Ondansetron for irritable bowel syndrome with diarrhoea: randomised controlled trial

Transit

5-HT3 receptor antagonists slow transit,10,11 an effect we found particularly marked in the left colon of IBS-D patients and rectosigmoid region, but the underlying mechanism was unclear. 12

Tone and motility

An early colonic barostat study showed a reduction in the postprandial rise in colonic tone with ondansetron in both carcinoid syndrome patients16 and healthy volunteers,17 which would be predicted to slow colonic transit. Previous studies of the impact of 5HT₃RAs on human colonic motility18–20 were however paradoxical in showing that the 5HT₃RAs tropisetron, alosetron and cilansetron all increased periprandial frequency of colonic contractions, and mean amplitude of contractions in the left colon. We hypothesised that 5HT₃RAs increase retrograde sigmoid motility, perhaps enhancing ‘brake’ function,21,22 which would be a novel mode of action.

Proteases

In our pilot study we showed the decrease in urgency associated with ondansetron use correlated directly with the reduction in faecal protease (FP),23 but whether this represents a true causal relationship, or just an epiphenomenon, is unclear. FPs have been shown to be increased in IBS-D and, at least in animal models, cause hypersensitivity to rectal distension via activation of protease-activated receptors type 2 (PAR2). 24 We have shown that most FPs are endogenous,23 representing pancreatic enzymes that have escaped degradation by colonic bacteria. We hypothesise that slowing gut transit with ondansetron reduces FP, by allowing time for bacterial degradation, and that this may contribute to the beneficial effects of ondansetron on urgency. Reducing faecal tryptase might improve the anal soreness that is commonly reported by patients with IBS-D.

Bile acids

Bile acids have also been shown to sensitise the rectum,25 and elevated faecal bile acids (FBAs) have been shown by several groups in patients with IBS-D. 26 Slowing transit may increase the time for bile acid deconjugation by colonic bacteria, and therefore enhance absorption, but how important this is in reducing rectal sensitivity, compared with the effects on FPs, is unclear.

Genetic factors altering sensitivity to 5HT₃RAs

We have shown that individuals vary widely in their responsiveness to ondansetron, explaining why trials using fixed doses of 5HT₃RAs result in severe constipation in around one in four patients. Meta-analysis gives a relative risk of constipation of 4.3 (3.3–5.6). 7 However, when patients were allowed to dose titrate we found that constipation was rare, occurring in only 9% of patients, most of whom responded to dose reduction and only 2% discontinuing ondansetron because of this. 12 However, the required dose of ondansetron ranged from 4 mg on alternate days to 8 mg three times a day (t.d.s. ). The reasons for this variation are unclear, but recent evidence suggests that responsiveness to 5HT₃RAs might be linked to polymorphisms in the genes controlling 5HT synthesis. Serotonin availability in the rectal mucosa is thought to be determined by the activity of the rate-limiting synthetic enzyme tryptophan hydroxylase-1 (TPH-1), which produces serotonin in enterochromaffin cells. A recent small study showed TPH-1 messenger ribonucleic acid (mRNA) levels in rectal mucosa (and thus presumably serotonin synthesis rate) were approximately doubled in responders to another 5HT₃RA, ramosetron, compared with non-responders, and that this was linked to the TPH-1 genotype. 27 TPH-1 rs211105 minor allele G was found in 44% of non-responders, but only 4% of responders, indicating that possessing the major allele increases responsiveness to the drug. It was also associated with worse diarrhoea, possibly because of the greater 5HT synthesis.

Inclusion criteria

These were as follows:

• patients had to meet modified Rome IV criteria for IBS-D (see as previously published);

• patients had to be aged ≥ 18 years;

• patients should have completed standardised workup to exclude:

  • microscopic colitis (colonoscopy or flexible sigmoidoscopy with colonic biopsies);

  • bile acid diarrhoea (BDA) [Selenium homocholic acid taurine (SeHCAT) results of > 10%, C4 results of < 19 ng/ml or failed 1-week trial of a bile acid binding agent (colestyramine 4 g t.d.s. , colesevelam 625 mg t.d.s. or equivalent)];

  • lactose malabsorption (suggested but not mandated negative lactose breath hydrogen test, negative clinical challenge or failure to respond to lactose-free diet);

  • coeliac disease [tissue transglutaminase (tTG) or duodenal biopsy];

• patients of childbearing potential or with partners of childbearing potential had to agree to use methods of medically acceptable forms of contraception during the trial and for 90 days after completion of trial medication;

• for women of childbearing potential, a negative pregnancy test was performed within 72 hours of confirmation of eligibility;

• weekly average worst pain score had to be ≥ 30 on a 0 to 100-point scale;

• stool diary needed to show stools with a consistency of 6 or 7 on the Bristol Stool Form Scale (BSFS) for 2 or more days per week.

Modified Rome IV diagnostic criteria for irritable bowel syndrome with diarrhoea

Must fulfil the following criteria for the past 3 months:

1. recurrent abdominal pain at least weekly;

2. pain is associated with two or more of the following criteria:

   1. related to defaecation;

   2. associated with a change in frequency of stool;

   3. associated with a change in form (appearance) of stool;

3. symptom onset at least 6 months prior to diagnosis;

4. > 25% of abnormal stools are loose (BSFS 6 or 7) but < 25% are hard (BSFS 1 or 2).

Reproduced under creativecommons.org/licenses/by/4.0/.

Exclusion criteria

• Gastrectomy.

• Intestinal resection.

• Other known organic gastrointestinal diseases [e.g. inflammatory bowel disease (Crohn’s disease, ulcerative colitis)].

• Unable or unwilling to stop restricted medication, including regular loperamide, antispasmodics (e.g. hyoscine butylbromide, mebeverine, peppermint oil, alverine citrate), eluxadoline, tricyclic antidepressant (TCA) doses > 30 mg/day or other drugs likely in the opinion of the investigator to alter bowel habit. These medicines should be discontinued for a 7-day washout period prior to registration.

• Corrected QT (QTc) interval ≥ 450 milliseconds for men or ≥ 470 milliseconds for women [assessed within the last 3 months by electrocardiogram (ECG)].

• Previous chronic use of ondansetron or contraindications to it.

• Pulse, blood pressure (BP) and laboratory blood values outside the normal ranges according to the site’s local definition of normal [assessed within the last 3 months]. Note minor rises in alanine transaminase (ALT) (< 2 × upper limit of normal) will be acceptable, but the patient’s general practitioner (GP) will be informed if they remain elevated at the end of the trial.

• Women who were pregnant or breastfeeding.

• Patients currently participating in, or who have been in, a trial of an investigational medicinal product (IMP) in the previous 3 months, where the use of the IMP may cause issues with the assessment of causality in this trial.

• Patients who had started or altered dosing of selective serotonin reuptake inhibitors or TCAs in the last 3 months, or who were planning to change the dose during the trial.

• Patients currently taking and unwilling or unable to stop apomorphine or tramadol (which interact with ondansetron).

• Patients with only stools of consistency 7 on the BSFS for 7 days a week.

Patients taking QT-prolonging or cardiotoxic drugs were reviewed by the local principal investigator (PI) to determine whether they were suitable for the trial.

Visit 1

Potential trial candidates attended their first visit for registration and consent by the PI or delegate. If required, further tests to exclude diagnoses other than IBS-D were arranged. These included a SeHCAT scan, serum C4 level (Edinburgh centre only) or a 1-week trial of a bile acid binding agent to assess for bile acid malabsorption (unless done within the last 5 years), and a colonoscopy (unless done within 2 years, or 5 years if they also had a current normal calprotectin) to assess for microscopic colitis. Baseline serum blood tests, vital signs, demographics (date of birth, gender, ethnicity and smoking history) and an ECG were obtained. Current medications were reviewed, and those unable to discontinue drugs likely to alter bowel habit were unable to enter the trial. Patients on QT-prolonging drugs and cardiotoxic drugs were reviewed by the PI for suitability for the trial, as high-dose ondansetron may increase the risk of QT prolongation and arrhythmias. Eligible and consenting patients were registered and allocated a unique trial ID.

All patients were asked to complete a 2-week daily diary recording stool frequency, stool consistency for each stool (using the BSFS), worst abdominal pain (on a scale of 0–100), worst bowel movement urgency (on a scale 0–100) and if they had used loperamide that day. In addition, patients had the option to be sent two automated text messages each day, something most participants initially agreed to do. The first message asked the patient if they had passed a stool of consistency 6 or 7 on the BSFS. They had to reply with either a yes or no. The second text message asked what their worst abdominal pain score was that day. The patient could respond with a number from a scale of 0–100 (where 0 is no pain and 100 is the worst imaginable pain).

Visit 2

Two weeks later the patient returned to confirm eligibility. The diary was used to confirm they had stool consistency BSFS 6–7 for more than 2 days a week and did not meet the exclusion criteria of having only BSFS 7. They also had to have a weekly average worst pain score ≥ 30. For patient consenting to the whole gut transit study, Transit-Pellet capsules containing markers were dispensed and the abdominal X-ray appointment confirmed for the morning of visit 3. If the patient had consented to one or both mechanistic studies, appointments for baseline assessment were arranged prior to visit 3.

Visit 3

On visit 3, patients underwent a pregnancy test if applicable, whole gut transit assessment by abdominal X-ray (if consented), rigid sigmoidoscopy (if consented), completion of baseline questionnaire booklet [including IBS-SSS, Short-form Leeds dyspepsia questionnaire (SFLDQ), Hospital Anxiety and Depression Scale (HADS), Patient Health Questionnaire 12 (PHQ-12) and IBS-QOL questionnaires], and collection of stool, whole blood and serum samples (if consented). Patients were then randomised and given a 6-week patient diary and the trial medication in accordance with their blinded randomisation allocation. Patients were asked to record the following on a daily basis: stool frequency, consistency of each stool, worst abdominal pain experienced that day, worst bowel movement urgency, number of trial medication capsules taken and whether they had used loperamide that day. Every week the diary required them to respond to the question whether they felt that ‘they have had satisfactory relief from their IBS symptoms that week’. If they agreed, the patient continued to receive two text messages each day for the rest of the trial, asking if they had passed a stool of a consistency of six or seven that day, and what their worst abdominal pain was that day.

During the first 2 weeks, patients were contacted every 2 days by the local site team to discuss bowel habit. The dose of ondansetron or placebo was titrated as required. Additional guidance on dose titration was given to each trial site in a standard operating procedure, and to the patient in a dose titration instruction leaflet (see Report Supplementary Materials 1 and 2). A check for serious adverse events (SAEs) was performed during each telephone call. The steady dose to be taken forward for the remainder of the trial was mostly confirmed in week 2, although patients could alter this during the 12 weeks, if required, to avoid constipation.

Visit 4

At 6 weeks on the trial treatment, patients returned for their fourth visit. Diaries were collected, and the investigator asked whether any reportable adverse events (AEs) had occurred since the last visit. A pregnancy test was taken (if required) and concurrent medication reviewed to ensure these do not interfere with the trial medication. A further 6-week patient diary, trial medication and Transit-Pellet capsules (for use 6 days prior to visit 5) were dispensed. If the patient consented to mechanistic studies, appointments were confirmed and arranged as convenient between 8 and 11 weeks of treatment. Daily text messages continued for a further 6 weeks for those patients who agreed.

Visit 5

After 12 weeks on the trial medication, patients returned for visit 5. A pregnancy test was taken (if required), concurrent medication reviewed to ensure that these did not interfere with the trial medication and the investigator asked whether any reportable AEs had occurred since the last visit. Unused medication and completed patient diaries were collected. The plain abdominal X-ray to assess whole gut transit was performed for consenting patients. Serum and stool samples were collected for consenting patients, and all patients completed the 12-week questionnaire booklet, including the IBS-SSS, SFLDQ, HADS and IBS-QOL questionnaires. Patients were issued with a follow-up diary and continued to respond to text messages for a further 4 weeks.

Visit 6

Patients returned for the sixth and final visit, where the diary was collected, and the investigator asked about any reportable AEs since the last visit.

See Appendix 1, Table 34 for a breakdown of activities per visit.

Primary outcome measure

Food and Drug Administration (FDA)-defined response in relation to abnormal defaecation and abdominal pain measured over 12 weeks post randomisation by patient diary and daily text message.

Definition: A FDA responder is a patient who records both a reduction in pain intensity and an improvement in stool consistency for at least 6 weeks of the 12-week treatment period, where:

• reduction in pain intensity is defined as ≥ 30% decrease from baseline in weekly average worst daily pain;

• improvement in stool consistency is defined as ≥ 50% decrease in the number of days per week with at least one loose stool [BSFS (21) 6 or 7)].

The two components to response (pain intensity and stool consistency) are also reported as individual outcomes.

Secondary outcome measures

Irritable bowel syndrome with diarrhoea symptoms measured by patient diary and daily text message over 12 weeks post randomisation and calculated for weeks 1–12 and for weeks 9–12:

1. abdominal pain (mean daily pain score, scored 0 = no pain to 100 = worst possible pain);

2. urgency of defaecation (mean daily urgency score, scored 0 = no urgency to 100 = worst imaginable urgency);

3. stool consistency (mean number of days per week with at least one loose stool; mean daily BSFS);

4. stool frequency (mean number of daily stools);

5. use of rescue medication (loperamide) over 12 weeks of treatment;

6. satisfactory relief assessed through answers to ‘Overall, have you had satisfactory relief from your IBS symptoms in the past week?’ Yes or No (measured by diary at the end of each week).

Self-reported symptoms, QoL and mood at 12 weeks post randomisation:

1. IBS symptom severity [measured by the IBS Severity Scoring System (IBS-SSS)];28

2. dyspepsia (using the SFLDQ);29

3. QoL and mood (IBS-QOL);30

4. HADS questionnaires;31

5. somatic symptoms [Patient Health Questionnaire 12 Somatic Symptoms (PHQ-12SS)] questionnaire. 32

Post-treatment (weeks 13–16 post randomisation) recorded in patient diary:

1. abdominal pain;

2. urgency;

3. stool consistency;

4. stool frequency.

Safety:

1. SAEs assessed throughout 12-week treatment period;

2. AEs assessed 4 weeks after the end of treatment to determine if there are any persisting effects.

Summary of any changes to the project protocol

A number of minor amendments were made; a summary of protocol changes can be found in Appendix 1, Table 35. There were no changes to end points or drug administration.

These minor amendments included:

1. raising the permitted calprotectin level to be < 100 ng/mg in view of new national recommendations;

2. allowing a 7-day trial of colestyramine to replace SeHCAT testing to exclude BDA to overcome the problem of long delays in SeHCAT testing;

3. simplifying the information sheets and including the new data on congenital malformations which became available during the trial;

4. adding GP surgeries within the East Midlands as Participant Identification Centres (PICs), October 2018 and January 2019;

5. allowing patients who have had a colonoscopy within 10 years to exclude microscopic colitis to be enrolled in the trial. The exclusion criteria have been amended to clarify that patients would need to stop taking high-dose TCAs to 19 July 2019;

6. allowing new adverts, July 2019;

7. both the treatment patient information sheet (PIS) and the treatment + tests PIS have been updated to reflect the new safety information regarding the use of ondansetron in pregnancy, October 2019;

8. the SMS Guidance for patients has been updated. A fully tracked document has been included for review. Changes have been made to this document to make it easier to understand and to add clarity on what is involved in this aspect of the study, October 2019;

9. an SMS messages flowchart has been added. Due to an oversight the content of the SMS messages has not been previously submitted for approval, October 2019.

Transit

Methods

Patients were asked to take the transit capsules daily for Transit-Pellets method™ supplied by Medifactia AB Limited, Stockholm, Sweden. Pellets were taken starting 6 days before their planned visit 3. Patients were instructed to take one pill at approximately 9 a.m. each day apart from day 6 when there were two pills to take, one at 9 a.m. and the second pill at around 6 p.m. This adaptation of the standard Metcalf method is designed to deal with patients with very fast transit when pellets taken at 9 a.m. may well have left the colon leaving zero pills to count. This then would raise the question of whether any pills had been taken whereas taking the second later pill makes this much less likely. On visit 3 a plain abdominal X-ray was taken and used to count the number of pellets in each of three regions designated right colon, left colon and pelvic area as defined on the plain abdominal X-ray by a vertical line to tip of the fifth lumber vertebra spinal process and then tangential to either pelvic brim as shown below. Colonic transit calculations from plain X-ray (Figure 1) were as described by Chan et al. 1 and used by Garsed et al. 2

FIGURE 1.

Colonic transit calculations from plain X-ray. Whole gut transit (WGTT) was calculated from total pellets × 2.4 hours, so in this example of a patient on ondansetron with 21 pellets, WGTT = 50.4 hours.

Regional transit was calculated from the number of pellets in the right, left and pelvic areas × 2.4 so in the example right transit time = 2 × 2.4 = 4.8, left = 2.4 × 3 = 7.2 and pelvic = 16 × 2.4 = 38.4 hours.

High-resolution manometry

Barostat assessment

This was performed at baseline in nine subjects and after 8–11 weeks of treatment in Nottingham and the Royal London Hospital; six of whom also underwent manometry. Patients from other sites were also offered the opportunity to travel to either site but only two patients accepted this offer.

Stool biomarkers: faecal proteases and bile acids

Primary end point

Treatment of irritable bowel syndrome using titrated ondansetron trial (TRITON) planned to recruit 400 patients from up to 24 sites across England and Scotland. This would have provided 90% power at 5% significance to detect a 15% absolute difference between the randomised groups in the proportion of patients achieving the FDA-recommended end point39 of a weekly responder for pain intensity and stool consistency for at least 6 weeks of the 12-week treatment period. This difference (15%) was considered to represent the minimum clinically important difference since in practice over the last two decades new IBS drugs with a lesser margin of efficacy are rarely prescribed in the NHS. We assumed a placebo response rate of 17%, as recently reported12 using this end point and allowed for a 15% attrition rate.

Mechanistic studies

Rectal compliance and sensitivity

Previous studies with alosetron showed an increase in compliance from 5.9 (SD 1.3) to 9.8 (SD 1.2) ml/mmHg in 22 patients. 40 We aimed to study 40 patients on each treatment to calculate correlations with symptoms, which typically require much larger numbers than just showing a change in mean values.

Frequency of analyses

Outcome data were analysed once only, at final analysis, although statistical monitoring of safety data was conducted throughout the trial and reported at agreed intervals to the DMEC. Final analysis took place 16 weeks post–last patient randomisation.

End-point analysis

All analyses were conducted on the ITT population, defined as all patients randomised, regardless of non-compliance with the intervention. A per-protocol (PP) analysis of the primary end point was carried out to indicate whether results were sensitive to the exclusion of patients who violated the protocol (e.g. those patients randomised but subsequently found to be ineligible). Primary and secondary analyses were blind to random allocation. Outcome measures were analysed by regression models appropriate to the data type. Such analyses adjusted for randomisation minimisation factors: site, completion of manometry assessment and barostat assessment, as well as baseline values where applicable, age and gender. Baseline characteristics were summarised by randomised group.

Primary analysis

The primary analysis compared the proportions of patients achieving the FDA-recommended end point between treatment groups at 12 weeks post randomisation using a logistic regression model adjusted for minimisation factors, age and gender. The plan was that any missing data were assumed missing at random (MAR) and imputed for the primary analysis. However, complete case analysis was undertaken and those without sufficient data (n = 4) to evaluate the primary end-point were assumed to be non-responders. The potential impact of missing data would be small with only 5% of missing responses. Odds ratios and corresponding 95% CIs are presented.

Sensitivity analysis was planned to assess the impact of missing data on the treatment effect was performed. This was planned to include complete case analysis and alternatives to multiple imputation (e.g. pattern mixture modelling), if missing patterns suggested data were missing not at random. However, given the small overall sample size and only four participants with missing data (5%), this was not undertaken.

Secondary analyses

The proportions of patients with satisfactory relief of IBS symptoms at 12 weeks post randomisation was compared between the treatment groups using logistic regression models, adjusting for minimisation, baseline values, age and gender. Odds ratios and corresponding 95% CIs are presented.

Differences between the treatment groups for the continuous secondary end points at 12 weeks post randomisation were compared using linear regression models, adjusted for the minimisation variables, baseline where applicable, age and gender. These end points were urgency of defaecation over the last month, stool frequency over the last month, number of days per week with at least one loose stool (BSFS > 5) over the last month, average stool consistency, number of days rescue medication used over 12 weeks, abdominal pain score, HADS depression and anxiety scores, SFLDQ score, IBS-QOL score and subscales, PHQ-12 and IBS-SSS severity scores. Treatment estimates and corresponding 95% CIs are reported.

The differences between the treatment groups post treatment over weeks 13–16 post randomisation in the following end points: stool frequency, abdominal pain and urgency of defaecation were compared using linear regression model adjusting for minimisation factors, baseline values and relevant baseline factors. Treatment estimates and corresponding 95% CIs are reported. Any missing data are assumed MAR.

Exploratory analyses on the daily measurements (worst abdominal pain, loose stools, number of stools passed, consistency of stool, worst urgency and use of loperamide) were carried out, using repeated measures models, which incorporate correlation between measurements from the same patient.

SAS software version 9.4 was used in the analyses of primary and secondary end points.

Safety analyses

All patients who receive at least one dose of trial treatment were included in the safety analysis set. The number of patients reporting a SAE (up to 28 days after the last dose of treatment), and details of all SAEs, are reported for each treatment group. The number of patients withdrawing from trial treatment is summarised by treatment arm, along with reasons for withdrawal. All safety analyses performed prior to final analysis were undertaken by the safety statistician (rather than the trial statistician), thus ensuring that the trial team remain blinded.

Subgroup analyses

No subgroup analyses were planned.

Mechanistic studies

Mechanistic studies were analysed by the site research fellow, blinded to the intervention allocation under supervision of the chief investigator and local supervising PIs. The differences between treatment groups for changes in whole gut transit times, colonic motility measures (% time of cyclical retrograde contractions and HAPC frequency), rectal compliance and thresholds for urgency and pain measured using the barostat, FBA concentrations and faecal tryptase were assessed using two-way analysis of variance (ANOVA) for analysis of treatment and time effects and t-test, Mann–Whitney U-test and Wilcoxon matched pairs test for paired comparisons between normally and non-normally distributed variables, respectively. In addition, exploratory mediator analyses were planned to explore whether treatment effects, in terms of changes in urgency or pain, are mediated through changes in FBAs or protease using Spearman/Pearson correlation as appropriate.

Recruitment strategies

A number of strategies were adopted during the trial to address the recruitment difficulties that were encountered. The initial plan had been to open 11 secondary care centres for recruitment, but it soon became clear that this would not be adequate to achieve the necessary numbers required to complete the trial. The source of patients varied widely by site (see Appendix 3, Table 37). As well as increasing the number of secondary care sites to 16, we utilised numerous other strategies to attract more patients to the trial.

We opened a number of secondary PICs. The secondary care PICs would identify potential participants from their clinics and refer them on to a recruiting centre for consideration for the trial. A total of six such PICs were opened.

A key issue in recruitment of patients within secondary care seemed to be changes in referral patterns over recent years. A trial in Yorkshire showed the value of screening patients with diarrhoea with calprotectin in primary care, reducing unnecessary colonoscopy and this has strengthened existing National Institute for Health and Care Excellence (NICE) guidance42 on the use of calprotectin in primary care. Evidence from NICE suggests a substantial reduction in referral from primary care with recent onset diarrhoea (56% in Cannock Chase, Staffordshire, UK). 43 Patients are thus less likely to be referred into secondary care for investigation of diarrhoea but are being investigated and managed in primary care. We soon realised that we would need to widen our recruitment pool to include patients in primary care and also in the general population.

We set up a self-referral pathway to allow patients to refer themselves to a local recruitment centre to be considered for the trial. Advertising was used to increase awareness of the trial and interested patients could visit the TRITON trial website for further information. The website included a simple self-screening questionnaire to confirm whether the patient was broadly eligible for the trial (see Appendix 10, Figure 31). Potentially eligible patients could then contact one of the trial research fellows to obtain further information about the trial and discuss their suitability before being referred on to their local recruiting centre.

We used a number of strategies to reach patients outside the secondary care centres. We approached GP practices in the areas surrounding the recruiting centres to set them up as PICs. The practices were asked to perform a search of their patient database to identify any suitable patients for the trial. They would then mail out a letter and leaflet introducing the trial to the patient asking them to visit the trial website if they were interested. Between March 2019 and February 2020, a total of 26 GP practices were set up as PICs and mail outs were sent to a total of 1381 patients. Of all the patients screened, two were identified in this way and of these one was randomised into the trial.

Starting in October 2019, we advertised within GP practices and pharmacies using posters and leaflets. GP practices that agreed to participate as PICs were asked to display the promotional materials in their waiting rooms. We also approached the Clinical Research Networks to arrange for promotional materials (see Appendix 10, Figures 32 and 33) to be displayed in the waiting rooms of GP practices not involved as PICs, as well as pharmacies.

In addition, we instigated a publicity drive to increase awareness of the trial outside of clinical settings. To achieve this, we advertised in local press in the areas surrounding our recruiting centres. The first advertisement drive was in May 2019 and regular advertising continued until March 2020. This advert (see Appendix 10, Figure 34) consisted of a quarter page in colour in publications such as the Metro, Manchester Evening News and London Evening Standard.

In November 2019, a Communications Officer was appointed to assist with the advertising strategy for the study, being responsible for engaging with Twitter (Twitter, Inc., San Francisco, CA, USA), TV, newspaper and radio, IBS charities and other patient forums to publicise the study. The trial Twitter account was used to engage with sites and improve awareness within trial teams in the hope that it would help attract more patients to the study. We actively encouraged staff at recruiting centres and collaborators to follow and retweet as well as using appropriate hashtags and tagging any relevant organisations in our tweets. We also advertised on social media, in particular Facebook (Meta Platforms, Inc., Menlo Park, CA, USA) and Google (Google Inc., Mountain View, CA, USA).

Professor Spiller and Karen Andrews (patient and public involvement representative) did an interview for the Mail on Sunday, which was the lead article on their health page on Sunday, 23 March 2019 and produced 3000 visits to our webpage and some 430 contacts of which approximately 10 entered the trial.

ContactME-IBS is a registry run through one of the TRITON recruiting sites, County Durham and Darlington NHS Foundation Trust. ContactME-IBS holds the contact information of approximately 2766 adults interested in hearing about, and taking part in, IBS research, of which 641 were suitable for TRITON having been identified as having IBS-D and associated pain (a key eligibility criterion). A mail shot to the 641 potentially eligible registry patients was sent in November 2019. Due to the nature of the registry setup, patients are mainly centred around areas where ContactME have had a drive to recruit GP practices, with 593 of the potentially eligible registry patients located in areas affiliated with TRITON secondary care sites, and 48 in areas without a suitable TRITON site.

Summary of patient identification method and screening approach is in Table 1.

Screening and recruitment

A total of 1582 potential participants were screened for entry into TRITON, across 13 hospital sites, including those referred to local sites by the Clinical Research Fellows (section Self-referrals below). Of these, 295 (18.6%) were found to be eligible for the study, 173 (58.6% of eligible) consented to take part in the study, 149 (86.1% of consented) were registered to the study, and 80 (53.7% of registered) were randomised to the study. Eligibility varied widely between centres from 10% to 90% with some being more efficient than others in deciding who to approach (see Appendix 3, Table 36). A detailed flow of these patients in the form of a CONSORT diagram can be seen in the supplementary document. A large proportion, 55.8% (882/1582), of screened patients were ineligible due to not meeting the Rome IV criteria. Of those who passed screening and were registered, insufficient pain scores were the largest single cause of being ineligible for randomisation (31/149). Overall, patients were screened over 28 months from March 2018 until June 2020.

Self-referrals

A total of 620 people self-referred via the TRITON website (Figure 2) and were sent information sheets. Subsequently, 297 (47.9%) did not respond further. A total of 155 (25.0%) were referred onto a local site, having confirmed their interest in taking part in the TRITON study and fulfilled the initial eligibility checks.

FIGURE 2.

Self-referral flow of participants.

The trial Clinical Research Fellows responded to each patient’s e-mail, and 64 patients were screened through the formal screening process. Twelve patients were randomised using the self-referral route (see Table 1). Self-referrals were received and screened over 20 months from November 2018 until June 2020.

Screening characteristics

Characteristics of screened participants (age and gender) summarised by site are detailed in Table 2. One site in particular screened more potential participants than the others, 49.2% (778/1582) of the total; but most of these screening forms were missing age and gender data. The percentage who were eligible varied from 0.6% to 91.7% of those considered indicating that some sites were much more selective in who they approached. For those potential participants with screening data available, the mean age of those screened was 44.9 years old (SD 15.4), and 67.1% were female (523/780). Further screening information by site can be found in Appendix 3, Tables 36–38.

Randomisation and participant flow

Registration data by site are provided in Appendix 3, Table 39. Participants were randomised over 23 months from July 2018 until May 2020 (Figure 3). Thirty-seven (46.3%) participants were randomised to ondansetron and 43 (53.8%) to placebo (see CONSORT diagram). By the end of treatment at 12 weeks, six participants discontinued treatment, four in the ondansetron arm (one of these patients withdrew from follow-up) and two in the placebo arm.

FIGURE 3.

Overall recruitment.

Baseline

At baseline, randomised participants were of a similar age to all screened participants, mean age 43.9 and 44.9 years, respectively. Overall, a higher proportion of males were randomised compared with the proportion of males screened; 41.3% and 32.9%, respectively. However, a high proportion of screening records had no age and sex data making comparisons uncertain (Table 4). In both arms, males and females reported similar non-gastrointestinal somatic symptoms associated with IBS measured by PHQ-12 (Table 5).

During the pre-treatment period, participants randomised subsequently to ondansetron reported, on average, higher abdominal pain scores (61.4 vs. 55.2 mean pain score), slightly more days per week with a loose stool (5.9 vs. 5.4 mean days per week) and a higher urgency to defaecate score (67.5 vs. 60.4 mean urgency score) than those randomised to placebo (see Table 5). Pre-treatment diary scores by site are provided in Appendix 3, Table 40.

Table 4 shows that the participants were well matched for demographics across randomised treatment arms. Demographic data by site are provided in Appendix 3, Table 41.

As part of the trials mechanistic substudy, participants were offered a barostat procedure and a colonic manometry. Eighteen participants underwent a barostat and 13 a colonic manometry assessment. Of these, 12 undertook both manometry assessments (see Table 5). Mechanistic test uptake by site is provided in Appendix 3, Table 42.

Further baseline summaries including summaries across time points are in Summaries and analyses of the secondary outcomes.

Losses and exclusions after randomisation

Analyses of the primary outcome

Analyses were conducted on the ITT population and all 80 participants were included in the primary analysis. There were 15/37 [40.5%, 95% CI (24.7%, 56.4%)] primary end-point responders in the ondansetron arm and 12/43 [27.9%, 95% CI (14.5% to 41.3%)] in the placebo arm. Four participants (two in each arm) did not provide sufficient data to calculate the primary end point and were therefore assumed to be non-responders. Multiple imputation was not used, as there were only 5% of missing responses the potential impact of the missing data would be small44 and the overall sample size was small.

Considering the individual components of the primary end point; there were 17/37 [46.0%, 95% CI (29.9% to 62.0%)] on ondansetron who were pain responders (i.e. met FDA criteria for reduction in pain intensity) and 16/43 [37.2%, 95% CI (22.8% to 51.7%)] on placebo. As regards stool consistency, 25/37 [67.6%, 95% CI (52.5% to 82.7%)] on ondansetron and 22/43 [51.2%, 95% CI (36.2% to 66.1%)] on placebo were stool consistency responders according to FDA criteria (Tables 9 and 10). There was no statistical evidence of a difference in the FDA-defined primary end-point responder rate between arms. The odds ratio from the logistic regression model, adjusted for the treatment group, minimisation variables (undergoing manometry or barostat), age and gender was 1.93 (95% CI 0.73 to 5.11, p-value 0.1869). Site was excluded from the model due to model convergence issues (see Table 10). Additionally, when assessing individual components of the primary end point, ondansetron had greater effect on stool consistency improvement than on pain intensity reduction when comparing between arms. However, none of these differences were statistically significant.

Supportive and sensitivity analyses

Analysis of the primary outcome on the PP population supported the primary analysis result that there is no evidence of a difference in the treatment arms (see Table 10). Throughout the study (pre-treatment, treatment and follow-up), participants were allowed to use rescue medication (loperamide) no more than twice per week. Similar proportions of participants used loperamide during pre-treatment and follow-up (off-treatment) across arms, but a higher proportion used loperamide in the placebo arm during treatment; 39.5% (n = 17) compared with 18.9% (n = 7) in ondansetron (see Appendix 6, Table 48). Therefore, as part of the sensitivity analyses, use of rescue medication was added as an independent variable and the primary analysis model was repeated. The outcome of this analysis is consistent with the primary analysis finding that there is no difference in the primary end point of FDA-defined responder rates between the treatment arms. Additionally, loperamide use was added as an interaction term with the treatment allocation and the findings were again similar. The number of responders using rescue medication is summarised in Appendix 6, Table 49; higher proportions of both responders and non-responders in the placebo arm used loperamide. Sensitivity analysis including HADS anxiety and depression scores as independent variables in the primary end-point model supports the conclusion of no evidence of a statistical difference between the treatment arms (see Table 10).

Summaries of overall responders and responders to individual components of primary outcome by week

A graphical summary of responders by week is provided in Figures 4–6. Weeks 1–12 are the treatment weeks and weeks 13–16 are the post-treatment follow-up weeks. Based on those summaries, participant response increased during the weeks in treatment in both treatment arms and decreased in the follow-up period. The largest difference between treatment arms is in stool consistency; a higher proportion of participants in the ondansetron arm were responders compared with those in the placebo arm (see Figure 6). This effect was apparent already in the first week (see Appendix 7, Table 54).

FIGURE 4.

Overall responders by week.

FIGURE 5.

Weekly abdominal pain responders.

FIGURE 6.

Weekly stool consistency responders.

The number and proportion of participants using loperamide as a rescue medication by week are summarised in Figures 7 and 8. Additionally, the proportion of participants using loperamide was added to the figures to complement the sensitivity analyses. Tables summarising number and proportion of responders by week are in Appendix 7, Tables 52–56.

FIGURE 7.

Loperamide use (proportion of participants).

FIGURE 8.

Loperamide use (number of participants).

Abdominal pain

Mean abdominal pain scores were higher in the ondansetron arm prior to randomisation compared with placebo (see Table 5), but were similar across the treatment arms during the treatment and showed a similar increase in the follow-up period (Table 11), indicating a greater reduction in pain scores in the ondansetron arm during treatment (see Appendix 6, Tables 44 and 53). However, there was no evidence of a statistically significant difference in abdominal pain scores between the treatment arms during or after the treatment periods (Tables 12–14).

Urgency

Mean urgency of defaecation at baseline was higher in the ondansetron arm compared with placebo (see Table 5). During the treatment period and in the follow-up, unadjusted urgency scores were slightly higher in the placebo arm compared with ondansetron, having reduced more in the ondansetron arm from baseline (see Table 11). Urgency scores increased in both arms in the follow-up period compared with the treatment period. There was no evidence of a difference in urgency of defaecation scores between the treatment arms during or after the treatment periods (see Tables 12, 13 and 15).

Irritable bowel syndrome severity scoring system

At baseline, participants on ondansetron reported higher IBS symptom severity scores compared with those on placebo. At the end of the treatment period, symptom severity reduced more in the ondansetron arm and a higher proportion of participants in the ondansetron arm reported symptoms that were in remission or mild compared with placebo (Table 17). A higher proportion of participants in the ondansetron arm 19/31 (61.2%) than in the control arm 18/43 (41.8%) reported a reduction of 50 or more points in IBS-SSS at 12 weeks compared to baseline, but this was not significant (p = 0.15), Fisher Exact test (see Appendix 6, Table 47). There was no evidence of statistically significant differences between ondansetron and placebo in IBS-SSS scores at the end of treatment period (see Table 14).

Short-form Leeds dyspepsia questionnaire

Participants in the ondansetron arm at baseline reported higher severity of dyspeptic symptoms compared with placebo. At the end of 12 weeks, both groups reported reduction of severity of symptoms, although the reduction was greater in the ondansetron arm. The largest reduction was in symptoms of indigestion and nausea (see Appendix 6, Table 50). There was evidence of a statistically significant difference between ondansetron and placebo in the total SFLDQ score, with an adjusted mean difference in scores, –3.2 points [SE 1.43, 95% CI (–6.1 to –0.4), p = 0.0275], suggesting that participants in the ondansetron arm experienced lower severity of dyspeptic symptoms (see Table 14). However, the result needs to be considered with caution due to the small sample size and missing data.

Irritable bowel syndrome quality of life

At baseline, participants in the placebo arm reported better IBS-specific QoL compared with those in the ondansetron arm. At 12 weeks, participants in both arms reported improvements in QoL, those in the ondansetron arm reported a greater improvement; there was improvement reported in all subscales of the IBS-QOL (see Appendix 6, Table 51). There was no statistically significant difference in IBS-specific QoL between the treatment arms (see Table 14).

Satisfactory relief of IBS symptoms (weeks 1–12)

There was no evidence of a statistically significant difference between treatment arms in terms of satisfactory relief of IBS symptoms (Table 18).

Use of rescue medication – number of days taken loperamide over weeks 1–12

Loperamide was used as a rescue medication by only a small per cent of subjects, mostly those that were not treatment responders (see Appendix 5, Table 55 and Figures 28 and 29). The number of days participants took loperamide over weeks 1–12 was planned to be evaluated. However, data on the number of days were right-skewed with considerable proportion of participants not using loperamide at all (see Appendix 7, Table 55). We attempted to fit two-stage model; however due to low frequencies, we were unable to fit it. Logistic regression was used instead (participant used loperamide 0 days or used it at least for 1 day during the time period). There was no evidence of statistically significant differences between the treatment groups in using loperamide during treatment or post-treatment period (see Table 18). Abdominal pain response and stool response according to loperamide use are detailed in Appendix 5, Figures 30 and 31).

Weekly summaries

Figures 9–14 provide graphical summaries of abdominal pain, stool frequency, stool consistency, number of days with loose stool and satisfactory relief by week. The largest observed difference between the treatment arms was in mean stool consistency during the treatment period. Participants in the ondansetron arm had improved stool consistency measured by the BSFS compared with those in the placebo arm; these differences diminished in the post-treatment period in weeks 13–16 (see Figure 12).

FIGURE 9.

Abdominal pain by week.

FIGURE 10.

Stool urgency by week.

FIGURE 11.

Daily number of stools by week.

FIGURE 12.

Mean stool consistency by week.

FIGURE 13.

Number of days/week with loose stools by week.

FIGURE 14.

Satisfactory relief by week.

Tabulated summaries of numbers, proportions, means and SDs of the end points described above are in Appendix 7, Tables 57–61.

Repeated measures models

In all repeated measures models, a week was fitted as a time variable and interaction between time and treatment arm was included in the model. Unstructured covariance structure provided the best fit for models. Models were run for all 12 weeks of treatment, with numerical results presented for the last 4 weeks of treatment due to space issues. The small number of participants needs to be considered in interpretation of repeated measures models described below.

Abdominal pain

There were no statistically significant differences between the treatment arms in weekly abdominal pain scores (Table 19 and Figure 15).

FIGURE 15.

Abdominal pain, output from repeated measures model, adjusted scores and 95% CI by week.

Stool urgency

There were no statistically significant differences between the treatment arms in stool urgency scores (see Table 20 and Figures 10 and 16).

FIGURE 16.

Stool urgency, output from repeated measures model, adjusted scores and 95% CI by week.

Urgency

Urgency responders were defined as participants that reported reduction in urgency score by at least 30% in at least 6 weeks of 12 weeks treatment. There was no evidence of a difference in proportions of responders between the treatment arms (see Appendix 6, Table 45 and Figure 17).

FIGURE 17.

Urgency responders by treatment arm.

Stool frequency

We were unable to fit the Poisson model due to dispersion, and a negative binomial model did not converge, so stool frequency was fitted using a linear regression model instead as stool frequency was fitted as continuous data. There were no statistically significant differences in daily stool frequency between participants in the ondansetron and placebo arms (see Table 21, Appendix 7, Table 58, Figures 11 and 18).

FIGURE 18.

Stool frequency, output from repeated measures model, adjusted scores and 95% CI by week.

Stool consistency

Due to the small number of participants, we were not able to fit a proportional odds model for stool consistency measured by BSFS. A linear model was fitted instead. Values from weeks 9–12 are shown in Table 22 and Figure 19. There was no evidence of statistically significant differences in mean daily stool consistency between participants in the ondansetron and placebo arms, Allocation*week interaction, p-value 0.3988.

FIGURE 19.

Stool consistency, output from repeated measures model, adjusted scores and 95% CI by week.

Number of days with loose stools

Number of days/week with loose stool was fitted as a model outcome. A Poisson regression model did not converge and so a linear regression model was used instead. There was no evidence of a statistically significant difference in the number of days with loose stools between the treatment arms (see Table 23, Appendix 7, Table 60, Figures 13 and 20).

FIGURE 20.

Number of days with loose stool/week, output from repeated measures model, adjusted scores and 95% CI.

Relief model

A logistic regression model was fitted; however odds ratios between the treatment arms need to be interpreted with caution due to missing data. Estimated treatment odds ratio varied by week (i.e. below and above 1) and there was no evidence of a statistically significant difference between arms.

Loperamide use model

A logistic regression model was fitted to the data and similarly to the relief model, odds ratios need to be interpreted with caution due to missing data and the small number of participants. Estimated treatment odds ratio varied by week and there was no evidence of a statistically significant difference between treatment arms.

Dose titration

Participants were allowed to self-titrate during the treatment period. Based on clinical experience, it was considered unlikely that participants would need to alter their dose after 6 weeks. The recommended titrated dose at 2 weeks, the requested dose at 6 weeks and the number and percentage of patients taking a constant dose from 6 weeks onwards are reported in Appendix 8, Table 67. A higher proportion of participants in the placebo arm required higher titrated doses of IMP after 2 weeks of treatment. Consequently, a higher number of bottles was requested for placebo participants at 6 weeks of replenishment.

In the majority of cases 54/80 (72%) overall, the number of capsules returned at 12 weeks was not consistent with the dose at 6 weeks (see Appendix 8, Table 62).

Unplanned trial drug usage

No incorrect bottles were provided to participants and there was one replacement of two damaged bottles.

Blinding and exit poll

There were no unblindings reported.

An exit poll completed by the study nurse is summarised in Table 24. There are differences in blinding index between the treatment arms. The blinding index45 ranges from –1 to +1 and a blinding index of 0 indicates compete blinding. The blinding index from the study nurse perspective was 0.40 [95% CI (0.17 to 0.63)] in the ondansetron arm compared with 0.10 [95% CI (–0.15 to 0.35)] in the placebo arm. As the lower limit of the 95% CI for the ondansetron arm is above zero; this suggests failure in masking in the ondansetron arm.

An exit poll completed by participants is in Table 25. There are differences in blinding index in participants’ responses. The participant blinding index in ondansetron was 0.42 [95% CI (0.18 to 0.67)] and in placebo 0.05 [95% CI (–0.18 to 0.28)], suggesting unblinding in the ondansetron arm.

Possible explanations for this occurrence in terms of response could be that participants in the ondansetron arm were more likely to experience (and/or report to the study nurse) some therapeutic effects. However, results of the exit poll should be treated with caution due to the small number of participants.

Certainty of choice on the scale 1–10 and reasons for choice which were completed by the site study nurse are summarised in Appendix 9, Tables 68 and 69.

Faecal bile salts

Faecal bile acids were assessed at baseline as shown in Table 32. Stools with more watery stool had lower total bile acid concentration, correlation with stool % water, Spearman’s r = 0.56, N = 48.

As Table 33 shows, at baseline there was wide variability. After 11–12 weeks of treatment, there was a numerical fall in primary bile acids on both ondansetron and placebo with a non-significant rise in the secondary/primary ratio.

Using the total sample of 51 at baseline, there was no correlation between the secondary/primary bile acids ratio and WGTT, Spearman’s r = –0.08 (0.39–0.75), p = 0.9.