Setting and recruitment

The planned trial is part of the project INDICATE-FH (improving diagnostics and therapy of food hypersensitivity, project no. 01EA2109A) consisting of seven subprojects. Patients suspected of having NCWS or a self-reported ARF—the trigger of which has not yet been found—are eligible for the study if they showed a positive reaction in the previous CLE-based endoluminal provocation test. In this test, provocation of the intestinal mucosa with milk, yeast, egg, soy and wheat, each dissolved in a 0.9% NaCl solution, will take place. However, the CLE measurement as such will not be part of this study.

The study is a prospective, double-blind, monocentric, randomised and controlled study. All examinations and personal meetings take place at the University Hospital of Schleswig-Holstein, Campus Lübeck. The study is conducted in compliance with the Declaration of Helsinki, the current revision of the International Conference on Harmonisation Topic E6 and the Guidelines for Good Clinical Practice. It will be reported according to current Standard Protocol Items: Recommendations for Interventional Trials and Consolidated Standards of Reporting Trials recommendations.15 16

Inclusion and exclusion criteria

Inclusion and exclusion criteria of the study are as follows:

Standard diet and elimination diet

The standard diet of this study is an oligoantigenic basic diet. Patients will not be allowed to consume milk, egg, wheat, yeast or soy products. Patients in the tailored diet group must avoid the allergen for which the CLE result was positive. The dietary recommendations will be communicated to participants through the Perfood Labs application (app) provided by Perfood, Lübeck. They include the forbidden allergen(s) and examples of food products in which the allergen(s) can be found. Furthermore, the app has a database of different recipes available to the patients, allowing them to cook at home. There is also a guideline on how to order food in a restaurant. This will enable both groups to have a varied and balanced diet despite dietary restrictions.

Randomisation and concealment of allocation

Potential participants will be primarily identified by the INDICATE-FH subproject 1. In that subproject, patients with GI problems are subjected to a detailed diagnostic flow including allergy routine in vitro tests, carbohydrate malabsorption, gastroendoscopy and finally CLE to find a culprit of their symptoms. The results of the CLE will not be shared with the patients to ensure that they remain blinded for the duration of this clinical trial. Eligible study participants will be randomised blinded to their CLE results into two groups (1:1 allocation) by block randomisation with variable block length stratified by sex (female/male). To unlock the app, patients will receive a unique one-time activation key depending on the randomisation and CLE results. Both groups will receive dietary recommendations via the app after a successful app activation.

Registration and randomisation will be carried out centrally at the Institute of Medical Biometry and Statistics (IMBS) using standard operating procedures to guarantee the concealment of allocation. The activation keys from Perfood will be transferred from the IMBS to the trial site after the completion of the registration form for a patient by the site. This will also ensure the blindness of the medical personnel.

Screening, baseline and follow-up examinations

An overview of the design and visits of the study is shown in figure 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Study design overview. The dashed boxes represent components that are not in a direct focus of the study. The top box (dashed) highlights a portion of subproject 1 within the larger framework of INDICATE-FH, while the bottom box (dashed) outlines the design for the follow-up phase of the study. BIA, bioelectrical impedance analysis; BMI, body mass index; BSFS, Bristol Stool Form Scale; CLE, confocal laser endomicroscopy; EQ-VAS, EuroQol-Visual Analogue Scale; FAQLQ-AF, Food Allergy Quality of Life Questionnaire–Adult Form; FU4/FU8/FU12, 4/8/12 weeks after FU0 (first entry in the app in phase 2); IBS, irritable bowel syndrome; IBS-QoL, IBS-Quality of Life; IBS-SSS, IBS-Symptoms Severity Score; NCWS, non-coeliac wheat sensitivity; W4/W8/12, 4/8/12 weeks after W0 (first entry in the app in phase 1).

In the first study phase, the patient’s fulfilment of the inclusion and exclusion criteria will be checked in the site at screening. If the criteria are met and the written informed consent is given, the patients will be randomised, and the baseline characteristics will be collected. For this purpose, the body composition and body mass index (BMI) will be recorded. Blood, stool and urine will be collected for proinflammatory profile, microbiome and metabolome analyses. Via email, the patients will receive an activation key for the app, in which they will complete the questionnaires Food Allergy Quality of Life Questionnaire–Adult Form (FAQLQ-AF), Irritable Bowel Syndrome-Symptoms Severity Score (IBS-SSS), IBS-QoL, EuroQol-Visual Analogue Scale (EQ-VAS) and Bristol Stool Form Scale (BSFS). The collection via app is advantageous because it means less effort for the patients (no travelling to trial site), no risk of losing data and the time of completion can be decided by the patients within acceptable time windows. Furthermore, experience with regard to the use of an app within other clinical trials was positive, for example, as described in Lelleck et al.17

The 12-week intervention will begin with the first entry in the app (W0). Further measurement time points will be 4/8/12 weeks after W0 (visits W4, W8 and W12, respectively) and the first visit in presence (V1, 13 weeks after W0). During the first study phase, patients will complete the app’s EQ-VAS, BSFS and adherence questionnaire daily. The app will remind the patients to respond. The FAQLQ-AF, IBS-SSS and IBS-QoL questionnaires will be conducted every 4 weeks. These questionnaires will be available for 4 days. Patients will receive two push notifications to remind them to complete the questionnaires, that is, on W4/W8/W12 and 3 days after the respective visit date. After that, the questionnaires for the corresponding time point can no longer be filled out or edited. The diet diary is to be filled in every week via the app. As soon as symptoms occur, the symptom diary will also be filled out via the app. All symptoms, also those that do not affect the GI tract, should be reported. Medications taken should be reported by patients in the medication log provided in the app. At V1, similarly to V0, the body composition and BMI will be recorded. Blood, stool and urine will be collected as at baseline. The first phase of the trial will end at V1 with the assessment of the primary endpoints.

For continuation in the second phase, patients’ responses to the diet will be analysed after W12 by calculating the change in IBS-SSS from W0 to W12. This information will be automatically sent from Perfood to the IMBS further to maintain the criteria of double blindness in the study. Accordingly, patients with a reduction of ≥50 points will be defined as respondents, the others as non-respondents as described by Francis et al.18 This will lead to four groups:

• Group 1: non-respondents in the standard diet group will not participate in the second phase because an improvement of the symptoms is not expected. Patients will be unblinded to their CLE result at V1.

• Group 2: respondents in the standard diet group will receive nutritional recommendations based on their CLE results to determine whether a further improvement of symptoms and QoL will occur following the new recommendations.

• Group 3: non-respondents in the tailored diet group will receive the opportunity to follow the standard diet to determine whether symptoms will improve following the new recommendations.

• Group 4: respondents in the tailored diet group will continue their tailored diet to determine whether symptoms and QoL will improve or remain stable.

Patients of groups 2–4 will remain blinded until the study ends. These patients will receive new activation keys depending on their respondent/non-respondent status and their randomisation result.

The second study phase will start again with the first entry in the app (FU0) after V1. Further measurement time points will be 4/8/12 weeks after FU0 (visits FU4, FU8 and FU12, respectively) and the second visit in presence (V2, 13 weeks after FU0). During the follow-up, the FAQLQ-AF, IBS-SSS and IBS-QoL questionnaires will again be completed via the app every 4 weeks, as the weekly food diary, and the EQ-VAS, BSFS, and adherence questionnaire daily. The body composition will be recorded at V2, the same as in V1, and the BMI will be measured. Blood, stool and urine will be collected as before.

Endpoints

Microbiome analyses will be performed at the University of Hohenheim. Metabolite analyses of the urine, blood and stool samples will be performed at the Institute of Chemistry and Metabolomics at the University of Lübeck. The proinflammatory cytokine profile in the blood will be assessed at the Institute of Nutritional Medicine at the University Hospital of Schleswig-Holstein, Campus Lübeck. The IMBS will analyse all other endpoints.

Statistical analysis

All statistical analyses will be described in detail in the statistical analysis plan, which will be finalised before randomising the last patient in the study. No interim analyses or design adaptions are planned.

For all analyses, only patients will be considered who were successfully randomised, gave their informed consent and did not withdraw their consent during the course of the study. Safety endpoints will be evaluated in the safety analysis set. Primary and secondary endpoints will be assessed in the full analysis set (FAS). As sensitivity analyses, all endpoints will be evaluated in the intention-to-treat population and per-protocol (PP) population.

In this trial, non-compliance, that is, discontinuation of following the recommendations or not starting the intervention, can occur. There is a higher risk of study discontinuation in the standard elimination diet group as this diet is very restrictive and hard to follow for a long period of time. Medication intake concerning GI symptoms is allowed during the study if the medication intake will be defined as stable for 3 months at screening. We assumed that both groups would be affected equally. However, intake of new GI medication or higher doses than known and documented in the electronic case report form at baseline will be considered intercurrent events (IEs). Medication intake for non-GI-relevant disease areas only used in the short term will be allowed and will not affect the endpoints. Group switching is not possible per se by the patients’ wish, as the access to the nutritional recommendations will be steered by the app, and the patient will not see the other diet recommendations. Deviations from the recommendations will be counted as protocol deviations (PDs).

Additionally, a wrong activation key could be provided, leading to a group switch or a wrong tailored recommendation. However, this is very unlikely and counted as a PD. For the possible IEs mentioned above, the treatment policy strategy will be used primarily (ICH E9(R1), https://www.ema.europa.eu/en/documents/scientific-guideline/ich-e9-r1-addendum-estimands-sensitivity-analysis-clinical-trials-guideline-statistical-principles_en.pdf, accessed 7 January 2023). Estimands will be only considered for the first study phase.

The first co-primary endpoint will be tested with a linear model assuming variance homogeneity using a two-sided Wald test at a significance level of 5% in the FAS. The corresponding 95% CI will be estimated. For the second co-primary endpoint, the severity of symptoms at W12, an IBS-SSS difference of <50 points will be counted as non-inferior. The inferiority margin was chosen according to Francis et al.18 A stratified t-test will test the corresponding hypothesis at a one-sided significance level of 2.5% in the PP set. Mean difference and corresponding 97.5% CIs will be reported, secondarily two-sided 95% CIs. The efficacy of the tailored diet will be shown if there is a significant difference regarding the first co-primary endpoint and non-inferiority regarding the second co-primary endpoint. If only one of the two hypotheses can be rejected, none of the co-primary results can be interpreted as significant.

The same analyses described for the first co-primary endpoint will be used exploratorily for the secondary endpoints. The median per weekly assessment will be calculated for the VAS.

According to the respective manuals, handling of missing values in the QoL questionnaires will be accounted for. The compliance endpoint for any week is based on counting the number of compliant days within the respective week. The separate endpoint will be set to missing if patients made entries for <70% of the days, that is, for less than 5 out of 7 days. If patients made entries for ≥70% but <100% of the days, the missing compliant days would be imputed by the average of the valid entries. If the primary and secondary endpoints are missing, multiple imputations with at least 50 imputations will be used to impute missing data. The imputation model will include group and present values (usually W0 measurement, if applicable, measurements at W4 and W8) for the first study phase. The combination will be made using Rubin’s rule. In the second study phase, if applicable, missings will only be imputed by the manuals. Otherwise, no imputation will be performed. Exploratory analyses will include the analysis of primary and secondary endpoints in subgroups of age and compliance.

Safety endpoints will be analysed as follows:

• Body compositions, that is, fat mass, total body water, muscle mass, fat-free mass, body cell mass, extracellular mass and dehydration: differences between baseline and V1 and V2 using exploratory paired t-tests, respectively; between-group comparisons for each measurement time point.

• BMI: intraindividual differences between baseline and V1 and V2 using exploratory paired t-tests, respectively.

• AEs/SAEs: absolute and relative frequencies, no statistical testing, tables and listings according to ICH E3 guideline (https://www.ema.europa.eu/en/documents/scientific-guideline/ich-e-3-structure-content-clinical-study-reports-step-5_en.pdf, accessed 7 January 2023).

Proposed sample size calculations

For the calculation of the first co-primary endpoint, we expect the control group of patients who tested positive but in whom the result of the CLE will not be used for recommendations to improve in HRQoL to a similar extent as those patients reported in the study of van der Velde et al19 because those also were tested positive but did not receive a specific recommendation. Therefore, we expect them to improve on average from 4.56 to 4.21 points, thus by 0.35 points, with an SD of about 1 point. For patients in the tailored diet group, we expect an improvement similar to that observed in negatively tested probands in the study by van der Velde et al. This assumes that the improvement in HRQoL is mainly based on better knowledge about which food compounds to avoid, which will also be granted in our intervention. Thus, we expect an average improvement from 4.24 to 3.36 points, thus by 0.88 points, which we pessimistically set to 0.85 points. We assume that both groups show equal variability in change and that the SD is 1 point. To detect these effects using a two-sample t-test, a power of 80% and a significance level of 5%, 64 subjects per group would be required, that is, 128 patients. We assume a dropout rate of 25% because of the long period of 12 weeks for following nutritional recommendations, especially in the case of no improvement of QoL or symptoms, totalling in 172 patients, that is, 86 patients per group.

For the sample size calculation of the severity of symptoms, we used the reported values from Bojarski et al.14 Here, patients with any CLE reaction to food antigens following a gluten-free diet (GFD) had an SD of 94 for the IBS-SSS 2 months after the GFD started. However, the other group, without reacting to food allergens following the GFD, had an SD of 88. We assume a difference of IBS-SSS at W12 of 0 points between the groups and an SD of 100 points (a more conservative assumption than Bojarski et al14 and a minimal clinically relevant effect of 50 points). With a power of 80% and a one-sided significance level of 2.5%, 64 patients per group would be required in the t-test setting, that is, 128 patients. Assuming a dropout rate of 25%, 172 patients in total would be needed.

So, in this trial, 172 patients will be required to confirm both primary hypotheses. nQuery V.4.0 was used for sample size calculations.

Data collection, handling and management

Each patient will receive a study identification (ID) like CLEXXX, where XXX is a three-digit number. This study ID will label and identify all samples collected and generate the email address needed for registration in the app. The key linking the study ID to the patient’s name will be documented in a hard copy in the signed consent forms folder. These folders will be held in a locked cabinet at the Institute of Nutritional Medicine. Access to this folder will be limited to the study director, the study coordinator and the study dietitian. After the study is finished, the key linking the study ID to the patient’s name will be destroyed, and connecting the two will no longer be possible.

The results of the questionnaires, the dietary and symptom diary, and the medication intake will be documented using the Perfood Labs app and stored in the associated database. After the last patient will have completed the study, the database will be closed, and the data will be transferred to the IMBS for analysis. Therefore, the study data are exported from the data management system, stored in a read-only and edit-protected manner, and thus unchangeable. Perfood ensures that it is documented which employees were involved in this process and that only authorised employees have access to this document. The document is archived for the legally required period. To register for the Perfood Labs app, patients receive the email address (CLEXXX@the-inum.com) and the activation key. Perfood does not have access to the patient’s personal data.

For metabolome and microbiome analyses, the group at the Universities of Lübeck and Hohenheim receive the samples labelled with study ID. In addition, they also receive the following data from the patients: gender, age, CLE result and dietary intervention in the last 12 weeks. The results of the metabolome and microbiome analyses are sent directly to the study centre, where they are stored on an external hard drive in duplicate.

The data collected during the V0, V1 and V2 visits will be documented via the electronic case report form. The data transfer is done via a typical web application consisting of a front-end, backend and database. The application is hosted on network-attached storage (NAS) server. Docker containers are located on the application’s NAS running the backend and database. Access to the NAS server is only possible via a corresponding password-protected account. Via the iPad, the user (medical staff) can access the application (or the front-end) and fill out the questionnaire. The front-end is connected to the database via the backend.

To establish the connection between the iPad and the database, the iPad must have an internet connection and be logged into a corresponding NAS account, for which the appropriate rights must be enabled. The patient’s name and date of birth are not stored in the database. They are only entered on the iPad to create a pseudonym via which all visits of the corresponding patient can be found in the database. An outsider who knows the patient’s date of birth and name cannot deduce the pseudonym from this. Subsequently, the data can be extracted from the database via an NAS account with the appropriate rights. In summary, data generation, transmission, storage and analysis of health-related personal data in this study are performed strictly in compliance with the applicable legal provisions on data protection.

Patient and public involvement

The German society of the coeliac disease is a partner in INDICATE-FH and as such is involved in project planning.