Ethics

This study was approved by the Medical Ethics Review Committee Utrecht, the Netherlands (NL75824.041.20), and was registered at ClinicalTrials.gov (NCT04790422). The study was conducted in accordance with the principles of the Declaration of Helsinki (Fortaleza, Brazil 2013), and the requirements described in the EU Clinical Trials Directive 2001/20/EC, transposed in the Revision of the Dutch Medical Research involving Human Subjects Act (WMO, effective as of 1 March 2006).

Study design and procedures

This randomised, double-blind, placebo-controlled intervention study included 70 subjects with IBS, had a total duration of 4 weeks and 5 parallel treatment arms. The primary outcome of this study was the relative abundance of faecal Bifidobacterium. Secondary outcomes were microbiota composition, SCFA concentrations, stool frequency, stool consistency, IBS-related complaints, and QoL. Subjects collected faecal samples at baseline and after the 4-week intervention. After collection, the samples were immediately frozen at home, and transported to the research facility on dry ice within 14 days. Subsequently, the faecal samples were stored at –80 ºC until further analysis. Furthermore, daily questionnaires were completed via an EMA app (LifeData, LLC), to report information about their stool frequency (numerical), stool consistency by using the Bristol Stool scale (BSS, scored from 1 to 7, where 1 indicates constipation and 7 indicates diarrhoea), and IBS-related complaints such as bloating, flatulence, and abdominal cramping (rated on a scale of 0 to 10). Additionally, information about supplement compliance and medication intake was reported on a daily basis in this app. Both at baseline and after the 4-week intervention subjects completed a food frequency questionnaire (FFQ), a validated questionnaire to assess symptom severity (IBS-SSS, ranging from 0 to 500; scores <175 denote mild symptoms, 175–300 moderate symptoms, and >300 severe symptoms),^{(Reference Francis, Morris and Whorwell32,Reference Mujagic, Keszthelyi and Aziz33)} the validated 34-item Irritable Bowel Syndrome Quality of Life questionnaire (IBS-QOL, ranging from 0 to 100; higher scores indicate a better QoL),^{(Reference Andrae, Patrick, Drossman and Covington34,Reference Patrick, Drossman, Frederick, DiCesare and Puder35)} and the validated Hospital Anxiety and Depression Score assessment (HADS, with scores ranging from 0 to 21 for both anxiety and depression; scores ≥ 8 suggest symptoms of anxiety or depression).^{(Reference Zigmond and Snaith36)}

Study subjects

Subjects were recruited in May 2021. The study was performed from June 2021 to July 2021 in the Netherlands and was conducted in a corona-proof setting, completely at home and online. Subjects were recruited using the Wageningen University & Research subject database, by advertisements on social media such as Facebook and Instagram, and via the website and social media of the Dutch IBS patient association. An online information meeting was organised prior to the start of the study to explain all study procedures and measurements. All subjects willing to participate in the study signed the informed consent form. Inclusion criteria were: IBS subjects meeting the Rome IV criteria, age between 18 and 65 years, and BMI between 18.5 and 30 kg/m². The main exclusion criteria were: 1) having gastrointestinal or other relevant diseases that could affect study results, 2) a history of intestinal surgery, 3) antibiotic treatment within the 3 months before the study, 4) use of medication or irregular use of nutritional supplements, such as prebiotics and probiotics, that could interfere with the study outcomes, or 5) following a FODMAP-restricted diet. Eligibility was assessed using an online screening questionnaire. Subjects were enrolled by the study coordinator.

Nutritional intervention products

The four tested nutritional supplements were Resistant starch type 4 (RS, NOVELOSE®3490, Ingredion, USA), Pea Inner Fibre (PIF, I50M, Roquette, France), Chondroitin Sulphate (CS, Bioiberica S.A.U., Spain), and Protein hydrolysate cow milk (PHCM, an αS1-casein hydrolysate, containing α-casozepine, Ingredia, France). The placebo supplement consisted of Maltodextrin (GLUCIDEX® IT 19 P, Roquette, France). Each of these supplements was consumed twice daily, one sachet in the morning and one sachet in the evening taken together with two spoons of banana compote (Frulla, Italy). For all supplements, the most effective and regularly consumed dosage was chosen based on literature and information provided by the providers of the supplements (Table 1). PIF and RS are insoluble fibres, with study dosages set at 20 g/d for both RS and PIF (PIF consists of 10 g fibre, the other 10 g consists of starch and protein). The CS, a soluble fibre supplement, was given in a daily dose of 1.2 g and the PHCM in a daily dose of 300 mg. Both CS and PHCM supplements were mixed with Maltodextrin to make supplements of 10 grams per sachet. The second dose of 10 grams, which was consumed in the evening, consisted of the placebo supplement for the CS and PHMC arms. Because a sudden increase in fibre might lead to GI complaints, subjects in all arms had a five-day run-in period where they consumed only 1 sachet in the morning, followed by both sachets for the remainder of the four weeks. All sachets were uniformly packed, to ensure similarity of the interventions.

Table 1. Overview of supplement dosages per day

Sample size and randomisation

No data on the effect size of the intervention products tested in the current study were available. Therefore, the sample size calculation was based on data from a previous study investigating the effect of 3.5 gram Galactooligosaccharides (GOS) on the relative abundance of Bifidobacterium in individuals with IBS (relative abundance; mean difference (delta) = 2.26, SD of the difference = 0.67).^{(Reference Silk, Davis, Vulevic, Tzortzis and Gibson37)} To account for the possibility of a smaller effect in our study, we assumed approximately half this effect size for our calculation, resulting in a mean difference of 1.1, with a variance of 0.6. Based on a power of 80% and a significance level of 0.05, 12 subjects per treatment arm would be needed to demonstrate a significant difference. To compensate for potential drop-outs, 14 subjects per treatment arm were included (70 subjects in total, divided across 5 arms).

Subjects were stratified according to age (2 groups) and BMI (2 groups) and randomly assigned to one of the five treatment arms using block randomisation by the study coordinator. Blocks of five subjects were created, and the allocation sequence within each block was generated using random numbers in Excel. An independent researcher assigned a unique code to each of the study products. All researchers of the project team, with the exception of the independent researcher, were kept blind to the assignment of treatment, and so were the study subjects. After database lock and signing of the statistical analysis plan, the key to the treatment allocation was shared with the other researchers.

Microbiota analysis

DNA was isolated from 0.25 gram of wet faeces using a previously published protocol.^{(Reference Salonen, Nikkilä and Jalanka-Tuovinen38)} Isolated and purified DNA was used for PCR amplification of the V4 region of 16S rRNA gene with barcoded primers 515F (5’-GTGY CAGC MGCC GCGG TAA-3’)^{(Reference Parada, Needham and Fuhrman39)} and 806R (5’-GGAC TACN VGGG TWTC TAAT-3’)^{(Reference Apprill, McNally, Parsons and Weber40)} in duplicate. PCR was performed in 50 µl reaction mix containing 10 µl 5x HF buffer (Thermo Fisher Scientific, Vilnius, Lithuania), 1 µl dNTPs (10 mM, Thermo Fisher Scientific), 0.5 µl Phusion Hot start II DNA polymerase (2 U/µl, Thermo Fisher Scientific), 36.5 µl nuclease-free water (Promega, Madison, WI, USA), 1 µl DNA template (20 ng/µl), and 1 µl sample-specific barcoded primer (10 µM).^{(Reference Ramiro-Garcia, Hermes and Giatsis41)} PCR was performed as follows: 98 ℃ for 30 s, followed by 25 cycles of 98 ℃ for 10 s, 50 ℃ for 10 s, 72 ℃ for 10 s, with a final extension of 7 min at 72 ℃.^{(Reference Ramiro-Garcia, Hermes and Giatsis41)}

The PCR product was purified with the CleanPCR kit (CleanNA, The Netherlands), and quantified using the Qubit dsDNA BR Assay kit (Invitrogen by Thermo Fisher Scientific, Eugene, OR, USA). An equimolar mix of purified PCR products was prepared and sent to Novogene (Cambridge, United Kingdom) for sequencing. Raw sequencing data were processed with NG-Tax pipeline 2.0 with default settings.^{(Reference Ramiro-Garcia, Hermes and Giatsis41,Reference Poncheewin, Hermes, Van Dam, Koehorst, Smidt and Schaap42)} Amplicon Sequencing Variant (ASV) picking and taxonomic assignments were performed using the SILVA 138.1 database. Raw sequencing data are available at the European Nucleotide Archive under accession number PRJEB56219.

Short-chain fatty acid analysis

The concentrations of acetate, propionate, and butyrate were measured in faecal water, using High-Performance Liquid Chromatography (HPLC, Shimadzu Prominence-I LC2030C-Plus, Shimadzu, Duisburg, Germany) equipped with a Shoedex SH1821 column (Showa Denko, New York, USA) and RID-20A refractive index detector (Shimadzu).

To prepare the sample, 0.2 grams of wet faeces was mixed with 0.8 ml of distilled water, and subsequently centrifuged at 21130 RCF for 5 min at 4°C. The supernatant went through a pre-treatment step with Carrez reagents (A and B) to deproteinise the samples. The Carrez A solution consisted of 0.1 M K₄Fe(CN)6·3H₂O and the Carrez B solution consisted of 0.2 M ZnSO₄·7H₂O. Both solutions were stored at 4°C. Next, 500 µl of supernatant was mixed with 250 µl cold Carrez A solution, followed by mixing with 250 µl cold Carrez B solution. Afterwards, the sample was centrifuged at 15000 RCF for 5 min at 4°C, and the clear supernatant was collected for measurement. The HPLC conditions included eluent composition of 0.01 N H₂SO₄, eluent flow rate set at 1 ml/min, column oven temperature maintained at 54°C, flow rate of 0.8 ml/min, and utilisation of 10 mM DMSO as the internal standard. All data were processed by the Chromeleon software (version 7.2.9, Thermo Fisher Scientific).

Data and statistical analyses

Intention-to-treat (ITT) analyses were performed for all outcomes and for all randomised subjects who received any study product and who completed all four weeks of the intervention period (see Fig. 1). Per protocol (PP) analyses excluded subjects with major protocol deviations related to medication intake or illnesses, that interfered with the study outcome, or with a low questionnaire compliance or intervention product compliance. Intervention product compliance was reported by subjects themselves, using the daily questionnaire in the app.

Figure 1. Study participant Flow-chart. Flowchart of study subjects from recruitment and screening to final per protocol (PP) data analyses.

Continuous data are presented as mean ± standard deviation if normally distributed or median [interquartile range] when skewed. All questionnaire data were analysed with R (version 1.3.959), microbiota data were analysed using R (version 4.2.2). All statistical analyses were performed on blinded data, except for the placebo group.

16S rRNA gene read count data were transformed to microbial relative abundance, as implemented in the Microbiome R package.^{(Reference Leo and Sudarshan43)} To measure the change in relative abundance of Bifidobacterium over time, fold changes were calculated. To determine differences in relative abundance of Bifidobacterium over time between the placebo and treatment groups, a paired sample T-test was used when the data was normally distributed, and the paired Wilcoxon test was used when the data were not normally distributed.

Microbial diversity indices including Inverse Simpson (richness and evenness of species in a community) and Faith’s Phylogenetic Diversity (phylogenetically weighted measure of richness) were calculated at ASV level, as implemented in the Phyloseq^{(Reference McMurdie and Holmes44)} and Picante^{(Reference Kembel, Cowan and Helmus45)} R packages, respectively. Since these data were not parametric, (un)paired Wilcoxon test was conducted to determine the difference between groups. Pairwise weighted Unifrac^{(Reference Lozupone, Hamady, Kelley and Knight46)} and unweighted UniFrac^{(Reference Lozupone and Knight47)} distance-based principal coordinates analysis (PCoA) were used to visualise microbial community variation at the ASV level. Permutational multivariate analysis of variance (PERMANOVA) was used to test for significant differences between groups, as implemented in the Vegan R package.^{(Reference Oksanen, Blanchet and Kindt48)} Differences in microbiota data between week 0 and week 4 were compared with the paired sample T-test when normally distributed or the paired Wilcoxon test when not normally distributed. RDA analysis was performed with the Vegan package. All figures were generated using the ggplot2 package.^{(Reference Wickham and Wickham49)}

Unpaired T-test was used to compare the post-intervention differences for faecal SCFA concentration data between placebo and treatment groups. Unpaired T-test was also used to compare differences between each treatment group and the placebo group as implemented in the rstatix R package.^{(Reference Kassambara50)}

Stool frequency, stool consistency, and IBS-related complaints (abdominal pain, bloating, and flatulence) were reported daily. For stool consistency, a daily mode per individual was calculated. Differences over time between the treatment groups and placebo (Time/Intervention interaction) were assessed using linear mixed models for repeated measures with subject as random effect. Questionnaire data on symptom severity, QoL, and HADS, reported at baseline and after the 4-week intervention, were analysed using linear mixed models, with time and treatment as main effects and subject as random effect. Interaction effects were also examined. Within-group effects for questionnaire data were analysed using paired sample T-test. To test for changes in fibre intake over time, delta values were calculated and analysed using one-way ANOVA.

False discovery rate (FDR) correction based on Benjamini–Hochberg procedure was used to correct for multiple testing, when applicable for microbiota and SCFA markers. In all cases, (adjusted) P-values < 0.05 were considered statistically significant, and adjusted P-values between ≥ 0.05 and < 0.1 were considered a trend.