A randomised controlled trial of a probiotic ‘functional food’ in the management of irritable bowel syndrome

A randomized placebo controlled trial to compare symptom and quality of life in patients with IBS taking a probiotic or non- probiotic fermented dairy product was conducted in primary care practices.

13 general practices in the Midlands Research Practices Consortium (MidReC) situated within the West Midlands, UK. Patients were identified through searches of electronic patient records for symptomatic individuals with a diagnosis of IBS > 6 months or having received two or more prescriptions in the previous 24 months for anti-spasmodics, bulking agents or laxatives. A confirmatory screening questionnaire based on ROME III criteria [27] was sent to all patients. Lower than anticipated yield, led to administration of the screening questionnaire to all patients aged 18–65 in five of the thirteen practices. No other alterations were made to the study protocol during trial conduct. General practitioners applied discretion to prevent contact with patients considered inappropriate for contact or inclusion, for example individuals known to be in a terminal stage of illness or experiencing acute and severe mental illness.

Study eligibility was assessed from the information provided on the screening questionnaire plus telephone follow-up to verify responses. Individuals were considered eligible for inclusion if they were aged 18–65, met ROME III [27] criteria for a diagnosis of IBS with symptoms being present for > 6 months and reported a constipation element to their symptom profile (either constipation predominant or a mixed symptom profile).

Eligible patients were mailed a symptom diary for completion over the two weeks prior to their randomization appointment. At the baseline recruitment clinic research nurses confirmed eligibility and applied further study exclusions: individuals reporting passing more than 3 soft stools per day in the absence of laxative mediation during the pre-study phase; individuals without problematic symptoms judged by an IBS Symptom Severity Score (IBS-SSS) [28] of <75; individuals with known organic disease or reporting clinical signs of alarm; individuals who were pregnant, breast-feeding or lactose intolerant; individuals with a BMI >35 or <18, as this was deemed suggestive of abnormal diet or function; and those with a recent change of IBS medication (in the preceding month) to reduce likelihood of co-intervention bias.

Blocked randomization lists (1:1 allocation) were produced by the lead statistician (RH) for each general practice site and held centrally. Because of the short shelf-life of dairy products randomised allocation lists were sent to the sponsors for product production prior to randomisation clinics so that two weeks supply of the relevant allocated product was available at the point of randomization. All volunteers gave written informed consent before inclusion in the study.

The test product was a commercially available fermented product containing Bifidobacterium lactis (strain number I-2494 in French National Collection of Cultures of Micro-organisms (CNCM, Paris, France; Danone collection number DN-173 010) together with the two classical yoghurt starters, S. thermophilus (CNCM strain number I-1630) and L. bulgaricus (CNCM strain numbers I-1632 and I-1519). The test product contains 1.25×10¹⁰ colony forming unit (cfu) of Bifidobacterium lactis CNCM I-2494 per cup and 1.2×10⁹ cfu/cup of S. thermophilus and L. bulgaricus.

The control product was a milk-based non-fermented dairy product without probiotics and with similar lactose content to the test product. Both the test and control products were without flavour. Each pot contained 125 g. Both products were specifically prepared for the study and provided by Danone Research (Palaiseau, France). Nurses, GPs, patients and the research team remained blind to product allocation until all analyses had been completed.

In both groups consumption of the product twice daily was instructed and all patients received a two week supply of the allocated product in the first instance and instructions to collect repeat supplies fortnightly for 12 weeks.

All patients returned completed daily diaries at follow-up visits after 4, 8 and 12 weeks. The daily diary assessed stool frequency and consistency (evaluated in accordance with the Bristol stool scale) and bowel movement difficulty (using a 5-point Likert scale which ranged from no difficulty to extreme difficulty). Weekly entries assessed subjective global assessment (SGA) [29, 30] of symptom relief (“Do you consider that in the past week you have had adequate relief of your IBS symptoms?”) and individual IBS specific symptom assessment of bloating, pain and flatulence (6 point Likert scale ranging from 0 (none) to 5 (very severe)).

Participants completed the Birmingham IBS Symptom Score [31], the IBS SSS [27] and an IBS specific quality of life (QoL) tool [32] at baseline and at weeks 4, 8 and 12. The Birmingham Symptom Score [31] uses 11 questions to assess pain, diarrhoea and constipation, scores range from 0–100 for each dimension with higher scores indicating greater well-being. The IBS –SSS [27] utilises five questions each generating a maximum score of 100 and a total possible score of 500 with higher scores indicating a greater symptom burden. The IBS specific QoL [32] tool evaluates dimensions; dysphoria, interference with activity, body image, health worry, food avoidance, social reaction, sexual and relationship using 34 five point Likert statements. Items are summed to create a total score with higher scores indicating better QoL.

The primary outcome was pre-specified as SGA of symptom relief at week 4 in line with the recent trial evidence on which the study was powered [33] and after discussion with the product manufacturer and sponsor. Comparison of all other outcomes at weeks 4, 8 and 12 between groups formed specified secondary outcome measures.

The sample size calculation was based on a previous study [33] demonstrating a 20% between group difference (62% active vs. 42% control) for SGA. Therefore, 107 patients with IBS-C were required in each group to demonstrate a 20% difference between groups in proportions reporting adequate IBS symptoms relief at week 4, with a significance of 5% and 80% power. This was inflated to 240 patients to be divided in two groups of 120 to allow for drop-outs. No stopping guidelines were in place due to the short trial duration and assumed safety of a commercially available product. Data analyses for all outcomes were performed on an intention to treat basis (ITT) and further analyses were undertaken to compare differences in those not deviating from protocol (per protocol). Analyses were performed on available data for each parameter. ANCOVA with primary care centre as a random effect, baseline as a covariate and product as a fixed effect was used to compare the intervention and control groups with respect to all outcomes. Ordinal logistic regression with centre and product as fixed effects was used to explore change within each arm of the study. Analyses were conducted using MINITAB version 14, S Plus version 8 and ML wiN version 2.1 statistical software.

The study was funded by Danone Ltd. Contracts including a right to publication of all findings were approved and signed by both parties prior to commencement of the research. The funders contributed to discussion about study design and selection of outcome measures. All subsequent data review was blinded and the analysis was undertaken independently of the funders. The paper was reviewed by the sponsor and an expert nominated by them with some requested changes included in the final version.

Among the 179 randomized patients, 109 had available data for SGA at week 4 (60 in active group; 49 in control group). There were no between group differences in SGA noted at week 4 with 56.7% of the active group and 53.1% of the control group reporting adequate symptom relief (p = 0.71) (Table 3).

Large numbers of withdrawals accrued, with dropout rates of 39% (70/179) at week 4, 48% (86/179) at week 8 and 55% (98/179) at week 12, despite efforts to follow-up all participants. Withdrawal reasons were recorded where provided by participants. Across both groups drop-out was attributable to a range of factors including dislike of product taste, changes in personal circumstances and reported adverse events such as nausea.

At week 8 greater benefit in terms of adequate symptom relief was reported in the control group (68.3% reporting relief versus 46.2%, p = 0.03) and this was sustained at week 12 (75.8% versus 45.8%, p = 0.004) (Table 3).

During the initial 4 weeks of the trial statistically significant (at the p < 0.05 level) improvement from baseline was reported in both groups for the majority of outcomes, including improvements in symptoms scores, bloating, flatulence, ease of bowel movement and quality of life (Table 4). Although both groups reported improvement there was no significant between group difference when considering changes in either of the symptom measures used, nor for any individual symptoms or QoL outcomes (Table 4). The improvement from baseline scores demonstrated at week 4 in both groups remained significant for most outcomes at week 8. The absence of a between group difference, however remained constant with both groups demonstrating similar improvements (Table 5).

By week 12, however, change in scores differed significantly between groups for the following outcomes; IBS symptom severity score, IBS symptoms of constipation and associated total score as measured by the Birmingham score, IBS specific symptoms of pain, bloating and associated total score. For all statistically significant outcomes those receiving the control product demonstrated greater improvement than those receiving the active product (Table 5).

Per protocol analyses undertaken at each of the time points (4, 8 and 12 weeks) did not significantly alter the findings reported in the ITT analyses.

The sustained and large improvement observed in both groups suggests there may be benefit from regular consumption of a dairy product but does not suggest any additional benefit of the addition of a probiotic to such products. If such a benefit is accrued the mechanism through which this works is unclear and product effect, regularisation of eating habits and increased fluid intake are all possible explanations. The fact that benefits were also accrued by control participants supports these potential mechanisms of effect. In the absence of a non-intervention group this effect cannot be confirmed but the substantial effect size is worthy of further consideration and exploration of sub-groups in which greatest benefit is accrued would be advisable. In this study there was the suggestion that drop-out may be related to baseline symptom severity with those remaining in the study having greater baseline symptom severity scores than those withdrawing (IBS SSS 264.15 versus 238.09, p = 0.04). This could suggest greater benefit was accrued by those with more severe symptoms encouraging continued participation or may reflect and greater motivation to pursue therapy in this group.

This primary care based RCT is one of the largest trials of probiotics in IBS to date and one of the only functional food trials to be conducted outside small hospital populations. The main study limitation was not reaching the target recruitment of 240 patients between the two arms, with only 179 recruited of whom 109 provided data for the primary outcome. The failure to meet the trial target recruitment was due to a contractually fixed end date for the trial and the 109 participants providing 4 week data reduced study power to 47%. However, at study close the trial showed futility of continuance given the observed lack of treatment benefit. A post-hoc re-estimate of sample size, based on observed difference in the study primary outcome (proportions 0.57 versus 0.53), indicated 2477 patients would have been required in each arm to demonstrate a statistically significant treatment effect.

The observed imbalance in the randomization numbers was due to the study requirement to randomize eligible patients prior to the baseline visit (to enable product manufacture) and there was differential clinic attendance by chance. Since the study power was reduced through a high level of drop-outs and differential drop-out, it therefore remains a possibility that an effect of the active product was missed. The high drop-out rate highlights the difficulty of conducting functional food trials but does call into question the acceptability of such interventions to patients over even short periods.

This study focused only on individuals who had a constipation element to their IBS since earlier data suggested this might be the population who might obtain most benefit from consumption of a probiotic: prior trials [34‐36] have suggested that daily consumption of fermented milk products containing Bifidobacterium animalis DN-173 010 improves transit and alleviates bloating in individuals with low stool frequency. However the heterogeneous nature of the cohort recruited (some with a constipation predominant form of IBS and others with a mixed profile) may further mask an effect which would be demonstrated in a more homogeneous IBS constipation predominant patient group. Despite screening to ensure all participants were symptomatic at baseline and with a constipation feature to their IBS, the mean number of stools passed daily at baseline was greater than one and the mean stool consistency was 3 on the Bristol Stool Form Scale which equates to only a slightly drier than average stool suggesting this cohort did not comprise those with more significant forms of constipation. Recruitment of participants with a more extreme manifestation of constipation may have more closely replicated the small hospital populations in previous trials where modest benefits of probiotics were observed. However, given the dominance of small positive trials in the literature, it is also possible that the literature in this area is subject to publication bias. This highlights the importance of establishing the efficacy of treatments across the wide range of IBS phenotypes if they are to be made widely available. If functional foods are to be advocated for a common disorder like IBS it is important to have clear indications as to the symptoms most likely to accrue benefit. This study importantly illustrates that data from the presumed more persistent and extreme IBS phenotypes seen in hospitals cannot routinely be extrapolated to the general population.

The interest in the use of functional foods is demonstrated by the large numbers of reviews conducted in recent years [20, 21, 37‐41]. Despite this major body of review evidence, the trial evidence upon which it is based remains limited. This carefully conducted double blind randomized 12 week study assessing a wider range of outcomes in a general population of IBS with a constipation element, did not reproduce these earlier hospital based findings. Explanations for this may be attributable to differences in participant groups, outcomes and follow-up. However this study best reflects current practice for the majority of patients with IBS who might consider probiotic milk products to try to reduce symptoms.

Clinicians advising patients with IBS managed in the community featuring a constipation element may wish to suggest the inclusion of a fermented dairy product, given that significant improvements were reported for most outcomes in all trial participants. The requirement of such products to contain a probiotic is not supported by this study.

Further research is required to consider the mechanism via which improvement in symptoms may be effected in trials of this nature through consideration of dietary habits and fluid intake. Such work will ensure the most accurate dietary advice can be provided to patients.