified criteria of a Rutgeerts score of ≥ i2. However, on post hoc analysis, fewer patients on placebo than on MP had endoscopically normal appearances at the anastomosis. The endoscopic subgroup analysis showed an intriguing difference in treatment effect between thiopurine-naive and non-naive groups, with patients previously exposed to thiopurine having greatest evidence of benefit.

The apparent discrepancy between clinical outcome measures and endoscopic outcomes is noteworthy. There are several possible explanations for this, and clearly no consensus regarding whether or not to prioritise clinical outcomes over endoscopic outcomes has emerged from previous trial data. The stringency of our primary outcome and the internal correlation with faecal calprotectin assessment and with thiopurine metabolite concentrations are arguments in favour of prioritising and relying on the clinical primary outcome. Indeed, the validity and reproducibility of a Rutgeerts score of ≥ i2 – defined in the primary study from Rutgeerts et al. ,26 but not subject to detailed replication as a predictor of clinical recurrence – has been debated.

With respect to clinical practice, the data emerging from the trial strongly highlight the importance of cigarette smoking in the pathogenesis and progression of CD, and as a target for therapeutic intervention. Not only do the data show that the effect of MP on primary outcome was of statistical as well as clinical significance in smokers, with a number needed to treat of only 3, but the analysis of factors predictive of primary outcome is also noteworthy. Only smoking habit was predictive of outcome, whereas age, sex or need for previous surgery were not implicated as risk factors for recurrence. The data underline the importance of smoking cessation in disease management, and prioritising this aspect using all available methodologies is of undeniable importance in clinical research. Previous studies have demonstrated that surgical recurrence increases with the number of cigarettes smoked each day, and that smoking cessation reduces clinical and surgical recurrence. 32,33 To date, the constituents responsible in cigarette smoke, and pathogenic mechanisms involved in this association, have not been identified.

Limitations

The present study does have a number of limitations. There was a lower primary event rate than initially anticipated: the actual primary outcome rates in the treatment group were 12.5% in the MP group compared with 23.2% in the placebo group as opposed to expected rates of 30% and 50%. Of the 240 patients recruited into the study, 37 (15%) were also withdrawn or lost to follow-up. This compares favourably to the rates reported in earlier trials. 10,18,19,34 In the present study, 80 (33%) patients discontinued therapy as a result of an adverse event [39 (30%) in the MP group vs. 41 (37%) in placebo group] and an additional 18 (8%) patients withdrew as a result of an abnormal blood test result picked up on safety monitoring [12 (9.4%) in the MP group vs. 6 (5.4%) in the placebo group]. Ardizzone et al. 19 reported treatment discontinuation rates attributable to adverse events of 15 (10.7%) in the azathioprine group and 6 (4%) in the mesalazine group, although figures on treatment discontinuation in other trials are more difficult to gauge from the published figures.

The number of discontinuations does raise questions in terms of adherence in routine clinical practice and, together with the numbers lost to follow-up, raises tantalising questions about the impact on the significance of the primary analysis. In this sense, however, the present study may reflect the real-life use of thiopurines in a postoperative population. We would argue that this trial has considerable value as a pragmatic trial in which the results and experiences of participating patients are highly generalisable to the clinic population seen not only in the UK but in much of Western Europe and North America.

Another limitation was the number of missing values in the analysis of the secondary outcomes, which could affect the overall results. However, the proportion of missing values between treatment groups was assessed and found to be similar.

There were no worrying signals with respect to adverse events in the present study. Indeed, lower than expected rates of pancreatitis and malignancies were reported. Fourteen pregnancies were reported, with 12 healthy births, one miscarriage and one abnormal birth to a mother in the placebo group. The stringent precautions for blood monitoring may well be pertinent to the safety profile demonstrated in the trial.

Strengths

The strengths of this study include the placebo arm, the assessment of both clinical and endoscopic outcomes, dose adjustment by TPMT status, the stringency of the primary end point and the attention to maintenance of blinding throughout, as well as its generalisability to routine clinical practice.

A further issue worthy of discussion and of exploration in further studies is whether or not a stronger treatment effect would have been seen if the TPMT metabolite profiles were available for dose adjustment during the trial, as is now routine in clinical practice. The data presented suggest that a significant proportion of patients may have benefited from higher doses of active drug; indeed, almost 60% had levels below the accepted lower limit of the therapeutic range.

Applicability in UK hospitals and beyond

The study population and the overall findings of the study are highly generalisable in clinical practice in the UK and indeed across Western Europe and North America.

Taken with other recent data from the POCER study investigators,13 the present study helps make progress towards a treatment algorithm in postoperative CD that may prove to be generalisable in clinical practice. We suggest that cessation of cigarette smoking emerges as the key intervention to emphasise ahead of drug therapy. In those who continue to smoke, but not non-smokers, thiopurine therapy appears to be justified in the early postoperative period, and we would now suggest the need to explore dose optimisation by metabolite testing. In thiopurine-intolerant patients, anti-TNF-α therapy may be justified. In all patients, reassessment of clinical status, endoscopic appearances and calprotectin levels in 6 months will determine need for escalation of therapy.

Reducing variance in clinical practice

The present study provides useful insights by identifying size of effect of thiopurine therapy overall, and the subgroup in which it is most effective. The identification of smoking habit as the key risk factor for recurrence is also of considerable importance in practice.

All these represent important points generalisable to clinical practice and intervention.

The present data are suggestive of utility of calprotectin levels in predicting clinical recurrence, and utility of metabolites in optimising thiopurine therapy. These data will need to be explored in further studies but both represent potential biomarkers that may help optimise intervention.

Potential risks not explored in this study

The study had a 3-year treatment period but did not have a follow-up period for recruited patients. The longer-term management of this patient population and the duration of therapy remain unexplored.

Unanticipated findings

The magnitude of effect of smoking habit on primary outcome was unexpected, and represents a key finding.

Patient public involvement

Patient public involvement in TOPPIC was limited. A lay representative formed part of the TSC and, as such, was invited to all formal meetings and provided with the TSC reports and updates with the initial meetings attended.

Implications for health care

TOPPIC was designed in 2003–4 and completed in 2015 as the largest single study of the efficacy of MP in reducing postoperative recurrence in CD. This has since remained an important question in the context of disease management over time, although the introduction of new biological therapies and the widespread use of biomarkers and therapeutic drug monitoring need to be taken into account in assessing current impact and implications.

We have defined the rate of clinical and endoscopic disease recurrence in the UK population, and have characterised the key risk factor amenable to public health intervention as smoking, while also demonstrating that other proposed risk factors do not have a significant effect in this study. The last finding is not to be underestimated – age, disease behaviour and previous exposure to thiopurines do not appear to influence outcome adversely.

From the clinical perspective, the need to educate and re-educate patients and clinicians regarding the deleterious effects of smoking on outcome is critical, as is the need to put in place all available strategies to help with smoking cessation.

From the trial itself, MP was not effective in reducing the frequency of clinical postoperative recurrence of CD overall, but the data suggest that it has clinically meaningful effect among the subgroup of patients who continue to smoke after surgery. Importantly, with the benefit of the results of therapeutic drug monitoring of TPMT levels available after the trial, it appears that many patients were underdosed per protocol, and the data generated in the trial may underestimate the potential benefit in the clinic.

Taking these data in the context of other published data in this field, we suggest that risk stratification of relapse after surgery is principally influenced by smoking habit, and would offer prophylaxis with thiopurine therapy to those individuals who continue to smoke, if the drug is tolerated. In all patients, these data, and the data from the POCER study, make the case for monitoring faecal calprotectin levels and offering endoscopic assessment within 6–12 months of surgery to those with elevated calprotectin values, or clinical concern of recurrence. At this stage, the introduction of thiopurine therapy or biological therapy needs to be assessed in terms of the presence or absence of symptoms, and endoscopic features.

Recommendations for research

A number of areas in which further research is required emerge from the present study.

• Although the present study defines the role of MP in postoperative prevention, we also provide evidence that the effect might be underestimated in the present study.

• A further trial designed to incorporate concomitant real-time measurements of faecal calprotectin levels, TPMT metabolites and other biomarkers in the clinic, and using the combination of these parameters to optimise and target therapy on appropriate individuals, is likely to provide a more accurate estimate of the full therapeutic potential of thiopurines in the postoperative setting.

• Most importantly, there is a research need to explore the efficacy of other agents as monotherapy or co-therapy with thiopurines – anti-TNF-α therapies, vedolizumab and ustekinumab will all be available in this context in 2017, with other new oral agents in advanced stages of analysis. This area is likely to require specific trials, incorporating the understanding generated in the present study.

• The trial has also demonstrated the need for research to identify the optimum biologically meaningful outcome measure in studies of postoperative recurrence, with discrepancies between the results when clinical, biochemical and endoscopic end points are considered.

• With TOPPIC strongly highlighting the effect of smoking on recurrence rates, the data demonstrate the need for directed intervention studies to assess the relative success of alternative strategies in smoking cessation in patients with CD.

• The trial also highlights the need for mechanistic studies to explore the very well-documented association between smoking and development, with epigenetic, immune-mediated mechanisms currently of particular interest.

• Exploration of efficacy of other agents in preventing postoperative relapse – these will include anti-TNF-α therapies and anti-adhesion strategies.

Contributions of authors

Professor Jack Satsangi (Consultant in Gastroenterology, Division of Gastroenterology) was the chief investigator. He devised the study, was involved in the design and conduct of research, and was a member of the trial development group, project management group and the TSC. He was also involved in the development and refinement of the protocol, the conduct of the research interpretation and the reporting of results, and he revised the manuscript for important intellectual content.

Dr Nicholas A Kennedy (Consultant Gastroenterologist and Honorary Clinical Senior Lecturer) was a research fellow at the Edinburgh site. He was involved in the conduct of the research interpretation and the reporting of results, including the analysis of the predictive data and samples, and he revised the manuscript for important intellectual content.

Dr Craig Mowat (Consultant in Gastroenterology, Gastroenterology Department) was the principal investigator at the Dundee site. He was involved in the design and conduct of research, and was a member of the trial development group and the TSC. He was also involved in the development and refinement of the protocol, the conduct of the research interpretation and the reporting of results, and he revised the manuscript for important intellectual content.

Dr Ian Arnott (Consultant in Gastroenterology, Division of Gastroenterology) was the principal investigator at the Edinburgh site. He was involved in the design and conduct of research, and was a member of the trial development group, project management group and the TSC. He was also involved in the development and refinement of the protocol, the conduct of the research interpretation and the reporting of results, and he revised the manuscript for important intellectual content.

Mrs Catriona Keerie (Statistician) was the study statistician who wrote the statistical analysis plan and undertook the end-of-study analysis. She prepared the final study statistical report and was involved in the interpretation and reporting of results, and the drafting and revision of this manuscript for important intellectual content.

Dr Steff Lewis (Reader in Medical Statistics) was the lead study statistician and was involved in the design and conduct of research, the TSC, the interpretation and reporting of results, and the drafting and revision of the manuscript for important intellectual content.

Dr Holly Ennis (Trial Manager from August 2013) was involved in the day-to-day trial management, data management and was a member of the TMG.

Publication

Mowat C, Arnott I, Cahill A, Smith M, Ahmad T, Subramanian S, et al. Mercaptopurine versus placebo to prevent recurrence of Crohn’s disease after surgical resection (TOPPIC): a multicentre, double-blind, randomised controlled trial. Lancet Gastroenterol Hepatol 2016;1:273–82.

Data sharing statement

All available data can be obtained from the corresponding author via the ECTU in accordance with the ECTU Data Sharing Policy. This policy is currently under development, but will be based on the MRC Hubs for Trials Methodology Research ‘Good Practice Principles for Sharing Individual Participant Data from Publicly Funded Clinical Trials’, published in April 2015. 36

Disclaimers

This report presents independent research. The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, the MRC, NETSCC, the EME programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the EME programme or the Department of Health.