Comparison Of iNfliximab and ciclosporin in STeroid Resistant Ulcerative Colitis: pragmatic randomised Trial and economic evaluation (CONSTRUCT)

Trial sites

Via the British Society of Gastroenterology we asked consultant gastroenterologists to express interest in taking part in the study and to complete a questionnaire. We considered sites that had treated four or more patients with steroid-resistant UC in the previous 12 months to be eligible and invited them to seek local approval. As a result we initiated the trial in 67 NHS trusts or health boards, covering both teaching and district general hospitals in England, Scotland and Wales. We phased the initiation of these sites, reflecting the time needed to gain local research and development (R&D) approval.

Participants in cohort and trial

The target population for the cohort were inpatients with ASC, known or suspected, who were potentially eligible for the trial. The target population for the trial were cohort members who failed to respond to a course of about 2–5 days of intravenous steroid medication, but did not then need surgery. We invited eligible patients into the cohort as soon as feasible and into the trial once the clinical team had confirmed steroid resistance. The treatment of patients who did not consent to cohort or trial did not change in any way.

Participants in qualitative study

We used purposive quota sampling to identify 12 representative consenting participants from each arm of the trial for interview on two occasions (see Appendix 1). To include participants from sites starting later in the trial pro rata, we maintained a list of eligible participants.

We also interviewed principal investigators (PIs) and nurses responsible for administering and monitoring the drugs across the trial sites. We included both professions to explore drug administration, physical effects of both drugs on patients, personal preferences and their participation in CONSTRUCT. We used purposive sampling to recruit 12 PIs from three strata: sites that recruited well to cohort and trial; sites that recruited well to cohort but less well to trial; and sites that recruited poorly to both cohort and trial. However, we recruited all eight nurses from good recruiting sites as others would have little relevant experience of the trial drugs.

Informed consent

Patients eligible for the cohort received cohort participant information sheets (see Appendix 2) and oral explanation from consultant gastroenterologists or research professionals (usually research nurses); in response, patients gave written consent by signing and dating a cohort consent form (see Appendix 3). Cohort participants who became eligible for the trial received the patient information sheet (RCT) (see Appendix 4), and gave written consent by signing and dating a trial consent form (see Appendix 5). For both cohort and trial, those taking consent countersigned and dated the form to confirm that the participant had fully understood the nature of the study and had an opportunity to ask questions; they also put a copy of that consent in the participant’s medical record and gave another copy to the participant.

Research professionals could take consent to the cohort if authorised to do so on the site delegation log following appropriate training, including in good clinical practice (GCP). Although they could also explain the trial to cohort patients, responsibility for countersigning lay with the site PI or another doctor with delegated authority on the site delegation log. The trial consent form included consent to take part in qualitative interviews.

Withdrawal

The procedure for consenting participants stressed that they could withdraw from the cohort or trial whenever they wished without giving a reason and without affecting their care in any way. However, we documented any reasons given. Participants could stipulate the level of withdrawal: from the allocated treatment, from completion of further participant questionnaires, from consent to the use of any of their routine, or from any combination of these. We also encouraged site staff to trace participants lost to follow-up and document reasons when possible.

Between randomisation and the end of the trial, there were decisions to change allocated treatments, but these did not require withdrawal from this pragmatic trial. These included failure to respond to treatment with infliximab or ciclosporin, usually leading to surgical intervention.

Trial inclusion criteria

Patients admitted as emergency admissions with severe colitis (according to Truelove and Witts,4 a Mayo score of at least 2 on endoscopic finding or clinical judgement) who fail to respond to about 2–5 days of intravenous hydrocortisone therapy, who also had either:

• histological diagnosis of UC in this episode

• histological diagnosis of indeterminate colitis in this episode when clinical judgement (based on macroscopic appearance, disease distribution or previous history) suggested a diagnosis of UC rather than Crohn’s disease

• symptoms typical of UC awaiting histology; or

• history of UC confirmed histologically.

Trial exclusion criteria

• Aged < 18 years of age on admission.

• Histological diagnosis before randomisation inconsistent with UC.

• Enteric infection confirmed before randomisation by stool microscopy, culture or histology (including Salmonella, Shigella, Clostridium difficile, Campylobacter and cytomegalovirus).

• Vulnerable patient.

• Unable to consent.

• Positive pregnancy test or current lactation.

• Woman of childbearing potential unwilling to use contraception during and for 6 months after treatment with infliximab in accordance with the Summary of Product Characteristics (SPC).

• Current malignancy, except basal cell carcinoma.

• Serious comorbidity, including immunodeficiency, myocardial infarction within last month, moderate or severe heart failure (New York Heart Association class III or IV), acute stroke within last month, respiratory failure, renal failure, hepatic failure and active or suspected tuberculosis.

• Other severe infections including sepsis, abscesses and opportunistic infections.

• History of hypersensitivity to infliximab, ciclosporin or polyethoxylated oils (notably Sandimmun Concentrate for Solution for Infusion).

• Current use of tacrolimus or rosuvastatin.

• English not good in absence of local translator.

• Need for emergency colectomy without further medical treatment.

• Currently taking part in another clinical trial.

• Treatment with either infliximab or ciclosporin in the 3 months before admission.

• Any other contraindication to treatment with infliximab or ciclosporin.

Sample size and power

Our original target analysable sample size was 360 participants, based on a primary outcome of a change in HRQoL over 2 years. 54 However, in 2012 slower recruitment than predicted led us to seek agreement from the Trial Steering Committee (TSC), Data Monitoring and Ethics Committee (DMEC) and Health Technology Assessment (HTA) programme to revise the primary outcome and reduce the analysable sample size to 250. We also proposed to analyse the area under the curve (AUC) of scores on the CUCQ collected every 6 months up to 3 years from randomisation; and to include participants who had undergone colectomy by developing and validating a post-colectomy extension to the CUCQ, which we termed the Crohn’s and Ulcerative Colitis Questionnaire with post-colectomy extension (CUCQ+).

The changes required statistical imputation to exploit the resulting data set. As these techniques are difficult to incorporate into power calculations, we used a simpler calculation based on t-tests of mean CUCQ scores at 12 months. To detect an effect size of 0.35 in these scores with 80% power when using a 5% significance level, required that we analyse at least 250 trial participants. We used the techniques of statistical imputation applied successfully by the Cancer of the Oesophagous or Gastricus: New Assessment of the Technology of Endosonography58 and Folate Augmentation of Treatment – Evaluation of Depression (FolATED)59 trials to achieve an effective sample size of 250 for CONSTRUCT.

Randomisation

We allocated at random between infliximab or ciclosporin all participants who completed baseline assessment, met the trial inclusion criteria and gave informed consent. We used a password-protected website that accessed an adaptive algorithm to protect against subversion while ensuring that each trial arm was balanced by centre. 60 To validate each request for randomisation, the website asked:

• for the participant’s trial number, and month and year of birth

• for the name of person requesting randomisation (limited to those trained and authorised)

• if consent had been given

• if the participant had met the inclusion criteria

• if the participant had none of the exclusion criteria

• if the baseline questionnaire been completed.

If the responses to all four questions were ‘yes’, the website gave the name of the drug allocated to the participant and immediately confirmed the trial number and drug by e-mail, and recorded those on the randomisation database.

Hospital pharmacies at trial sites held the trial drugs. After each randomisation, the research staff confirmed study number and drug by fax to the relevant pharmacy who labelled the drug with the European Union Drug Regulating Authorities Clinical Trials number, sponsor, participant’s trial number, name and address of supplier, dose and ‘For Clinical Trial Use Only’.

Primary outcome

The primary outcome was the AUC of CUCQ scores. This is equivalent to QAS weighted by scores on the disease-specific CUCQ. We concurrently validated the CUCQ, which extends the validated UK-IBDQ61 to cover acute illness and colectomy.

Secondary outcomes

1. (a) Disease-specific QoL, measured by the CUCQ.

2. (b) and (c) Generic QoL, measured by the SF-12. 62

3. (d) Mortality.

4. (e) Colectomies, both emergency and planned.

5. (f) AEs.

6. (g) Readmissions, including those for causes other than UC.

7. (h) Malignancies.

8. (i) Serious infections.

9. (j) Renal disorders.

10. (k) Disease activity, using the criteria proposed by Truelove and Witts. 4

Economic outcomes

1. (l) NHS costs.

2. (m) HRQoL, measured by EQ-5D. 63

3. (n) Participants’ time off work.

Qualitative outcomes

1. (o) Participants’ views of their drugs and their consequences.

2. (p) Professional views of both drugs and their consequences.

Definitions

The Medicines for Human Use (Clinical Trials) Regulations 2004: SI 2004/1031 describe thus:64

• AE: any untoward medical occurrence in a subject to whom a medicinal product has been administered, including occurrences which are not necessarily caused by, or related to, that product.

• AR: any untoward and unintended response to a medicinal product which is related to any dose administered to the subject.

• Unexpected adverse reaction: an adverse reaction, the nature and severity of which is not consistent with the information in the SPC for the medicinal product.

• SAE or serious adverse reaction (SAR) or suspected unexpected serious adverse reaction (SUSAR): any AE, adverse reaction or unexpected adverse reaction, respectively, that:

  1. results in death

  2. is life-threatening

  3. requires hospitalisation or prolongation of existing hospitalisation

  4. results in persistent or significant disability or incapacity; or

  5. is a congenital anomaly or birth defect.

Causality

Causality is the degree to which an untoward medical occurrence can be attributed to the trial intervention rather than the underlying UC. We used the subjective scale: unrelated, unlikely to be related, possibly related, probably related or definitely related. We classed only the last three as ARs or SARs having a causal relationship. However, we did not class events caused by UC, notably worsening colitis or initial steroid resistance, as ARs. We reported SARs to the Medicines and Healthcare products Regulatory Agency (MHRA) annually.

Expectedness

We considered AEs, ARs, SAEs and SARs as ‘unexpected’ if their nature and severity was not consistent with the relevant SPC. Expected events included:

• progression or exacerbation of the participant’s underlying UC, including clinical sequelae of progression, such as worsening diarrhoea or abdominal pain

• medical or surgical procedures including surgery and endoscopy; however, we considered events that led to procedures and surgical sequelae, such as pleural effusion or small bowel obstruction, separately

• conditions or symptoms present or detected before the first dose that did not worsen

• recognised undesirable effects found in the current SPCs for Remicade, Sandimmun or Neoral; as SPCs updated regularly during the trial, we recommended that PIs consult the versions online.

Monitoring adverse events

In view of the extensive side effects of both trial drugs documented in their SPCs, and the many sequelae of disease progression, the protocol stipulated that we did not require expedited reporting of serious events from sites, unless they were unexpected, in which case we asked for notification within 24 hours.

We designed adverse event screening forms (AESFs) to enable local PIs to assess seriousness, causality and expectedness in a logical sequence (see Appendix 6). Sites sent AESFs, once countersigned by local PI or authorised person, securely via FaxPress to the trial office. The CONSTRUCT data manager was responsible for initial screening of AESFs, paying particular attention to completeness, raising queries with the local team and immediately notifying the chief investigator of events that might be classified as SUSARs. If uncertain, the chief investigator discussed them with the local PI before the final decision. We entered screened AESFs onto our Generic Clinical Information System (GeneCIS; version 10i, Swansea University, Swansea) and regularly checked that they were consistent with data on colectomies reported through case report forms (CRFs). Clinicians within the trial team reviewed the accumulating data on AEs, and commented on GeneCIS, if necessary after discussion with the local team.

At the end of the study, two clinicians in the trial team reviewed all SAEs to ensure consistency of interpretation in the final report, informed by data from sites on the duration of use of the drugs beyond the 3-month intervention period specified in the protocol. Once that period was complete, clinical management varied from participant to participant according to their progress and the clinical judgement of the local team. The reviewers also took the different pharmacokinetic profiles of the trial drugs into account because the bio-availability of ciclosporin is short-lived, whereas bio-availability of infliximab can persist until 6 months after the last infusion. With the exception of malignancy, therefore, we judged relatedness unlikely if the event occurred more than 1 month after ciclosporin, or more than 6 months after infliximab.

Problems caused by UC and the trial drugs may coincide in a single event. When an AESF documented a serious event, usually admission, and included more than one problem (e.g. abnormal liver and renal function tests), we used clinical judgement to decide which related to the drug and which was the prime cause of admission, and classified SAEs and reactions by body system.

Baseline and follow-up data

At recruitment to cohort, we collected:

1. sociodemographic details: age, sex, ethnic group and truncated post codes, used to generate measures of social deprivation (Indices of Multiple Deprivation for England, Welsh Index of Multiple Deprivation, Carstairs Deprivation Scores for Scotland and Townsend scores for all three countries)

2. details of admission

3. disease history, including presenting complaint, time since first diagnosis and previous treatment, including any previous surgery, or biologic or steroid therapies

4. comorbidities, in particular cardiorespiratory, liver or renal disease, diabetes mellitus or hypertension

5. UC signs and symptoms, including duration of symptoms in current episode, stool frequency, blood pressure, pulse and temperature

6. current treatment, including type, dose and duration of steroid therapy

7. pathology results, including full blood count, inflammatory markers, liver and renal function tests, and total cholesterol

8. site and extent of disease according to Montreal classification of IBD65

9. histopathology results, including stool culture and histological diagnosis

10. family history of IBD

11. height, weight and smoking status.

At 3 and 6 months after randomisation, and at 6-monthly intervals until 36 months, we collected clinical and resource-use data from participants’ records (see Appendix 7).

Process of data collection

Research staff at trial sites asked all trial participants to complete the Participant Baseline Questionnaire including the CUCQ, SF-12, EQ-5D and questions on primary care resource use (see Appendix 8); and the similar Participant Follow-up Questionnaire (PFQ) which also included questions about intercurrent events at 3, 6 and 12 months, and at 18, 24, 30 and 36 months if they reached these time points before March 2014 (see Appendix 9). Following admission a baseline CRF was completed by research staff to document demographic and clinical details (see Appendix 10).

We asked trial sites to arrange outpatient appointments to coincide with these times whenever compatible with routine clinical management. Local research staff posted PFQs to participants and asked them to bring completed questionnaires to clinics. Participants could either complete and return them by post or seek help at the clinic. We also asked sites to complete CRFs recording intercurrent events, secondary care resource use and drug treatment at these times, preferably at the outpatient appointments, or else from medical records (see Appendix 11).

The trial office sent Post-Colectomy Questionnaires (PCQs) and business reply envelopes to trial participants who underwent colectomy on discharge following surgery and at 4, 8 and 12 weeks thereafter (see Appendix 12). This PCQ included the CUCQ+, the post-colectomy version of the CUCQ and the EQ-5D.

Electronic data capture

We captured designed data on our existing GeneCIS, based on the Oracle(R) object-relational database management system (RDBMS version 10g, Oracle Corporation UK Ltd, Reading). The system is implemented in a three-tier architecture and remotely hosted in a professionally managed, secure environment. With support from local information technology departments, we provided access to hospital sites over the NHS N3 network.

We customised GeneCIS to support the trial following a detailed evaluation of requirements, including process mapping. The resulting electronic data capture structure reflected trial data requirements organised in a clinically logical manner. GeneCIS includes data validity controls, including predefined pick lists, and format and range constraints. We provided contextual guidance as help text and used system alerts to warn users when specific combinations of entered data affected eligibility, the potential for AEs or other important items.

Sites could choose either to enter data collection forms into GeneCIS locally or send completed forms by secure FaxPress to the trial office. Sixteen sites used local data entry; the rest faxed paper forms. The basic system and user interface were identical at sites and the trial office.

Data management and record keeping

All data acquisition, storage, transmission and use complied with the Data Protection Act. 66 The trial office recorded forms received on a bespoke Microsoft Access^® 2007 database (Microsoft Corporation, Redmond, WA, USA) for tracking paperwork, and stored all forms in locked cabinets within a secure office with controlled access. We backed up GeneCIS and Access databases every day.

All GeneCIS users had role-based access to the system, including trial staff engaged in system configuration, data entry, helpdesk support or quality assurance. Those with authorised access to identifiable data did not contribute to analysis. Qualitative researchers had access to identifiable data to contact participants, but not to their clinical data. Trial sites had access only to identifiable data for their participants and could not view any other records.

We extracted data for analysis in pseudonymised form identified by participants’ trial numbers.

Data quality and information governance

We asked PIs to maintain site delegation logs authorising staff to perform defined tasks, to sign off any changes and to send them to the trial manager. When we updated trial documents including CRFs, we circulated new versions to sites with appropriate instructions. Sites replaced previous versions but retained them in trial site files.

We asked those making essential changes to data collected on paper to strike the original entry only once, insert the new data, and record the date and initials of the person responsible. GeneCIS maintained an electronic audit trail by annotating all data with the user, date and time of entry. It also checked data at entry, notably for correct format and within specified ranges. We subjected the resulting data to rigorous quality assurance, compared all anomalies with the paper source documents and consulted local staff when necessary.

We transcribed and reviewed interview data as soon as possible. We analysed the final transcripts without identifying patient, professional or hospital.

Routine data

Funded research data collection continued until March 2014. We plan to supplement the designed baseline data collected from cohort and trial participants with routinely collected data, including Hospital Episode Statistics (and the equivalent in Wales and Scotland), and mortality from the Office for National Statistics. We plan to continue follow-up for up to 10 years by linking routine mortality, inpatient and primary care data. Record linkage will use the existing facilities of the Farr Institute at Swansea University Medical School. Using these information sources, we plan to monitor all participants’ long-term outcomes, notably mortality, colectomies both emergency and elective, and major morbidity including hospitalisation and surgery, and thus most of their NHS costs. Hence we aim to achieve long-term follow-up of both trial participants and a large comprehensive cohort of patients with UC.

Trial participant interviews

To understand trial participants’ experiences and perceptions of treatment by infliximab, ciclosporin and surgery, we conducted telephone interviews at about 3 and 12 months after recruitment. We aimed to investigate their priorities for their health and well-being, and their perceptions of taking the drugs, side effects and response to treatment.

We first sought consent to interview when recruiting participants to the CONSTRUCT cohort. We sampled interviewees from those subsequently recruited to the trial. Hence they had all been admitted with ASC and received either infliximab or ciclosporin. We contacted those sampled following their first outpatient appointment after discharge, invited them to take part and arranged convenient 1-hour slots to enable them to talk freely without feeling rushed. As more than 50 trial sites across the UK contributed participants, interviews by telephone to participants’ homes enabled us to cover the trial population, minimise cost and ensure confidentiality.

Trained qualitative researchers used a semistructured approach to guide participants through the interview questions and give them the opportunity to develop their responses and raise issues that were important to them. The opening questions of the first interview schedule encouraged participants to think and talk about their health, what was important about it, and what good and bad health meant to them (see Appendix 13). More specific questions followed about their experiences of treatment, drug regimes and health care, and about their outcomes. We soon recognised that some participants had already required surgery for their colitis. We therefore adapted the interview schedule to capture the views of these participants (see Appendix 14).

At the end of each of these initial interviews, we asked participants to take part in a second interview about 12 months later to explore their subsequent experiences. For those, we used similar interview schedules but added questions, notably to explore changes in their health, in their opinions of treatment and in their interactions with health-care professionals (see Appendix 15). Again, we arranged convenient 1-hour slots for telephone interviews.

Health-care professional interviews

These aimed to gain insight into professional preferences between the two drugs and their personal contribution to the trial. Our specific objectives were to explore their views about:

• administration of the two drugs, including ease of handling

• effects of the drugs on health-care provision

• personal preferences between the drugs

• drug regulation and current policy

• surgery for UC

• equipoise in recruiting to the trial.

We therefore planned semistructured interviews lasting 30–40 minutes with flexibility for interviewees to expand on important issues and access their broad knowledge base. Our schedule covered interviewees’ beliefs and ways of working; aspects of drug provision that may affect preferences; their interaction with patients and others; and their contribution to the trial (see Appendices 16 and 17). We offered interviews face to face or over the telephone. We conducted separate interviews with consultants and nurses to provide a richer understanding of differences between professional groups. With participants’ consent we recorded and transcribed interviews for analysis.

Background

We are entirely committed to the philosophy, expounded by NICE and implemented by the National Institute for Health Research (NIHR), that the ultimate criteria for interventions in health care are clinical effectiveness and cost-effectiveness in improving the survival and QoL of patients over extended periods. Although we admire how the GETAID investigators implemented the CySIF trial with little funding, they recruited only 110 patients, followed them for only 98 days and reported only on ‘treatment failures’. This is not a good basis for NHS investment decisions worth hundreds of millions of pounds. Instead we chose QAS as our primary outcome and recruited enough participants to yield the power to discriminate between two contrasting drugs. That still left us with the task of developing and validating a comprehensive patient-reported outcome measure (PROM) applicable across the broad spectrum of disease severity within UC.

There are several disease-specific QoL measures for patients with IBD. 18,67–71 The most widely used is the McMaster Inflammatory Bowel Disease Questionnaire. 18,67,68,72,73 As it had not been validated for the UK, we previously conducted a development and validation study to anglicise it as the UK-IBDQ. 61 As this was a community study, however, we recognised in designing CONSTRUCT within acute care that, to avoid ceiling effects, we had to modify the UK-IBDQ by extending the range of responses for all items to include more severe replies; the number of items to include more questions addressing severe symptoms; and the frequency of administration in participants undergoing surgery. Furthermore, the UK-IBDQ has no questions about the impact on QoL of colectomy, which reportedly happens to 9–21% of those with UC. 74–76 We therefore derived from the UK-IBDQ a PROM for patients suffering from moderate to severe symptoms of Crohn’s or UC with or without stoma. We called this the CUCQ in its basic form, and the CUCQ plus stoma extension (CUCQ+) in its extended form. Preliminary validation of the CUCQ confirmed that it met essential psychometric criteria,77 with the result that the national IBD registry adopted it. Although we developed the CUCQ as a tool for both conditions, we validated it only on UC patients. As the development and concurrent validation of CUCQ and CUCQ+ underpin the primary analysis of CONSTRUCT, we now describe that validation.

Methods

We used the standard psychometric approach outlined by Streiner and Norman78 to develop and validate the CUCQ and CUCQ+ in three stages:

• item generation

• initial development in the CONSTRUCT cohort

• definitive validation in the CONSTRUCT trial sample.

We used the 32 UK-IBDQ items [see Appendix 8, Section A: Crohn’s and Colitis Questionnaire (CCQ)], as the basis for developing the CUCQ and CUCQ+. We also reviewed the literature on PROMs in gastroenterology to identify additional items. After drafting the CUCQ and CUCQ+, we recruited an expert panel of gastroenterologists, outcome specialists, statisticians and patients to review the resulting questions and response options, and ensure they were appropriate for UC patients.

We piloted the draft questionnaires on a sample of 20 UC patients with or without stoma from Neath Port Talbot Hospital, Port Talbot, who were not participating in CONSTRUCT. We asked them to complete CUCQ or CUCQ+ as appropriate and added four supplementary questions:

• Did you find any of the questions difficult to understand?

• Was there any question you did not want to answer?

• Was there any aspect of your bowel condition not covered by these questions?

• Did you find any of these questions not applicable to you?

For our initial validation, we used patients recruited to the CONSTRUCT cohort but not the trial, who had therefore completed the CUCQ at baseline. Of the 32 CUCQ questions (see Appendix 18), six were not relevant to post-colectomy participants:

• Q1: On how many days over the last 2 weeks have you had loose or runny bowel movements?

• Q2: On how many days in the last 2 weeks have you noticed blood in your stools?

• Q6: On how many days over the last 2 weeks have you opened your bowels more than three times a day?

• Q9: On how many days over the last 2 weeks have your bowels opened accidentally?

• Q24: On how many days over the last 2 weeks have you wanted to go back to the toilet after you thought you had emptied your bowels?

• Q26: On how many days over the last 2 weeks have you had to rush to the toilet?

We therefore designed the CUCQ+ (see Appendix 9, For patients with stoma), with 10 stoma-specific questions replacing these six questions, making a total of 36 questions.

• S1: On how many days over the last 2 weeks have you been afraid that other people might hear your stoma?

• S2: On how many days over the last 2 weeks have you been worried that other people might smell your stools?

• S3: On how many days over the last 2 weeks have you been worried about possible leakage from your stoma bag?

• S4: On how many days over the last 2 weeks have you had problems with care for your stoma?

• S5: On how many days over the last 2 weeks have you found the skin around your stoma irritated?

• S6: In the last 2 weeks have you felt embarrassed because of your stoma?

• S7: In the last 2 weeks have you felt less complete because of your stoma?

• S8: In the last 2 weeks have you felt less attractive as a result of your stoma?

• S9: In the last 2 weeks have you felt less feminine/masculine as a result of your stoma?

• S10: In the last 2 weeks have you been dissatisfied with your body as a result of your stoma?

To calculate scores for participants who had not undergone surgery, we used the 32 CUCQ questions. For post-colectomy participants, we used the 10 stoma-specific questions and the 26 stoma-relevant CUCQ questions to calculate CUCQ+ scores. In analysing and validating both CUCQ and CUCQ+, we calculated scores as follows:

1. We scored questions with four responses as 0, 1, 2 or 3 in ascending severity.

2. We scored questions with responses between 0 and 14 days as the actual value.

3. We reversed the scoring of questions with wording in the reverse direction (Q7, Q22 and Q32) to code all questions in the same direction.

4. We rescaled questions between 0 and 1 by dividing actual responses by their maximum score (3 or 14).

5. We calculated CUCQ scores for non-colectomy participants, and CUCQ+ scores for post-colectomy participants, by summing all valid responses and dividing by the number of completed questions.

6. So the lower the CUCQ+ (or CUCQ score), the better the respondent’s health. In analysing the AUC, however, it is necessary to subtract the total CUCQ+ (or CUCQ) score from 1, so that higher scores show better health, consistent with both EQ-5D and SF-12.

However, we calculated CUCQ and CUCQ+ scores only when participants had responded to at least 75% of the questions – 24 out of 32 or 27 out of 36, respectively. If participants had completed fewer than 75% of questions, we treated the total CUCQ or CUCQ+ score as missing. To give equal weight to each question answered, our analysis plan used the original 32 CUCQ questions for non-colectomy participants and the 36 CUCQ+ questions for post-colectomy participants. To test the sensitivity of our findings to this simple approach, we repeated the analysis in two ways. First, we gave equal weight to core scores and stoma-specific scores, thus giving more weight to the latter. Second, we gave the stoma-specific scores 6 out of 26 of the weight given to core scores, as the original UK-IBDQ comprised 26 questions relevant to stomas and 6 questions inapplicable to stomas, thus giving less weight to stoma-specific scores.

Still following Streiner and Norman,78 we conducted initial psychometric development of the CUCQ on CONSTRUCT cohort patients (who had not had a colectomy):

1. We examined the 32 sets of response frequencies for floor or ceiling effects.

2. We calculated the Kaiser–Meyer–Olkin measure of sampling adequacy and Bartlett’s test to judge whether or not principal component analysis was appropriate.

3. We calculated Cronbach’s alpha (which should exceed 0.7 for good internal consistency).

4. We calculated item-total correlations for each question (which should exceed 0.2 for good homogeneity).

5. We undertook principal component analysis to assess the underlying structure; we considered factors important if their eigenvalues were clearly > 1, and individual questions as useful if their factor loadings exceeded 0.4.

6. We assessed the construct validity of the scale by examining the correlation between the CUCQ and two generic QoL questionnaires (EQ-5D and SF-12).

Then we undertook definitive validation of the CUCQ and CUCQ+ on the CONSTRUCT trial sample. First, we compared demographic characteristics of that sample with the cohort to ensure that they were similar. Then we analysed the trial sample in essentially the same psychometric way as the cohort, and tested whether or not the principal components arising from these two analyses were consistent. Finally, we repeated the principal component analysis of the trial sample after 12 months and tested whether or not principal components arising from the CUCQ for non-colectomy patients and those arising from the CUCQ+ for post-colectomy patients were consistent.

We also analysed reliability and responsiveness for the trial sample, initially combining CUCQ and CUCQ+ and then comparing them. We assessed test–retest reliability of the scales on participants who reported no change in their condition at successive assessments; we considered scales reproducible if intraclass correlation exceeded 0.75. We assessed responsiveness of the scales to change on participants who reported a change in their condition; we considered scales responsive if the responsiveness ratio exceeded 0.5.

Clinical effectiveness

Primary analysis was by treatment allocated, reflecting the pragmatic nature of the trial design. Figure 1 illustrates the primary outcome measure – the AUC defined by CUCQ scores at baseline and all available time points before 31 March 2014 – also known as ‘QAS’. Within research into PROMs, there is a general convention that for generic PROMs ‘higher is better’ whereas for condition-specific PROMs ‘lower is better’. However, in order to be consistent with both EQ-5D and SF-12 we have subtracted calculated CUCQ scores from 1, so that higher is better.

FIGURE 1.

Primary outcome measure: area under the CUCQ curve.

We calculated this area by summing the areas of the component trapezia, thus assuming linearity between successive CUCQ scores, and setting the score at the end of follow-up equal to the last recorded score (last one carried forward). This method accommodates both missing values and extra values like those arising from PCQs.

The main analysis used a general linear model to estimate differences in QAS between groups, adjusting for covariates that may affect this measure, including trial site; data collected while assessing eligibility for cohort and trial, notably the sociodemographic variables age, gender, ethnic group, QoL, disease severity; current immunosuppressant therapy (using a binary indicator set equal to 1 for participants taking azathioprine, 6-mercaptopurine or methotrexate); and time in follow-up. We combined rare categories in factors like ethnic group, taking account only of observed numbers in each category and the coherence of new groupings.

Secondary analyses adjusted for the same covariates as primary analysis and compared between groups: QAS per day (again using general linear models); QoL scores (using methods for repeated measures); proportion of participants undergoing colectomy (using binary logistic regression); time to colectomy (censored at the end of follow-up, and analysed by Cox regression); proportion of participants suffering one or more AEs (using binary logistic regression); and mortality.

We examined residual diagnostics in analyses that assume normality, with the options of data transformation and bootstrapping when residual distributions were markedly non-normal. We excluded identified outliers and reanalysed the revised data sets. We supplemented analyses by descriptive comparisons between groups in accordance with Consolidated Standards of Reporting Trials (CONSORT) guidelines, including estimates with 95% confidence intervals (CIs), representing two-tailed tests at the 5% significance level.

Imputation of missing data

We used statistical imputation of censored and missing data to impute costs and QoL for all participants who generate data on survival, colectomy or QoL after randomisation. Thus we excluded participants without follow-up data, as calculation of QAS requires one or more CUCQ scores in follow-up.

None of the questionnaires has an official algorithm for imputing missing answers. So we imputed missing values within participant interviews using the Expectation Maximisation option in the Missing Value Analysis module within Statistical Package for the Social Sciences version 22 (SPSS Inc., Chicago, IL, USA), and calculated scores according to the instructions for the measure in question. When necessary we imputed missing scale and subscale scores by regression using the available values of that score at other times and the allocated treatment group. To avoid introducing outliers, we restricted imputed scores to the range observed for that measure in that group.

Step 1: identify, categorise and code effects

We derived the potential effects of CONSTRUCT from its aims and added effects that emerged over the course of the trial; we grouped them in three columns:

1. on participants

2. on gastroenterology services and professionals

3. on the rest of NHS and society.

We populated layer 1 with these effects using one cell per effect under the three defined columns, and numbered each effect.

Step 2: code methods used by CONSTRUCT

We listed each method or instrument used (e.g. CRFs, questionnaires, interviews and routine data) in layer 2 and gave each method a unique letter.

Step 3: create joint alphanumeric codes

We assigned letters to each effect in layer 1 showing how we had investigated them. For example, if disease-specific QoL was effect 2 investigated by the CUCQ questionnaire labelled B, we recorded B alongside effect 2 in layer 1. Similarly, we assigned numbers to each method in layer 2 showing all the effects they had investigated. For example, we recorded effect 2 alongside method B in layer 2. Effects investigated by more than one method have extra letters in layer 1; methods investigating more than one effect have extra numbers in layer 2.

Step 4: identify and code all research findings

We listed the individual findings from each component of CONSTRUCT in layer 3 of the MATRICS proforma. We illustrated them textually or by summary statistics, and labelled them with codes derived from layers 1 and 2 to identify the corresponding effects and methods.

Step 5: synthesise complementary findings and reorder contrasting findings

We merged comparable findings into composite statements, irrespective of which effects were investigated or methods used and listed all codes associated with each. This enabled us to show that more than one effect or method had reported the same outcome. When findings were not comparable, even opposing, we kept them as separate rows in layer 3, but put them together.

Ethics and research governance

The Research Ethics Committee (REC) for Wales gave ethical approval (Ref 08/MRE09/42) and each participating trust or health board gave NHS R&D approval. The trial has European Union Drug Regulating Authorities Clinical Trials number (2008-001968-36) and clinical trial authorisation from the MHRA. We have sought to conform with the Research Governance Frameworks for England,91 Scotland92 and Wales;93 the principles of GCP outlined by the International Conference on Harmonisation (www.ich.org/), the European Union Directive 2001/20/EC94 and The Medicines for Human Use (Clinical Trials) Regulations 2004. 64

Trial management

We established:

• a TSC which oversaw the trial by meeting twice per year

• a DMEC which monitored trial data in accordance with an agreed charter (see Appendix 21) and reported to the TSC

• a Trial Management Group (TMG) comprising academics, health professionals, researchers and a representative of service users which undertook general management of the trial, met every month and reported to the TSC and funders through reports every 6 months.

The available members of the research team also met every week.

Service users

In accordance with the relevant West Wales Organisation for Rigorous Trials in Health standard operating procedure,95 we included service users as active contributors at all stages of this trial. They provided separate members of TSC, DMEC and TMG, contributed to the research process and provided valuable insights into UC. In particular they contributed to the development of the CUCQ+, the primary outcome measure; helped to maximise recruitment rates; attended the series of report writing days; and gave much valuable help and advice outside these meetings. Appendix 22 summarises the personal experience of one service user.

Before protocol publication

Since protocol publication

Item generation: devising the items for the CUCQ and CUCQ+

Following exploration of the UK-IBDQ and the gastrointestinal literature related to PROMs, we developed the CUCQ and CUCQ+. Appendix 18 shows the major differences between the UK-IBDQ and the CUCQ and CUCQ+. Question 7 from the original UK-IBDQ (related to admission to hospital) was removed from the CUCQ and CUCQ+. A new question was added to the CUCQ/CUCQ+ (Q15, related to getting up to use the toilet after going to bed). In addition, where the original UK-IBDQ had four fixed-response options relating to questions ‘in the past 2 weeks, how often . . .?’, the CUCQ/CUCQ+ used open-response options to take account of any potential floor and ceiling effects.

As the UK-IBDQ did not have any questions for post-colectomy patients, we developed 10 specific stoma-related questions for the CUCQ+ (S1–S10).

Piloting the CUCQ and CUCQ+

We did not modify the CUCQ or CUCQ+ following piloting as none of the patients regarded any of the questions difficult to answer or that there were any specific aspects that were not covered by the questionnaire. The patients with a stoma did not answer the six questions that we had previously identified as not being relevant and instead completed the additional 10 stoma questions.

Baseline demographics of the development and validation samples

We had data from a total of 1240 patients in our cohort development sample and 270 patients in our RCT validation sample (total 1510). We compared the baseline demographic characteristics of the two samples. Table 7 gives details of the characteristics of the two samples.

Psychometric validation of the CUCQ in the validation sample (CONSTRUCT cohort sample)

We examined the data prior to undertaking principal component analysis. The Kaiser–Meyer–Olkin measure of sampling adequacy was 0.924 and the Barlett’s test of sphericity was 0.000, indicating that the data were suitable principal component analysis.

All 32 CUCQ questions had a response rate of at least 93% in the cohort development sample (Table 8). Based on our set criteria of responding to at least 24 out of the 32 (75%) total questions, there were 1226 (99%) patients for whom we were able to calculate a CUCQ score.

The internal consistency of the CUCQ was excellent with a Cronbach’s alpha of 0.900. Correlations of all 32 questions with the total score exceeded 0.2 (Table 9). None of the questions had a response frequency > 80%. The data therefore did not exhibit any floor or ceiling effects.

The principal component analysis indicated that there were seven factors with an eigenvalue of > 1 and which explained approximately 57% of the variance in the data. The scree plot indicated that there was an elbow in the scree plot between the fourth and fifth factors. We rotated the solution based on four factors using direct oblimin principal component analysis (Table 10). The principal component analysis identified that the first factor covered emotional symptoms; the second bowel symptoms; the third social activities; and the fourth general symptoms.

The CUCQ scores achieved significant correlations with the two generic HRQoL scales (SF-12 mental and physical component scores and the EQ-5D) scores demonstrating good construct validity (Table 11).

Table 12 illustrates the percentage variance that each of the 32 questions contributed to the total CUCQ score in the CONSTRUCT cohort development sample. Ten of the 32 questions contributed over 90% of the variance in the total score. These 10 questions could be future candidates for a shortened version of the CUCQ. We will undertake further work to test the validity of a shortened version of the CUCQ.

Psychometric validation of the CUCQ in the validation sample (CONSTRUCT randomised controlled trial sample)

For the purposes of this report, we calculated CUCQ+ scores for stoma extension patients on the basis of equal weighting for each score. Further analysis will be carried out to explore weighting of the stoma-specific questions.

The internal consistency of the CUCQ in the RCT sample was excellent, with a Cronbach’s alpha of 0.845. Table 13 illustrates the percentage variance that each of the 32 questions contributed to the total CUCQ score in the RCT validation sample.

Table 14 indicates the number of patients for whom we collected a CUCQ/CUCQ+ score at each time period within the RCT validation sample.

The CUCQ scores achieved significant correlations at baseline with the two generic HRQoL scales (SF-12 mental and physical component scores and the EQ-5D) in the RCT validation sample scores demonstrating good construct validity (Table 15).

Baseline characteristics

We compared the trial participants in the two groups in terms of various characteristics (Table 16); there are no statistically significant differences between the groups.

Quality-adjusted survival and quality-adjusted survival per day

Figures 4 and 5 show the mean QAS and QAS per day for the two groups, together with 95% CIs for the means. For both variables, the mean for the ciclosporin group is higher than its infliximab counterpart, although the clear overlap in the 95% CIs indicate that observed differences are not statistically significant, as is confirmed in our analysis. There are relatively high values for the intracluster correlations in these variables, albeit with wide CIs (reflecting the relatively small – in statistical terms – sample sizes).

FIGURE 4.

Means and 95% CI for QAS (the QAS or area under the CUCQ curve from randomisation until end of follow-up on 28 February 2014) for the two groups.

FIGURE 5.

Means and 95% CI for QAS per day [defined as QAS divided by the time (in days) from randomisation until the end of follow-up] for the two groups.

Proportions undergoing colectomy: time to colectomy

Table 17 shows that the observed difference in the proportions of participants subsequently undergoing colectomy in the two groups is not statistically significant; nor is the difference in time to colectomy. Figure 6, the Kaplan–Meier curves of the time to colectomy in the two groups, illustrates the higher proportion of colectomies in the ciclosporin group.

FIGURE 6.

Kaplan–Meier plots showing the proportion remaining colectomy-free in the two groups over time (in days) in follow-up, censored at the end of follow-up on 28 February 2014.

Post-randomisation length of stay

Figure 7a displays box plots of the post-randomisation length of stay data for the two groups, and, together with the numerical summaries in Table 17, illustrates the skewed nature of these values. Transformed values (based on taking natural logarithms) are shown in Figure 7b, which shows that the transformation has considerably reduced the original skewness. Formal analysis shows that, although differences in the raw data are deemed to be not statistically significant, the corresponding analysis of the transformed data leads to the contrary conclusion that differences on the logarithmic scale are statistically different. As assumptions underpinning statistical models are more appropriate to the analysis of the transformed data, we should not readily accept conclusions based on the analysis of the raw data: the evidence on the equality of post-randomisation lengths of stay is thus rather more nuanced than an initial examination indicates.

FIGURE 7.

Box plots of post-randomisation LOS for the two groups. (a) Post-randomisation LOS in days; and (b) natural logarithms (Ln) of post-randomisation LOS. LOS, length of stay.

Quality-of-life profiles

Table 18 presents numerical summaries of the three QoL measures [namely CUCQ, Short Form Questinnaire-6 dimensions (SF-6D), EQ-5D] at the various stages in follow-up. As discussed elsewhere, these summaries include, for the CUCQ, values obtained from the CUCQ+ when the extended version of the questionnaire was administered to patients post colectomy, and treats these as pari passu with other CUCQ scores; however, the summaries do not include values generated via specific PCQs, as such questionnaires are not always readily associated with a specific time in follow-up. (For the avoidance of doubt, it may again here be emphasised that CUCQ+ scores obtained from PCQs are included in participants’ CUCQ profiles, and hence contribute to QAS and QAS per day.)

The profiles of the three QoL measures, shown in Figures 8 (CUCQ), 9 (SF-6D) and 10 (EQ-5D) are broadly very similar. They show an initial rise over 1 year from values at baseline followed by a levelling off; the profile for the two groups are also very similar to each other. Formal analysis, summarised in Table 18, confirms that the observed differences are not statistically significant.

FIGURE 8.

Mean CUCQ scores and 95% CI for the two groups at specific points in follow-up.

FIGURE 9.

Mean SF-6D values and 95% CI for the two groups at specific time points in follow-up.

FIGURE 10.

Mean EQ-5D values and 95% CI for the two groups at specific time points in follow-up.

Analysis of missing data

Table 31 reports the number of observations missing at each data collection time between the two arms of the trial.

The amount of missing data was greater for items reported on the PFQ which was completed by the participant as compared with the CRF which was completed by the research nurse. Participants who withdrew from the study were more likely to withdraw only from responding to further questionnaires (PFQ) than from accessing their medical records (CRF) leading to more missing data from the PFQ.

Data on trial drug administration costs were missing for 41 infliximab and 24 ciclosporin participants. Data on oral ciclosporin were missing for two participants who had completed the protocol single infusion.

We used age, gender, study arm, ethnicity, smoking, weight, and baseline SF-6D, EQ-5D and CUCQ scores to assess randomness of missing data.

For CRF data, the pattern of missing data on tests and investigations and drugs delivered to inpatients was studied at 18, 24 and 30 months and no significance was shown.

At 3 months, the pattern of missing PFQ data showed statistical significance for gender, with 41 males (25%) and 12 females (12%) not providing responses (p = 0.017). At 12 months, younger people (mean age 36.35 years, SD 14.59 years) were less likely to provide responses than the rest of the cohort (mean age 40.44 years, SD 15.85 years) (p = 0.045). This was also the case at 24 months, younger people (mean age 36.89 years, SD 14.10 years) versus the rest of the cohort (mean age 43.52 years, SD 14.56 years) (p = 0.012).

At each follow-up point, however, missing data were evenly distributed between study arms, suggesting that the imputation process adopted should not have led to any estimation bias.

Trial drug costs

A total of 42 of 52 of sites that recruited participants to the RCT responded to our survey regarding preparation and delivery of trial drugs. Pharmacists prepared infliximab infusions at 11 sites and prepared ciclosporin infusions at seven sites. All other infusions were prepared by nurses. For infliximab, the mean administration costs were £17.66 (SD £25.05) per infusion. For ciclosporin they were £35.41 (SD £28.48) per day during which the infusion was received.

Eighty-eight infliximab participants received the three infusions specified in the protocol; 98 received two infusions and 134 received a single infusion. Thirty-four participants received additional infusions (minimum = 1, maximum = 13) with a mean of 1.20 (SD 2.71) additional infusions.

One hundred and thirty-two ciclosporin participants received the single infusion specified in the protocol with a mean (SD) infusion duration of 5.0 (SD 1.8) days. Of these, 100 were switched to oral following their single infusion with a mean duration on oral ciclosporin of 52.0 (SD 48.8) days.

Table 32 shows the mean costs of treatment with the two trial drugs. As most participants on oral ciclosporin completed their treatment within 3 months, all oral ciclosporin costs have been attributed to the first 3 months. Infliximab is clearly the more costly treatment with a mean cost in the first 3 months of £4823 (SD £2199) compared with £880 (SD £649) for ciclosporin and with costs for additional infliximab infusions continuing beyond this period.

Baseline spell in hospital: length of stay

All participants were recruited and randomised while inpatients. Mean length of stay of this baseline spell in hospital (i.e. from randomisation to discharge) was slightly higher for ciclosporin participants but the difference was not statistically significant: ciclosporin 12.21 (SD 10.18) days, infliximab 10.32 (SD 13.55) days (95% CI –1.06 to 4.69 days; p = 0.20) as shown in Table 33.

Twenty-eight participants in the infliximab arm (20.7%) had surgery during the baseline spell compared with 34 in the ciclosporin arm (25.2%). The mean length of stay for those who had surgery was similar between groups: infliximab 22.89 (SD 14.90) days versus ciclosporin 21.38 (SD 14.65) days (95% CI –9.03 to 5.91 days; p = 0.68).

A total of 107 participants in the infliximab arm (79.2%) were discharged from their baseline spell in hospital without having surgery compared with 101 in the ciclosporin arm (74.8%). For those who did not have surgery, the mean length of stay was higher for ciclosporin participants at 9.13 (SD 5.45) days versus 7.02 (SD 11.07) days for infliximab, although the difference was not statistically significant (95% CI –9.03 to 5.91 days; p = 0.08).

Cost of hospitalisation without surgery

The cost of the baseline spell in hospital for participants who did not have surgery is shown in Table 34, which also shows costs for all subsequent non-surgical hospital admissions. The mean differences were not statistically significant in any period.

Cost of hospitalisation with surgery

Table 35 shows the number of participants who had colectomies or other surgical procedures by follow-up period. None of the differences was statistically significant (chi-squared test).

Table 36 shows the costs of surgical admissions at each follow-up point. There were no statistically significant differences between groups at any point.

NHS resource use

The costs of NHS resource use in the 3 months prior to baseline are shown in Table 37. There were no statistically significant differences across individual resource items apart from prescription drugs in which the mean cost for ciclosporin participants was significantly higher: £224 (SD £223) versus £166 (SD £185), mean difference £58 (95% CI £9 to £107; p = 0.020). Total costs were not significantly different with a mean cost for ciclosporin participants of £1115 (SD £1291) versus with £1034 (SD £1047), mean difference £80 (95% CI –£194 to £352; p = 0.573).

The total costs of all NHS contacts (as itemised above) are shown in Table 38. Table 38 also shows the costs of prescribed drugs other than the trial drugs given to participants as inpatients and by prescription in the community, and the costs of tests and investigations. There were no statistically significant differences between arms for any items during any period.

Incremental cost-effectiveness

Results relate to a ‘within-trial’ cost-effectiveness analysis using primary data collected within the period of study and without lifetime extrapolation. Base-case costs and QALYs to be reported below cover 30 months adjusted by the number of days over which each participant contributed data. Table 39 shows that the mean number of days that ciclosporin participants contributed CRF data was slightly higher than for infliximab participants: 645.9 (SD 204.8) days versus 623.7 (SD 224.0) days, but the mean difference of 22.2 days was not statistically significant (95% CI –32.2 to 77.4 days; p = 0.433). A similar pattern was shown for PFQ data: 673.0 (SD 226.0) days versus 653.2 (SD 224.8) days, mean difference 19.7 days (95% CI –36.8 to 76.5 days; p = 0.509).

Base-case costs and QALYs are shown in Table 40. The mean difference in QALYs was 0.023 in favour of ciclosporin, but this is not statistically significant (95% CI –0.053 to 0.101; p = 0.563). The mean difference in costs was –£5632 in favour of ciclosporin and is statistically significant (95% CI –8348 to –2880; p < 0.001).

The cost-effectiveness plane representing 5000 bootstrap replications is shown in Figure 12. Most observations are in the south-east quadrant suggesting that ciclosporin dominates infliximab.

FIGURE 12.

Base-case (30-month) cost-effectiveness plane adjusted for by participants’ length of time in study.

The CEAC shown in Figure 13 shows ciclosporin to have a 73–74% probability of being cost-effective over virtually all willingness-to-pay thresholds. The almost horizontal curve is due to the difference in effects being close to zero. 59,98

FIGURE 13.

Base-case (30-month) CEAC adjusted by participants’ length of time in study.

Table 41 shows base-case costs and QALYs over 30 months further adjusted using baseline EQ-5D and CUCQ scores and participants’ weight as covariates for QALYs and baseline cost as a covariate for costs. The mean difference in QALYs remains in favour of ciclosporin but is slightly reduced by 0.002 to 0.021 QALYs, which is not statistically significant (95% CI – 0.032 to 0.096; p = 0.35). The mean difference in costs (–£5632) is unchanged and is statistically significant (95% CI –£8305 to –2773; p < 0.001).

The cost-effectiveness plane is shown in Figure 14. The decrease in spread is due to reduced standard errors. Ciclosporin continues to dominate infliximab with the CEAC (Figure 15) showing ciclosporin to have an 85% chance of being cost-effective over all willingness-to-pay thresholds.

FIGURE 14.

Base-case (30-month) cost-effectiveness plane weighted for participants’ time in study and adjusted for baseline covariates.

FIGURE 15.

Base-case (30-month) CEAC weighted for participants’ time in study and adjusted for baseline covariates.

Sensitivity analyses

Table 42 below shows results at 12 and 24 months which are similar to the base case, in that the differences in QALYs at both time points are in favour of ciclosporin and are small (0.027 and 0.022 QALYs at 12 and 24 months, respectively). The mean difference at 12 months is marginally statistically significant (p = 0.049), but not at 24 months (p = 0.479). Mean differences in costs in both cases are again in favour of ciclosporin (–£4568 and –£4498 and are statistically significant; 12 months (p < 0.001), 24 months (p < 0.001). These two sensitivity analyses are consistent with the base case showing ciclosporin to dominate infliximab.

Participants’ time off work

Mean participants’ time off work in the 3 months prior to baseline was similar at 7.70 (SD 15.62) days and 8.27 (SD 14.32) days for infliximab and ciclosporin participants, respectively. There were very few differences between groups at other follow-up points suggesting that treatment had no differential impact on participants’ time off work as shown in Table 43.

Number of interviews completed

The qualitative section in Chapter 2 explained that we planned to conduct a total of 24 interviews at 3- and 12-month intervals with participants after randomisation and treatment. However, after 20 3-month interviews had been analysed, data saturation was reached and the TMG confirmed that no further interviews were needed.

The number of 3-month interviews completed was split evenly between those randomised to infliximab and those to ciclosporin, with 10 participants in each group, of which three in each group had also had a colectomy. There were three females and seven males in each treatment group and their ages ranged from 21 to 75 years, as shown in Table 44.

To assess the representativeness or otherwise of the 20 participants selected for interview, we compared the primary outcome, QAS per day, in terms of group and interview status. The numerical summaries of means and SDs in Table 45 are consistent with the box plots in Figure 16, and show that, in both groups, those interviewed are reasonably similar to those not interviewed.

FIGURE 16.

Graphical summaries of QAS per day.

It is worth noting the wide age range of the interviewees, but that their views and experiences were similar and the only differences seemed to relate to views about reversal procedures for those who had had surgery to treat their UC. This is explained below.

The length of time since the participants had been diagnosed with UC varied from just a few weeks to as long as 30 years. However, those who had only recently been diagnosed had generally been experiencing symptoms for several months.

With many participants working, the interviews often took place when people returned home from work. The interviews lasted between 15 minutes (unusually short) and 45 minutes in length (more common), and all participants agreed that the interviews could be recorded.

As soon as possible after an interview had taken place, recordings were transcribed and then reviewed by the interviewer. The interview transcripts were anonymised and any names or geographical details were removed from the transcripts.

At 12 months, 15 of the interviewees agreed to a second interview (eight infliximab, seven ciclosporin); one participant had died, two did not want to take part in the second interview and two could not be contacted. One participant in each treatment group, one male and one female, had undergone colectomy since their first interview and one male participant who had undergone a colectomy before the first interview had since had two further operations, pouch surgery followed by ileostomy closure. In total, 35 interviews were conducted with 20 participants.

Analysis of data

Findings

Conclusion

The findings from these interviews confirm the liking that patients have for infliximab, because of the treatment regime described, fewer side effects and ease of handling the drug. Ciclosporin was less favourably described, with more side effects, greater problems of drug handing and a more onerous treatment regime.

The physical symptoms that UC sufferers described, particularly when experiencing a flare-up, are dramatic and debilitating clearly indicated by these data in terms of the effects of symptoms on patients’ lives, leading to social isolation as their health deteriorates. Efforts should be made to ensure prompt diagnosis, to encourage known UC patients to report changes in their symptoms as soon as possible and to offer appropriate treatment that patients favour, which should be administered as quickly as possible.

A significant issue that emerged was the need for better support and greater information to enable patients to manage the impact of unpredictable symptoms and ongoing treatment regimes.

More research needs to be conducted to explore the views of UC patients post surgery, and to provide more information that could support the decision-making process of those facing surgery. As the views of participants following surgery were generally positive, there should be further research into the surgical treatment of UC as a real alternative to medical treatment.

An awareness raising campaign about UC (and Crohn’s disease) including details about surgery would encourage people to seek help earlier and help to destigmatise UC, thereby reducing the embarrassment felt by sufferers and those living with the outcomes of surgery.

Main issues arising

The main issues that emerge from the findings above are:

• Professional interviews indicate a clear preference for infliximab among nurses.

• Most doctors are more equivocal and prepared to wait for evidence of effectiveness and safety before making up their minds.

• Some doctors are strongly in favour of infliximab, wishing to see it as the drug of choice in view of its ability to deal with the many complex symptoms this disease group displays; its ease of administration; fewer adverse side effects; greater familiarity; convenience; greater perceived effectiveness; and ease of handling.

• Ciclosporin is more cumbersome to administer and requires additional monitoring, which puts pressure on an already overstretched workforce.

• Professionals view nurse colleagues as more familiar with the administration of infliximab and note the fewer demands it puts on nurses’ time.

• Ciclosporin is more restrictive on patients’ movements, leaving them frustrated and in need of more intensive support from nurses who must be present to manage complications.

• Professionals question current NICE guidelines and government regulations around drug use and the restrictions that this places on their professional autonomy.

• Professionals want to gain a clearer understanding of how the drugs affect patients’ lives.

• Professionals complement the trial for shining a light on this area of study which they see as seriously under-researched.

Drug administration and management

Consultants and nurses, when asked about the administration of infliximab and ciclosporin, were keen to point out that consideration must not only be given to the administration process but also to the lead-up procedure (‘work up to treatment’, consultant, HP7.45). Patient work-up was seen as a vitally important step in administering the drugs. However, work-up was said to be something that was neither given enough consideration within the drug guidelines, nor discussed in terms of positive and negative consequences and its effect on the rest of the administration process, such as its effect on drug handling. Interviewees described the work-up necessary for the provision of infliximab and ciclosporin in similar terms; however, following initial work-up administration, infliximab was much easier to handle than the administration activities necessary for the provision of ciclosporin. First, there was less to be concerned about during that period of time, and, second, the process had less impact on workload, especially for nurses, while both drugs demanded due care and attention. Attention to detail was described according to the careful mixing of drugs and other preparations necessary for treatment, prescreening of patients and patient preparedness for the intravenous infusions. In this respect ciclosporin was consistently described as the more complicated of the two, with a longer administration time (continuous as opposed to 2 hours for infliximab), and with the need for frequent changes of intravenous bags (infliximab is a one-off infusion). This had a knock-on effect on health-care professionals’ work schedules:

It [ciclosporin] goes on over a longer period of time obviously, so the need [for nurses] to continually make up bags and things over a longer period of time rather than just the one off infusion.

Consultant, HP13.25

It’s [ciclosporin] time consuming for the nurses and slightly messy. You know it’s complicated and it’s work for the nursing staff.

Consultant, HP15.19

Nurses’ views accorded with consultant opinion on this matter. In addition, nurses were keen to emphasise the complexities resulting from contextual and geographical factors that had to be contended within a busy hospital setting, such as managing patients on different wards requiring an infusion and, as a consequence, the greater number of nurses with advanced prescribing capabilities necessary on the wards at any one time:

Having to make it up [ciclosporin] every 6 hours is time consuming, changing lines, always having two nurses to check it because the way the GI [gastrointestinal] unit is split is there’s a corridor up between the two wards so obviously bed cover etc. but only to have one trained [nurse] each side is a bit difficult . . . geography of the wards . . . we’ve no one else to check the drugs.

Nurse, HPN19.21/28/29

You have to be mindful that the continuous infusion [ciclosporin] has to be prescribed to cover the weekend until Monday.

Nurse, HPN18.27

The practicalities of a lengthier ciclosporin infusion had implications for patients’ well-being too, with health professionals having to spend more extensive periods of time in hospital dealing with patients who were ‘frustrated’ (nurse, HPN21.98) with highly restricted movement:

[Ciclosporin] fairly cumbersome for both the staff and equally importantly for patients because once you are tied to the drip and associated drip stand, it sort of restricts patients moving around.

Consultant, HP12.10

Patients don’t particularly like being hooked up to it for such a long period . . . it’s [ciclosporin] restrictive.

Nurse, HPN25b.52/57

Views on the ease of administering individual drugs were clearly influenced by an individual’s familiarity with the drug in question. Thus, those with more experience of ciclosporin tended to be more positive about that drug, and there were suggestions that nurses should be trained in both drugs as they faced a steep ‘learning curve’ (consultant, HP13.123) particularly in relation to ciclosporin. Nevertheless, support for the administration of infliximab was stronger and it was repeatedly described as the ‘easier’ (consultant HP5.18 and nurse HPN18.169) and more ‘convenient’ (consultant, HP12.78) drug:

Infliximab has an advantage (over ciclosporin) in that it’s just a 2 hour infusion and then it’s done . . . there’s no problem with infliximab, I think it’s a good drug to administer, I think it’s an easy drug to administer . . . I like infliximab because once you’ve done it, you’ve done it for two weeks.

Consultant, HP14.26/30/33

Although both drugs were perceived to be effective, ciclosporin was faster acting and the slower response time with infliximab presented a challenge in terms of whether to continue with infliximab or look towards a different treatment. Ciclosporin was more ‘clear-cut’ (consultant, HP11.81) and consultants felt more confident in their decision-making and timings with regard to this drug. In addition, consultants worried about the ambiguous information available for ciclosporin response rates, as these did not appear to match patients’ actual responsiveness, made all the more complex by the need for an extended administration period:

Tend to use ciclosporin as acting more quickly.

Consultant, HP4.171

This is the one thing I don’t like about infliximab, because there is no data to give us a clear timescale or timeline for decision-making.

Consultant, HP11.94

Sometimes it’s a bit challenging that you have given them infliximab if they are going to take sort of a week to 10 days time would you hold your nerve?

Consultant, HP12.47

In addition to response times, adverse effects were of major importance in treating patients with one or other drug. Ciclosporin in particular was seen to have a range of associated adverse effects, which included drug toxicity, renal failure, neurological impairment, seizures, tremors and hypertension. Consultants were ‘uneasy’ (consultant, HP9.56) about the drug, seeing it as ‘dangerous’ (consultant, HP6.72). Even those advocating its use tended to be apprehensive about the adverse effects in the longer term, thus viewing it as more of a bridging therapy. These views were reinforced by a perceived lack of evidence of ciclosporin’s longer-term benefits.

Don’t want to use it [ciclosporin] in the longer term because of side effects, so switch to azathioprine as soon as possible.

Consultant, HP6.99

I’m always a little bit more nervous with it [ciclosporin] . . . and I think that’s from the side effects of renal impairment . . . I think the side effects profile of ciclosporin, although maybe it has not come out in studies, still concerns us more.

Consultant, HP16.278/281

. . . potentially quite significant side effects that can happen . . . slightly uncomfortable feeling about it [ciclosporin].

Consultant, HP9.60

It’s [ciclosporin] a bit debatable although I use it, I worry about toxicity and the liver, use it half-heartedly, when no choice.

Consultant, HP3.29/32/33

An adverse risk profile for ciclosporin, we were told, results in increased patient monitoring and checking of drug and blood levels, which, in addition to a more resource-intensive process, impacts on nurses’ time. This is particularly noticeable on a busy ward where there are difficulties with nurse-to-patient ratios, and nurses had the sense that their profession was struggling with workload increases, impacting on their ability to give equal time to all patients under their care. They reported occasionally ‘forgetting’ (nurse, HPN24.255) to monitor ciclosporin patients fully. Infliximab, on the other hand, did not warrant any additional monitoring and no issues associated with monitoring were raised:

It’s [ciclosporin] time consuming with regards to observation, particularly on a busy ward when you’ve got one nurse to 10 patients, it can take quite a huge part of your workload.

Nurse, HPN21.16

We have to monitor this patient closely for any side effects . . . it [ciclosporin] takes 3 hours because you’ve got to do obs [observations] for a couple of hours, but then sometimes we tend to forget because we get busy with other patients . . . so ideally it should be one to one.

Nurse, HPN24.249/255

Although infliximab also carried a risk profile that included adverse reactions, causing particular hesitancy in prescribing it for the older patient, it was favoured for its fewer side effects and longer-term efficacy:

Most of the time it [infliximab] goes without incident.

Nurse, HPN19.83

Have only ever seen minor reactions to the infliximab.

Nurse, HPN21.82

However, consultants pointed out that they felt frustrated about policy restrictions and government guidelines that prevented them from using infliximab in the way in which they wished to, including the need to justify its use in the longer term:

We now have to request the funding for patients with UC on long-term infliximab so that’s a bit of a challenge and it means that decisions are made differently to whether a patient is going into maintenance infliximab.

Nurse, HPN19.154/156

It’s already an issue because we can’t really treat as maintenance with infliximab so obviously if a patient responds well we’ve had to get exceptional funding and things like that so there is a case for it I think.

Nurse, HPN22.103

Personal preference and trial involvement

Consultants clearly recognised the effect that their own experience of using one or other drug had on their personal preference. Some consultants, as indicated in Drug administration and management, have a preference for infliximab, based on its ease of preparation; easier administration; lower negative impact on nurse care provision, staffing and workloads; longer-term benefits; and reduced adverse effects. In addition, both consultants and nurses pointed to greater patient tolerance and enhanced patient convenience. Further reasons for stated preferences included familiarity with the drug, drug effectiveness, greater clinical expertise and knowledge and general sense of ‘unease’ (consultant, HP9.58) with the use of ciclosporin:

My personal view is that they probably have equal efficacy, I think infliximab has less side effects so if I was given a free choice, if I was asking it for me or for my loved ones, I would opt for infliximab.

Consultant, HP5.65

When a patient was given infliximab I was rather pleased and when they were given ciclosporin I was less enthusiastic . . . we wondered about the tolerance for the patient and convenience for the patient.

Consultant, HP15.380/394

Nurses’ views mimicked those of their consultant counterparts, expressing a preference for infliximab based on its ease of administration, perceived patient benefits and personal experience and familiarity with the drug. They highlighted the restrictions of patients being hooked up to a ciclosporin infusion for longer periods of time while suffering from bowel disturbances, and noted patients’ preference for infliximab in this respect. Ciclosporin was described as ‘high maintenance’ (nurse, HPN25a, 174):

I prefer it when they have infliximab . . . I think it’s easier for the patients because it’s just one infusion . . . it’s a couple of hours, instead of being hooked up. I mean keep in mind they have profuse diarrhoea . . . I think it’s easier to administer as a nurse and it’s easier for the patient to receive.

Nurse, HPN18.165/169/173/178/189

Personally I feel that the infliximab is better but that’s only because of the experiences we’ve had with the infliximab rather than the ciclosporin . . . seen patients do very well on it, particularly when before perhaps there wouldn’t have been any other outcome than surgery.

Nurse, HPN21.209/217

This personal preference was not, however, across the whole interview cohort, with two consultants describing their preference for ciclosporin, based on personal familiarity with the drug, its quicker-acting properties and its ability to disappear from the system once the drug administration had been halted. However, even these two interviewees stressed the need to retain a level of vigilance and to ‘be more wary’:

Rather more relaxed about it [ciclosporin], paradoxically, simply because I know when the IV [intravenous] infusion is stopped then the drug disappears.

Consultant, HP10.75

I’ve used it [ciclosporin] for a very long time and I’m comfortable with it.

Consultant, HP11.160

In spite of these personal preferences, there remained a genuine sense of uncertainty as to which drug was the more beneficial at the time of the trial, particularly in relation to people’s views of trial entry, trial management and patient recruitment. Most health-care professionals commented that effectiveness was equivocal across the two drugs, genuinely wanting answers from the trial team as to the ‘better’ of the two. This desire was described as patient-driven, based on which of the two was more effective and well received by patients.

If the result was that ciclosporin was a lot better we’d use it and if the results were that infliximab was a lot better we’d use it.

Consultant, HP16.622

The difference (between drugs) is not sufficiently dramatic that I would say it was unethical to put them into a randomised study.

Consultant, HP15.460

Professionals welcomed the CONSTRUCT trial and congratulated the trial team for moving this body of work forward so successfully, in spite of what was described as a somewhat ‘ambiguous’ area of research. They discussed the fact that they had attempted to actively recruit patients into the trial, despite, in the lower recruiting sites, a lack of success, which was put down to busy workloads and lack of joined-up working patterns.

Surgery

Health professionals had mixed views about using alternative therapies, such as surgery, for acute UC. Colectomy, for example, was generally seen as a ‘last resort’ (nurse, HPN21.247) when all other therapies had failed, with some health-care professionals describing feeling ‘disappointed’ (consultant, HP9.210) with colectomy being needed. Consultants argued that although surgery was an option, other medical avenues should be exhausted first before reaching that point:

It’s [colectomy] always a last resort . . . It’s not something I’m comfortable sending them off to have surgery.

Nurse, HPN21.247

Having a colectomy was also seen as a difficult decision to get patients to agree to, and discussions around surgery had to be highly individualised and dealt with on a patient-by-patient basis. Most professionals described colectomy as an option that needed discussion first, to provide both counselling for patients and an assessment of patients’ views: ‘once it comes out you cannot put it back’ (consultant, HP13.338). Patients were seen to be often ‘nervous’ (consultant, HP13.330) of surgery and wanted to put it off for as long as possible, and for this reason alone it was important to have medical options available to prolong the time before surgery. However, at times surgery was seen as being put off for ‘too long’ (consultant, HP20.192), resulting in more difficult decision-making and complications for patients:

I think it comes down to a personal decision, I think it’s very important to have lots of discussions about the options, medical or surgical . . . it’s very difficult if they’re newly diagnosed and they come in with their first presentation . . . some people can come with obvious pre-conceived ideas about things anyway and I think it’s important to explore those feelings.

Nurse, HPN19.254/267

Colectomy was recognised as a ‘lifesaving’ procedure for some patients to which patients responded positively once surgery was over and symptoms were alleviated:

My personal thoughts are that you know that’s their last resort and I think the majority of patients feel the same that they would try any form of medical management.

Nurse, HPN25b.137

Policy development and drug regulation

Health-care professionals were clearly aware of the cost of infliximab and saw the implications of the high cost to be driving policy development and guidelines for practice. Guidelines around the use of the drug ultimately affected clinical decision-making:

There is an issue of cost of course, we’re pushed all the time to stop it for cost reasons.

Consultant, HP15.120

Although health-care professionals understood the need for rigorous guidelines around the use of the drugs, many also linked guidelines to an enforced restriction of the use of infliximab, emphatically arguing that the NICE guidelines were outdated, outmoded and out of step with more recent evidence. This left them feeling frustrated and constrained in their ability to provide the best and most appropriate care for their patients. They described the situation as one in which they had ‘their hands tied’ (consultant, HP6, 127.131), indicating that the more recent clinical evidence and their own personal experience showed infliximab as the drug that needed to be used more widely. As a consequence, they urged that guidelines became more flexible in order to accommodate patients’ needs more appropriately:

It’s very hard to find anyone who supports the NICE guidance in its present format.

Consultant, HP16.516

I sort of treat it with a bit of contempt.

Consultant, HP9.239

The point was repeatedly reinforced that, although the guidelines were there to be adhered to, there was a clear sense of rebellion within the profession, with clinicians ‘ignoring’ (consultant, HP9.236) guidelines when they could find ways around them in order to best meet patient needs. Thus, health-care professionals welcomed a much-needed change to the NICE guidelines that allowed for the greater flexibility of drug use, led primarily from an assessment of patient need:

You can usually find a reason why ciclosporin would be contraindicated . . . I don’t think it’s a particularly sensible NICE guidance based on the lack of evidence that they have to make their decision and therefore in all honesty because we can, we slightly ignore it.

Consultant, HP9.232/236

I think it’s unduly restrictive and I think it will eventually change. I think it’s actually quite hard to adhere to as well. I mean you’ve got a patient who you want to give treatment to, you’re duty bound to give the best treatment and in some cases that’s going to be infliximab.

Consultant, HP14.208

Patient benefit and negotiated care

Throughout the discussions around the administration of the drugs, treatment of patients, personal preference, the value of surgery, and adherence to guidelines and regulations, was a professional focus on patient benefit and need. Nurses and consultants gave many examples of the decisions that were being made that took into account what was best for the patient, especially when it came to a consideration of a patient’s personal circumstances:

I like to tailor the treatment to the patient.

Consultant, HP10.144

The drug we choose ultimately depends on the patient.

Consultant, HP5.75

Nurses were more influenced by the patient experience than the consultants were, perhaps as a result of their extended contact with the patients, and nurses discussed their preference in the context of patient convenience and QoL:

It worked really well on that patient . . . there’s a quicker response time with infliximab . . . patients pass comments and they always tell me I feel much better after this one.

Nurse, HPN24.33/118

It’s whatever is best for the patient . . . I think it’s different for each patient and I think that they need to be given the choice and have all the information because for some patients it can be a very good treatment [colectomy] and give them back their quality of life.

Nurse, HPN22.139/157

As evident in the last quote, the notion of negotiated care was central to the consultation as far as consultants were concerned, and ongoing patient care and decision-making was perceived as being shared across and within multidisciplinary teams. This led to a shared professional responsibility, with the patient–clinician interaction the ultimate example of this:

Our surgeons are very active in decision-making, and multidisciplinary team meetings twice a week.

Consultant, HP5.121

Care is negotiated with the patient, patient input is important, but predefined pathways are according to previous discussions.

Consultant, HP4.113

Patient views are as important as anything else, combined with weighing up risks and benefits, we involve them hugely.

Consultant, HP7.103

The results of this interview study indicate that doctors would make a judgement between the two drugs largely on effectiveness but that there is a clear preference among nurses for infliximab. Ciclosporin is more complex and cumbersome to administer, requiring additional monitoring which tends to be particularly problematic on busy wards with extensive demands on an already overstretched health-care professional workforce. The longer administrative process restricts patient’s movements and leaves them feeling frustrated. It is also excessively demanding on the time of specialist nurses and other health-care professionals, who need to be present to manage any complications, and there is more intensive patient monitoring due to the drug’s adverse risk profile. Consequently, professionals question current NICE guidelines and government regulations around drug use and suggest a sense of rebellion about the restrictions placed on their autonomy and their practice.

Not only are professionals keen to change regulations surrounding current prescribing and drug administration, they also want changes to longer-term patient maintenance and monitoring, preferring infliximab as the longer-term drug of choice in view of its success in dealing with the many complex symptoms that the disease group displays, as well as it being the more familiar, convenient, effective and easy to handle of the two.