We identified 497 citations, 178 of which were duplicate records and were therefore excluded. Two authors screened 319 abstracts and identified potentially relevant articles. Twenty two retrieved articles were independently reviewed for inclusion and exclusion criteria. The reviewers were not masked to any aspect of the studies (e.g. journal type, author's names etc). On further analysis of retrieved articles, 8 trials were excluded for being non-controlled studies. Therefore, a total of 14 articles met the inclusion criteria [17–30] (figure. 1). As a number of studies included symptoms of flatulence this was included in the analysis. For example, in Gade's study [18] published in 1989, the IBS definition of constipation and/or diarrhoea, abdominal pain, meteorism, borborygmus and flatulence was used.
Trials were from USA (three), Poland (two), Ireland (two), UK (one) France (one), Israel (one), Finland (one), Italy (one), Sweden (one) and Denmark (one). One trial included women only [30]; the remainder included female and male participants with the majority being women (table 1). Two studies included children [17, 19], though Bausserman's study [17] included an age range of 6–20 years. Trials varied in relation to the type, dose and strengths of probiotic(s) used. Number of probiotics varied from one to multiple, e.g. VSL#3 (contains 8 different bacterial strains). Three studies included more than one arm in the intervention group[27, 29, 30] (table 1).
Table 1 Study characteristics
In six studies the length of treatment was four weeks, in three studies eight weeks, in two studies six weeks and in two studies six months (table 1). One study was a cross over of 20 weeks (two weeks wash in and washout) [28]. Diagnostic criteria for IBS used were as follows: ten trials used Rome II, two trials used Rome, and one trial used Manning. However, in Gade's study [18] published in 1989, the IBS definition of constipation and/or diarrhoea, abdominal pain, meteorism, borborygmus and flatulence was used. One study included constipation predominant IBS [20]; one included diarrhea-predominant symptom [22].
Measured outcomes varied, from overall gastrointestinal (GI) function to individual symptoms such as abdominal pain, flatulence and bloating. Four studies included quality of life using different questionnaires: Kajander et al [21] used RAND 36 item health survey [31] Guyonnet et al [20] used FDDQL questionnaire [32]. Whorwell [30] and O'Mahony [27] used IBS specific quality of life questionnaire [33].
Assessment of quality of studies
Methodological qualities varied (Table 2), with seven studies scoring two or less and four studies scoring a maximum of four.
Table 2 Assessment of methodological quality of randomised trials:
Overall improvement
Combined data suggested a modest improvement in overall symptoms after several weeks of treatment: for dichotomous data from seven trials [18–22, 24, 30] (895 participants) (figure 2) the overall Odds Ratio (OR) was 1.6 (95% CI, 1.2 to 2.2); heterogeneity I2 = 28%. For continuous data from six trials (657 participants) [20–22, 26, 27, 30], the standardised mean difference (SMD) was 0.23 (95% CI, 0.07 to 0.38), I2 = 0% (figure 3).
In the six trials [18, 20–22, 24, 30] of adults only (850 participants) the overall improvement in symptoms remained statistically significant, OR 1.59 (95% CI, 1.19 to 2.13), I2 = 35%. Results for overall improvement in symptoms were stable to sensitivity analysis [18–20]; (365 participants), OR 1.62 (95% CI, 1.06 to 2.48), I2 = 20%.
Abdominal pain
Seven trials [17–19, 21, 24, 26, 28] (398 participants) reported a statistically significant improvement in abdominal pain, OR 2.88 (95% CI, 1.84 to 4.50), I2 = 24% (figure 4).
In five studies of adults only [18, 21, 24, 26, 28] (297 participants) the effect remained statistically significant, OR 3.34 (95% CI, 1.99 to 5.61), I2 = 1%.
Two studies of children [17, 19] (101 participants) reported on improvement in abdominal pain. However, high heterogeneity (I2 = 63%) suggested pooling was inappropriate. In Bausserman's study [17] (64 participants), Lactobacillus GG (LGG) was not superior to placebo in relieving abdominal pain. There were no differences in other gastrointestinal symptoms, except for a lower incidence of perceived abdominal distension (P = 0.02 favouring LGG). In Gawroska's study [19] (37 participants), those in the LGG group were more likely to have treatment success (defined as no pain based on face pain scales) than those in the placebo group and had reduced frequency of pain (P = 0.02), but not severity of pain.
Using continuous data in nine trials [17, 19–23, 26, 27, 30] (792 participants) there was no statistically significant improvement in abdominal pain, SMD 0.05; (95% CI, -0.09 to 0.19), I2 = 51%. The heterogeneity was high. The most likely reason for this was the use of different scales in different studies; some used Likert scales ranging from 0 to 6 whereas others used visual analogue scales or face pain scales.
Flatulence
Five studies [18, 21, 22, 24, 26] (274 participants) reported a significant improvement in symptoms of flatulence, OR 2.31 (95%CI, 1.37 to 3.9), I2 = 7% (figure 5). Using continuous data in five studies [21–23, 26, 30] (388 participants), there was no statistically significant improvement in symptoms of flatulence, SMD 0.11; (95% CI, -0.09 to 0.31), I2 = 49%. The high heterogeneity was most likely to be due to use of different scales in different studies.
Bloating
Four studies [18, 21, 23, 28] (253 participants) reported a statistically significant improvement in symptoms of bloating, OR 1.75 (95% CI, 1.03 to 2.96), I2 = 0% (figure 6).
Using continuous data in six studies [20–23, 27, 30] (653 participants), there was no statistically significant difference in symptoms of bloating, SMD 0.05 (95% CI, – 0.10 to 0.21), I2 = 8%.
Quality of Life
Four trials aimed to report quality of life (QoL) but inadequate data prevented pooling of the results. In a study by O'Mahony et al [27] QoL was assessed by administration of an IBS specific questionnaire. They reported that for most domains, QoL scores were numerically lower than those for placebo for the patients randomized to the probiotics, but reached statistical significance versus placebo, during the treatment phase only, for health worry for bifidobacterium (at the 0.05 level) and dysphoria for lactobacillus at the 0.10 level. In Kajander's Study [21] the RAND 36 items health survey was used. They report that at baseline the mean QoL was somewhat higher in the probiotic group, but the difference between the groups was non significant compared with the baseline. There was no change in the mean score at three months or six months in either group.
Whorwell et al [30] used the IBS specific QoL questionnaire and the Hospital Anxiety and Depression Scale (HAD). No significant change in the QoL or HAD scores was reported with any of the probiotic dosages in comparison to placebo.
In a study by Guyonnet et al [20], health related QoL was assessed by the administration of the Functional Digestive Disorders Quality of Life questionnaire. They reported that the global score did not differ significantly between the probiotic group and the control group.
Adverse effects
Nine trials reported that there were no adverse effects with probiotics [17–19, 21–24, 26, 28], four trials reported adverse effects [20, 25, 27, 30] and one study [29] reported no data.
In the Niv study [25] (54 participants), the following adverse events were reported in the probiotic group: dyspepsia (one), headache (one) and nausea (none) and in the control group: dyspepsia (three), headache (none) and nausea (one). The differences were not significant. In the Whorwell study [30] (362 participants), 17 of subjects withdrew because of an adverse event. There was no significant difference between the groups and there were no details on the characteristics of adverse events. In O'Mahony's study [27] (80 participants), four subjects reported adverse events during the study. In Guyonnet's study [20], (274 participants) 23 subjects (ten from the control group and 13 from the probiotic group) reported minor adverse events. Four in the control group and three in the probiotic group stopped the product after an adverse event. No further details were given on the nature of these events.