|Home | About | Journals | Submit | Contact Us | Français|
It has been suggested that probiotics may improve gastrointestinal discomfort. Not all probiotics exhibit the same effects and consequently meta-analyses on probiotics should be confined to well-defined strains or strain combinations. The aim of this study was to evaluate the effectiveness of a probiotic fermented milk (PFM) that includes Bifidobacterium lactis (B. lactis) CNCM I-2494 and lactic acid bacteria on gastrointestinal discomfort in the general adult population.
Double-blind randomized controlled trials in the general adult population comparing PFM with a control dairy product for at least 4 weeks were searched from multiple literature databases (up to February 2015). Meta-analyses using random-effects models, with individual participant data were undertaken to calculate an odds ratio (OR) or standard mean difference (SMD), with a 95% confidence interval (CI).
The search strategy identified 12,439 documents. Overall, three trials with a total of 598 adults (female = 96.5%) met the inclusion criteria. Consumption of the PFM product was associated with a significant improvement in overall gastrointestinal discomfort compared with the control product (OR = 1.48; 95% CI 1.07–2.05), with a number needed to treat (NNT) of 10.24 (95% CI 5.64–55.93). PFM was also superior to the control in reducing digestive symptoms, as measured using a composite score (SMD = −0.21; 95% CI −0.37 to −0.05). Sensitivity analyses produced similar results, and the heterogeneity between studies was minimal.
This meta-analysis shows that the consumption of PFM with B. lactis CNCM I-2494 and lactic acid bacteria is associated with a modest but consistent and significant improvement of outcomes related to gastrointestinal discomfort in healthy adults.
Digestive function plays a key role in maintaining or improving health status. In their guidance on health claims related to gut and immune function, the European Food Safety Agency (EFSA) considers that reduced gastrointestinal discomfort is indicative of improved gastrointestinal function and is related to a beneficial physiological effect [European Food Safety Authority, 2011]. Gastrointestinal discomfort encompasses different digestive symptoms such as abdominal pain or discomfort, bloating, borborygmi and flatulence all of which are widely observed in the general population [van Kerkhoven et al. 2008; Tielemans et al. 2013]. The type, frequency and intensity of these symptoms varies between individuals [Heaton et al. 1992; Drossman et al. 1993; Longstreth et al. 2006] and may impact on daily life [Tielemans et al. 2013; Frexinos et al. 1998].
Probiotics have been defined as live microorganisms, which, when administered in adequate amounts, confer a health benefit on the host [Food and Agriculture Organization of the United Nations and World Health Organization, 2001]. Probiotics are available in a variety of forms such as powders, capsules, foods and infant formulae [Sanders et al. 2011]. The administration of probiotics has been recommended as a way of improving gastrointestinal symptoms by modifying the gut microbiota [Simren et al. 2013]. Indeed, the effect of probiotics in patients with constipation or irritable bowel syndrome (IBS) have been extensively investigated [Ford et al. 2014; Moayyedi et al. 2010; Clarke et al. 2012; Dimidi et al. 2014]. The ingestion of probiotics, such as lactobacilli and bifidobacteria, can modulate the composition and metabolism of the gut microbiome [Derrien et al. 2015]. Probiotics have been shown to have a whole range of activities that can potentially lead to improvement of gut function. These include a direct interaction with the gut luminal microbiota, metabolic effects resulting from enzymatic activity; an effect on barrier function and crosstalk with the central nervous system [Rijkers et al. 2010; Theodorou et al 2014]. It has been clearly shown that the beneficial effects of a particular probiotic are strain-specific and these beneficial effects cannot be extrapolated to another probiotic, even if the probiotic belongs to the same species [Marteau, 2010; Rowland et al. 2010; Shanahan, 2011]. Consequently, it is inappropriate to undertake a meta-analysis of the effect of a range of different probiotics on a particular condition, as the results may be misleading [Szajewska, 2010]. While it is acceptable to report the effects of a range of different probiotics, a meta-analysis should just be confined to a particular product or strain.
Some strains of bifidobacteria have been reported to improve symptoms in IBS [Ford et al. 2014; Moayyedi et al. 2010; Hungin et al. 2013] and to accelerate intestinal transit time [Miller and Ouwehand, 2013]. Probiotic fermented milk (PFM), with a specific Bifidobacterium strain (Bifidobacterium lactis CNCM I-2494) and four strains of lactic acid bacteria, has shown beneficial effects on gut function in several randomized controlled studies [Marteau et al. 2002, 2013; Guyonnet et al. 2007, 2009a; Agrawal et al. 2009]. Furthermore, this probiotic food has been shown to improve gastrointestinal well-being and symptoms in individuals from the general population without functional gastrointestinal disorders [Guyonnet et al. 2009a, 2009b; Marteau et al. 2013] in contrast with another bifidobacteria-containing product [Eskesen et al. 2015]. It has also been shown that the consumption of this PFM can influence the gut–brain axis by modulating the activity of the brain [Tillisch et al. 2013] as well as the composition of the gut microbiome in healthy people [McNulty et al. 2011].
The aim of this systematic review with meta-analysis was to evaluate the effectiveness of this well-defined PFM, (containing a specific combination of B. lactis CNCM I-2494 and four strains of lactic acid bacteria), on gastrointestinal discomfort, as defined at the beginning of this section, in the general population.
This systematic review was carried out according to the Cochrane Handbook for Systematic Reviews (http://handbook.cochrane.org/) [Higgins and Green, 2011]. The review was guided by a protocol, developed in advance of the review proper, which was registered on the PROSPERO database (http://www.crd.york.ac.uk/PROSPERO) with the registration number CRD42015016660.
A literature search was carried out in MEDLINE and MEDLINE In-Process, PubMed, EMBASE, the Cochrane Database of Systematic Reviews (CDSR), the Cochrane Central Register of Controlled Trials (CENTRAL), the Database of Abstracts of Reviews of Effects (DARE), Health Technology Assessment Database (HTA), Science Citation Index (SCI), Conference Proceedings Citation Index-Science (CPCI-S), Biosis, CAB abstracts, Proquest Dissertation and Theses: UK & Ireland, Global Health Library and WPRIM, up to February 2015.
Grey literature was identified via OAISTER, OpenGrey and NTIS. Unpublished studies were identified using ClinicalTrials.gov and the International Clinical Trials Registry Platform (ICTRP), and by searching conference proceedings for 3 years up to February 2015 (details of the searches are provided in online supplementary Appendix A).
The strategy searched for two concepts: Bifidobacterium and gastrointestinal discomfort/comfort. The detailed search terms and the strategy used for each database are presented in online supplementary Appendix A. The searches were not limited by date or language. Additionally, the reference lists of relevant reviews, trials and studies were searched to identify further eligible trials.
To be eligible for inclusion in the review, a study had to be a prospective, double-blind, randomized controlled trial (RCT) that compared oral consumption of any dose of a PFM containing a specific mix of B. lactis CNCM-I2494 and four lactic acid bacteria (Lactobacillus bulgaricus strains CNCM I-1632 and CNCM I-1519, Streptococcus thermophilus strain CNCM I-1630, and Lactococcus lactis ssp. lactis strain CNCM I-1631), with a control product for at least 4 weeks. Participants had to be aged 18 and over and recruited from the general population, with gastrointestinal discomfort at entry. Eligible outcomes were the effectiveness of PFM on gastrointestinal discomfort or comfort measured by a global assessment (i.e. overall assessment of gastrointestinal discomfort/comfort or gastrointestinal well-being), using a single integrated question with a dichotomous outcome (classifying each participant as a responder or nonresponder and enabling a rate of responders to be calculated), or a composite score comprising at least two of the following individual digestive symptoms: abdominal pain/discomfort; abdominal pain; bloating; borborygmi (rumbling); or flatulence.
Eligible control products included nonfermented dairy products without probiotics, but with the same taste and colour appearance (to allow double-blinding); a heat-treated active product (i.e. the bacteria were not alive); or another fermented milk product without ‘active’ probiotics (e.g. a standard yoghurt without bifidobacteria). Studies published as abstracts or conference presentations were eligible for the review as long as they provided adequate data.
Studies of true healthy participants without gastrointestinal symptoms were excluded, as were studies of participants with IBS, functional bowel disorders including constipation or diarrhoea, or any other significant intestinal disorder such as Crohn’s disease, ulcerative colitis, pouchitis, coeliac disease and colon cancer. Studies of the specific mix of five bacterial strains combined with other functional ingredients (such as prebiotics or vitamins) were not eligible. Studies in which control products included the PFM with additional ingredients, not present in the intervention product, were excluded.
Search results were assessed for relevance to the review question. First, one reviewer removed obviously irrelevant database records, including those records which reported animal studies, addressed products other than PFM, or involved children. Second, two reviewers independently assessed the remaining database records in detail, using information in the title and abstract. Third, the full texts of possibly relevant studies were obtained and were assessed for relevance by one reviewer and the decision was checked by a second reviewer. Disagreements were resolved through discussion and, where necessary, by consulting a third reviewer.
Information on study characteristics, participant characteristics, methods, and results were extracted from each study. If no composite outcome score was reported, but at least two individual eligible digestive symptoms were reported in the publication or study report, we calculated the composite score. Study authors were contacted for additional information where required.
One researcher extracted the data from the full papers, and a second researcher checked the extraction. Any disagreements were resolved through discussion or by consulting a third reviewer.
The quality of the RCTs was assessed using criteria based on the Cochrane risk of bias tool [Higgins and Green, 2011]. Quality assessment was conducted by one reviewer and checked by a second reviewer. Any disagreements were resolved through discussion or by consulting a third reviewer.
Based on the data reported on global gastrointestinal discomfort or comfort, a single dichotomous ‘responder status’ was defined for each participant. Participants who improved on at least 50% of occasions were defined as responders (based on recommendations for global assessment of symptom relief in IBS trials) [Irvine et al. 2006]; a secondary outcome defined responders as those who improved on at least 50% of occasions and did not deteriorate on any occasion.
Based on the individual digestive symptom scores, a composite symptom score was defined for each participant, as long as at least two symptom scores had been recorded. This composite score was measured weekly over the course of the study.
The details of the definition of ‘improvement’, ‘deterioration’ and ‘composite score’ depended on the study and are provided in the results section below.
Analyses of the outcomes of interest were undertaken using individual participant data. The analysis was based on the intention-to-treat population, defined as all randomized participants included in a study.
The primary analysis for responder status was confined to completers. Two sensitivity analyses were conducted to include imputed values for the missing data. The first analysis assumed participants receiving the active product were responders and those receiving the placebo product were nonresponders. The second analysis assumed participants receiving the active product were nonresponders and those receiving the placebo product were responders. For the primary analysis for the composite score, the repeated measures analysis (see below) allowed the inclusion of all randomized participants.
Heterogeneity was assessed formally using the I-squared (I2) statistic, between-study variance (tau-squared) and the p value of the heterogeneity statistic Q. The I2 statistic measures ‘the percentage of total variation across studies that is due to heterogeneity rather than chance’; an I2 of 0–25% represents no heterogeneity, 25–50% represents moderate heterogeneity, 50–75% represents substantial heterogeneity, and 75–100% represents considerable heterogeneity [Higgins and Green, 2011].
Two steps were used to conduct the meta-analysis. The first step was to conduct an analysis of the effect of the PFM and control product within each study separately by calculating an odds ratio (OR), for overall gastrointestinal discomfort/well-being, or a standardized mean difference (SMD) for the composite score of digestive symptoms (using repeated measures analysis of covariance), along with a 95% confidence interval (CI). The second step was to combine these results in a conventional meta-analysis involving a frequentist meta-analysis based on the log OR or SMD (depending on the outcome evaluated in the analysis), and its standard error. Both a fixed-effect model and a random-effects model were fitted in each meta-analysis. The results were presented as forest plots showing the pooled effect sizes and 95% CI. For gastrointestinal discomfort/well-being, the number needed to treat (NNT) was also reported.
Sensitivity analyses were conducted to assess the impact of the heterogeneity across studies, and the impact of baseline covariates. For the analysis of overall gastrointestinal discomfort/well-being, sensitivity analyses were conducted using a stricter definition of a responder (excluding any deterioration, as defined above), and by assessing the effects of imputing values for missing participants. For the composite score, exploratory analyses were conducted to assess the impact of individual symptoms included in the composite score. In addition, exploratory analyses were conducted to investigate the relationship between the symptom scores and overall assessment of gastrointestinal well-being. Summary statistics were presented for the subgroup of participants with predominant bloating at baseline, as bloating has been shown to improve with this product in previous studies on IBS [Agrawal et al. 2009; Guyonnet et al. 2009b]. All analyses were conducted separately for the study by Donazzolo and colleagues [Donazzolo et al. 2007].
All treatment comparisons were two-sided, at the 5% level of statistical significance.
A total of 12,439 records were identified by the database searches and 5 records were identified from other sources. Following deduplication, 6588 records were assessed for relevance based on titles and abstracts. Of these, 37 documents were obtained and relevance was assessed from the full text. In total, three trials (six documents) were included in the review (Figure 1) [Guyonnet et al. 2009a; Marteau et al. 2013; Donazzolo et al. 2007; AtlanStat and Guyonnet, 2012; de la Motte et al. 2008; Tanguy et al. 2014]. A total of two systematic reviews [Miller and Ouwehand 2013; Didari et al. 2014] were also obtained and reviewed to identify further relevant studies, but none were found.
A total of two of the included studies were conducted in clinical centres in France [Donazzolo et al. 2007; Marteau et al. 2013], and one was conducted at a clinical centre in Germany [Guyonnet et al. 2009a]. All were randomized, controlled, parallel-group trials, with 4-week double-blind intervention periods. Guyonnet and colleagues randomized 202 participants, Marteau and colleagues randomized 336 participants and Donazzolo and colleagues randomized 60 participants [Guyonnet et al. 2009a; Marteau et al. 2013; Donazzolo et al. 2007].
The studies by Guyonnet and colleagues and Marteau and colleagues compared the effects of PFM with a nonfermented dairy product, on gastrointestinal well-being and digestive symptoms among adult women (18–60 years of age) without gastrointestinal disorders (including functional bowel disorders such as IBS) [Guyonnet et al. 2009a; Marteau et al. 2013]. Donazzolo and colleagues compared the effects of PFM with a nonfermented dairy product on stool frequency in men and women (18–65 years of age) with low stool frequency (<4 stools/week), but not functional constipation or IBS [Donazzolo et al. 2007].
In all trials, the culture count of the active intervention was B. lactis CNCM I-2494 (1.25 × 1010 cfu/pot), and S. thermophilus and L. bulgaricus (1.2 × 109 cfu/pot). In all trials, the participants consumed two 125 g cups/pots daily, one at breakfast and one at dinner.
Baseline digestive symptoms of the participants are presented in Table 1. The trial by Donazzolo and colleagues used a slightly different scoring system than Guyonnet and colleagues and Marteau and colleagues; Donazzolo and colleagues evaluated intensity of symptoms (on a scale from 0–5), whereas Guyonnet and colleagues and Marteau and colleagues reported frequency (on a scale from 0–4) [Guyonnet et al. 2009a; Marteau et al. 2013; Donazzolo et al. 2007]. Generally, the scores for individual and global digestive symptoms were mild to moderate in the Donazzolo and colleagues study [Donazzolo et al. 2007]. In the studies by Guyonnet and colleagues and Marteau and colleagues, abdominal pain was reported to occur once a week at baseline, and mostly twice a week for borborygmi, bloating and flatulence [Guyonnet et al. 2009a; Marteau et al. 2013]. The composite score of frequency of digestive symptoms, which ranged from 0–16, averaged 7.1 for the Guyonnet and colleagues trial and 6.9 for the Marteau and colleagues trial [Guyonnet et al. 2009a; Marteau et al. 2013]. For the Donazzolo and colleagues trial, the composite score of intensity of symptoms ranged from 0–15 and the average was 6.6 [Donazzolo et al. 2007]. For comparability across all three trials, these scores can be scaled by dividing by the maximum value. For all three trials the average scaled score at baseline was 0.4.
Based on information presented in the clinical study reports, full details on outcomes of interest were presented, and data were available for all outcomes planned for analyses.
All three included studies were considered to have a low risk of bias. A detailed table of the quality assessment is presented in online supplementary Table 1. As the quality assessments were the same for both outcomes, only one table is presented.
Guyonnet and colleagues and Marteau and colleagues both undertook the same overall assessment of gastrointestinal well-being involving weekly self-evaluation during the 4-week period of product consumption [Guyonnet et al. 2009a; Marteau et al. 2013]. Participants indicated whether their gastrointestinal well-being had remained the same, improved, or worsened (three-point Likert scale). In a second step, participants with improving or worsening symptoms rated the degree of change on a seven-point scale. This led to a 15-point Likert scale (−7, 0, +7). The percentage of participants who worsened, had no change or improved (percentage by class) and the percentage of participants who improved on at least 2 of the 4 weeks (% responders) were outcomes reported by both studies. Donazzolo and colleagues measured an overall assessment of ‘global digestive symptoms’ where each week, participants indicated the intensity of their symptoms using a six-point Likert scale, ranging from none (0) to very high (+5) [Donazzolo et al. 2007].
In the Guyonnet and colleagues and Marteau and colleagues studies, a participant was considered to be a responder if they were ‘improved’ on at least 2 of the 4 weeks [Guyonnet et al. 2009a; Marteau et al. 2013]. In the Donazzolo and colleagues trial, a participant was considered to be a responder if they had at least a 30% decrease in the intensity (compared with baseline) of global symptoms on at least 2 of the 4 weeks [Donazzolo et al. 2007]. A second definition of response, defined participants who recorded a deterioration of symptoms at any time as nonresponders, even if they reported an improvement during at least 2 of the 4 weeks.
Guyonnet and colleagues and Marteau and colleagues both reported a composite score (ranging from 0 to 16) of the frequency of four symptoms (abdominal pain/discomfort, bloating, flatulence/passage of gas and borborygmi/rumbling stomach) which were each self-evaluated using a five-point Likert scale that ranged from 0 (never) to 4 (every day of the week) [Guyonnet et al. 2009a; Marteau et al. 2013]. Donazzolo and colleagues measured a composite score (ranging from 0–15) of intensity of each of three symptoms (abdominal bloating, abdominal pain and flatulence) [Donazzolo et al. 2007]. The intensity of these symptoms was assessed using a six-point Likert scale, ranging from none (0) to very high (+5).
There were very few participants for whom responder status could not be calculated in each sample: 3 (1.5%) participants [Guyonnet et al. 2009a]; 3 (0.9%) participants [Marteau et al. 2013]; 1 (1.7%) participants [Donazzolo et al. 2007].
When individual participant data from all three studies were combined in a meta-analysis, there was a significant effect in favour of PFM, with no significant heterogeneity between the studies (I2 statistic = 0%; p = 0.54). For all three studies, the estimated OR for response was >1, supporting benefit of the product. Only the result for Guyonnet and colleagues was statistically significant on its own (OR = 1.90; 95% CI 1.08–3.35; n = 199) [Guyonnet et al. 2009a]. The overall estimate for the OR was 1.48 (95% CI 1.07–2.05; n = 591) for both the fixed-effect and random-effects models. The corresponding NNT was 10.24 (95% CI 5.64–55.93) (Figure 2).
Results for the sensitivity analyses were all similar, with low or zero heterogeneity in all cases except for the analysis of the second definition of response, which excluded any worsening in global symptoms. Here, I2 = 38.8% representing moderate heterogeneity, and the p value for heterogeneity was 0.20; the OR estimate (random-effects model) was 1.62 (95% CI 1.01–2.58; n = 591), with NNT = 8.90 (95% CI 4.51–363.10). The intervention effect was significant in all cases. The estimate of NNT varied across the analyses, ranging from 7.26–11.89, with wide CIs.
For the composite score, all participants had a baseline value and at least one later value. It was assumed that scores were missing completely at random and that therefore an intention-to-treat analysis could be achieved without imputation of the missing scores.
When individual participant data from all three studies were combined in a meta-analysis, there was a significant effect in favour of PFM with no significant heterogeneity between the studies (I2 statistic = 0%; p = 0.82). For all three studies the estimated SMD for the composite score was negative, supporting benefit of the product. Only the result for Guyonnet and colleagues was statistically significant on its own; however the product effect was significant overall: the estimate for the SMD was −0.21, with a 95% CI of −0.37 to −0.05 (n = 598) for both the fixed-effect and random-effects models (Figure 3) [Guyonnet et al. 2009a].
Table 3 shows the results of logistic regression models relating the responder status to all the individual symptom scores together, as well as age, sex and body mass index (BMI); including all time points.
The results show that all the symptoms are related to responder status, and that the lower the symptom score, the higher the probability of response (because the coefficients are negative). There was, however, one exception: in the Donazzolo and colleagues study, flatulence was weakly positively related to response.
Age was strongly positively correlated to response in the Donazzolo and colleagues’ population, but weakly negatively correlated with response in the other two studies. Sex and BMI showed no relationship with response.
When the data from all three studies are combined, there are many more participants who responded than deteriorated each week, and the percentage of participants reporting a deterioration of their gastrointestinal well-being is generally lower in the PFM group versus control group (Table 4).
The analysis of the change from baseline for composite symptom score in the subgroup of participants with bloating as the predominant baseline symptom, showed that the evidence in favour of an improvement is stronger in this subgroup compared with that in the overall population. This trend was clearer for the combined Guyonnet and colleagues and Marteau and colleagues data (Table 5).
This meta-analysis, using individual participant data for 598 participants to evaluate the effectiveness of a PFM containing a specific combination of bacteria strains including B. lactis CNCM I-2494, shows a statistically significant beneficial effect from consuming the PFM compared with control product for the two gastrointestinal discomfort outcomes evaluated: namely overall assessment as dichotomous measure and a composite score of digestive symptoms. A unique feature and advantage of this study was that all the raw data on every individual participant was available for inclusion in the analysis.
Probiotics have a modulatory effect on the gut microbiota and therefore should not be expected to have the rapid and dramatic effects that are observed with pharmaceutical agents [Sanders et al. 2011]. The magnitude of the effect of the PFM in our study is more appropriately compared with that obtained by other functional foods or probiotics in IBS treatment and also to agents targeting gut microbiota in IBS, such as rifaximin. A recent systematic review on the effect of probiotics on IBS and chronic constipation showed a significant reduction of global IBS or abdominal pain scores for probiotics [Ford et al. 2014]. This reduction was in the same range as the reduction of the composite score observed with PFM in this analysis: (SMD = −0.21; 95% CI −0.37 to −0.05) compared with all studies in Ford and colleagues (n = 24 for 2001 patients; SMD = −0.25; 95% CI −0.36 to −0.14) or compared with studies in Ford and colleagues of combinations of probiotics (n = 15 for 1038 patients; SMD = −0.24; 95% CI −0.37 to −0.12). Although the approach of combining different probiotic strains to assess their effect may be misleading, we refer to the data from Ford and colleagues to demonstrate that the effect of the PFM demonstrated in this study is of a similar magnitude to those of others in the probiotic area [Ford et al. 2014]. A recent meta-analysis of the antibiotic rifaximin for global IBS symptoms by Menees and colleagues reported an OR of 1.57 (95% CI 1.22–2.01) based on five RCTs (n = 1803) [Menees et al. 2012]. Rifaximin is a useful comparator, as it involves just one agent in contrast with most other meta-analyses of probiotics where different preparations are grouped together. The value obtained with rifaximin is comparable with the rate of responders for overall assessment in our analysis which found an OR of 1.48 (95% CI 1.07–2.05).
When considering the NNT of 10.24 obtained in this analysis, it is close to the value of 10.2 obtained in IBS for rifaximin [Menees et al. 2012]. The reported NNT for probiotics in IBS is 7–8 [Ford et al. 2014] and approximately 8 for antibiotic-associated diarrhoea (meta-analysis of 34 studies, N = 4138) [Videlock and Cremonini, 2012]. In making these comparisons, some caution is required regarding comparison of these NNTs due to the often large CIs and the variation in baseline risk between trials within a meta-analysis. With regard to responder rate, the pooled data from the three studies used in this meta-analysis gives a difference of 9.7% between active and control product [active: 51.9% (154/297), control: 42.2% (124/294)]. This compares favourably with rifaximin in IBS where the efficacy of rifaximin over placebo was 9.8% for global symptom improvement and 9.9% for bloating. Not surprisingly, higher therapeutic gains have been reported for the recently developed specific drugs for IBS such as 16.2% for linaclotide [Lacy et al. 2014]. Higher therapeutic gain is also observed when changing the diet in IBS, using a low FODMAP diet. The observed response is usually >20% [Staudacher et al. 2011, 2012] which may be explained by the minimization of a major stimulus to symptom induction (luminal distension) as opposed to changing the microbiota per se, such as with the introduction of a single food as in the PFM studies.
The strengths of this study include that it was conducted using rigorous systematic review methods and included meta-analyses using individual participant data. In addition, the three included trials were randomized, controlled, double-blind, had a low risk of bias, with low or no dropouts and included relatively large participant numbers. The included studies were carried out over a 4-week period, which provides enough time for a clinical effect to be detected [Irvine et al. 2006]. A recent editorial stated that the inclusion of small-sized trials, with poor or variable quality of design or conduct, and susceptibility to high risk of bias minimizes the strength and applicability of results from meta-analyses in the probiotic and prebiotic fields [Whelan, 2014]. This systematic review addresses these perceived weaknesses by including relatively large participant numbers with a low risk of bias and only included trials of one probiotic food with a specific combination of bacteria strains, including B. lactis CNCM I-2494. The three studies had the same duration of product consumption (4 weeks), the same daily amount of product (2 × 125 g), the same amount of probiotics, and the same control product.
A limitation of this study is that the majority of participants were women, reflective of the fact that the majority of the yoghurt-eating population are women [Crichton et al. 2010]. In addition, in any review of evidence there is a potential for publication bias. We could not employ a funnel plot to identify patterns that may be indicative of publication bias, as it is generally recommended that such an analysis should include a minimum of 10 studies [Higgins and Green, 2011].
Unlike many drug-based treatments for digestive discomfort, there is no requirement for medical advice prior to administration of a probiotic. It is available over-the-counter and may offer benefit in people with mild digestive discomfort who seldom seek medical advice and are unlikely to be suffering from any serious disease. Consequently, the addition of a probiotic food to the daily regimen of healthy adults experiencing mild gastrointestinal discomfort is readily achieved and, based on the findings reported here, is likely to actually offer them significant benefit. The observation in this review that more severe symptoms responded less well, indicates that these products are probably best suited to individuals with mild symptoms. Importantly, this study also demonstrates that this PFM is not likely to cause an exacerbation of symptoms. The safety profile is such that the probiotic approach is a reasonable one, without the necessity for medical advice. Participants with gastrointestinal discomfort often opt to follow restrictive diets (e.g. a gluten-free diet or low FODMAP diet) without medical advice in an attempt to minimize symptoms. However, in contrast with probiotics, such diets have potential effects on nutritional status and may also unfavourably change gut microbiota [Staudacher et al. 2012; Cenit et al. 2015; Halmos et al. 2015]. Of course, as with any ‘over-the-counter’ remedy, individuals with persisting symptoms not responding to probiotics should be encouraged to seek medical advice.
In conclusion, this systematic review with meta-analyses shows that the consumption of this PFM with B. lactis CNCM I-2494 and lactic acid bacteria is associated with a consistent and significant improvement of outcomes related to gastrointestinal discomfort in a large cohort of people who do not carry a medical diagnosis for those symptoms. Most benefit is achieved in those with mild symptoms and future studies should investigate the profile of those individuals most likely to respond.
Funding: This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Conflict of interest statement: Danone commissioned York Health Economics Consortium (YHEC) Ltd. (an independent research organization) and Quantics Consulting (a statistical consultancy) to carry out this systematic review and meta-analysis. Dr Jacqui Eales (YHEC) served as a researcher on this review. Dr Eales has previously received funding from Danone Institute International (DII) to deliver a conference presentation of the outcome of another systematic review commissioned by DII and undertaken by YHEC. Dr Peter Gibson has served as a consultant or advisory board member for Abbvie, Ferring, Janssen, Merck, Nestle Health Science, Danone, Allergan and Takeda. His institution has received speaking honoraria from Abbvie, Janssen, Ferring, Takeda, Fresenius Kabi and Pfizer. He has received research grants for investigator-driven studies from Abbvie, Janssen, Falk Pharma, Danone and A2 Milk Company. His department financially benefits from the sales of a digital application and booklets on the low FODMAP diet. He has published an educational/recipe book on diet.
Dr Peter Whorwell has acted as a consultant for, or received research grant support from, the following companies in the last 5 years: Almirall Pharma, Boehringer-Ingelheim, Chr. Hansen, Danone Research, Ironwood Pharmaceuticals, Salix, Shire UK, Sucampo Pharmaceuticals and Allergan. Dr John Kellow is a Board member of the Rome Foundation and is on the Advisory Board of Danone Australia. Dr Ann Yellowlees is a Director of Quantics, and declares no additional personal interests. Dr Richard Perry (Quantics) declares no additional personal interests. Mary Edwards, Dr Sarah King, Hannah Wood and Julie Glanville are employed by YHEC, and declare no additional personal interests.
Jacqui Eales, York Health Economics Consortium, University of York, Heslington, York, UK.
Peter Gibson, Monash University, Alfred Hospital, Melbourne, Australia.
Peter Whorwell, Wythenshawe Hospital, Manchester, UK.
John Kellow, University of Sydney, Royal North Shore Hospital, Sydney, Australia.
Ann Yellowlees, Quantics Consulting Ltd, Edinburgh, Scotland.
Richard H. J. Perry, Quantics Consulting Ltd, Edinburgh, Scotland.
Mary Edwards, York Health Economics Consortium, University of York, Heslington, York, UK.
Sarah King, York Health Economics Consortium, University of York, Heslington, York, UK.
Hannah Wood, York Health Economics Consortium, University of York, Heslington, York, UK.
Julie Glanville, York Health Economics Consortium, Enterprise House, Innovation Way, University of York, Heslington, York YO10 5NQ, UK.