Diarrhoea is defined by the World Health Organization (WHO) as three or more loose or watery stools (taking the shape of the container) in a 24-hour period. Diarrhoea is acute if the illness started less than 14 days previously, and persistent if the episode has lasted 14 days or more (Anonymous 1988). Normal infants who are exclusively breast fed may pass loose, "pasty" stools frequently. In this group the definition is usually based on what the mother considers to be diarrhoea (WHO 1990). Infectious diarrhoea is an episode of diarrhoea that is caused by an infectious agent.
Incidence and disease burden
Infectious diarrhoea occurs much more commonly in developing countries than industrialized countries (Guerrant 1990). Attack rates in developing countries are typically six to 12 episodes per child per year, compared with two in the USA (Savarino 1993). In a systematic analysis of population health data available for 2001, diarrhoeal diseases accounted for 1.78 million deaths (3.7% of total deaths) in low- and middle-income countries (Lopez 2006). Most of these deaths occur in children under five years of age. Although 50% or more children with diarrhoea receive oral rehydration therapy and continued feeding in only six of 60 priority countries, and only seven countries include zinc in diarrhoeal management (Bryce 2006), diarrhoeal deaths have reduced in this age group. However, diarrhoea still accounted for about 1.6 million deaths in 2001 (15% of all deaths in the under fives; Lopez 2006). In industrialized countries deaths from infectious diarrhoea occur mainly in the elderly (Savarino 1993).
More than 20 viruses, bacteria and parasites are associated with acute diarrhoea (Gadewar 2005). Worldwide, rotavirus is the most common cause of severe diarrhoea and diarrhoea mortality in children (Cunliffe 1998). Other important viral pathogens are astrovirus, human caliciviruses (norovirus and sapovirus) and enteric adenoviruses. Important bacterial pathogens are diarrheogenic Escherichia coli, Salmonella, Shigella, Yersinia, Campylobacter, and Vibrio cholerae. The main parasitic causes of diarrhoea are Cryptosporidium and Giardia (reviewed by O'Ryan 2005). An aetiological study of young children attending hospitals in China, India, Mexico, Myanmar, and Pakistan showed that rotavirus, enterotoxigenic E. coli and Shigella spp. were the most commonly isolated pathogens (Huilan 1991). Acute diarrhoea is frequent among travellers, in whom enterotoxigenic E. coli is particularly common (Black 1986). In practice, most episodes of acute diarrhoea that are assumed to be caused by an infectious agent are treated without the causative agent being identified. The major causes of acute infectious diarrhoea differ according to local factors, such as availability of clean water and sanitation. In contrast with acute infectious diarrhoea, infection is likely to be only one of several factors that contribute to the pathogenesis of persistent diarrhoea (Walker-Smith 1993).
The aim of treatment is to prevent or reverse dehydration, shorten the duration of the illness (important for preventing progression to persistent diarrhoea, which is associated with adverse outcomes such as malnutrition), and to reduce the period that a person is infectious. Treatment options available are oral rehydration solution, antibiotics, and gut motility-suppressing agents such as loperamide, codeine, and probiotics. This review considers the use of probiotics only.
Probiotics have been defined as microbial cell preparations or components of microbial cells that have a beneficial effect on the health and well-being of the host (Salminen 1999). Although organisms used in clinical trials may not have a proven health benefit for the indication being investigated, we have used the term "probiotic" in this review for simplicity. Fermenting foods to enhance their taste and nutritional value is an ancient and widespread practice. Well-known probiotics are the lactic acid bacteria and the yeast Saccharomyces (Naidu 1999). The taxonomy of the lactic acid bacteria relied traditionally on their phenotypic characteristics. Modern molecular techniques have shown these to be unreliable, and polyphasic taxonomy using both phenotypical and molecular techniques is now recommended (Klein 1998). Even closely related probiotic strains can have different clinical effects, and the Food and Agricultural Organization (FAO) of the United Nations and WHO expert consultation committee have emphasized that the beneficial effects observed in one strain cannot be assumed to occur in other strains (FAO/WHO 2001). This implies that the reliable identification of organisms at the strain level is necessary for clinical studies.
The rationale for using probiotics in infectious diarrhoea is that they act against enteric pathogens by competing for available nutrients and binding sites, making the gut contents acid, producing a variety of chemicals, and increasing specific and non-specific immune responses (Gismondo 1999; Goldin 1998; Vanderhoof 1998). No serious adverse effects of probiotics have been suggested in well people, but rarely, infections have been reported in people with impaired immune systems or indwelling catheters (Hata 1988; Piarroux 1999; Salminen 1998; Saxelin 1996; Sussman 1986).
Six systematic reviews of probiotics in acute diarrhoea have been published. Szajewska 2001 included only published, randomized, placebo-controlled double-blind studies of acute diarrhoea lasting three or more days in infants and children. A score was used to assess the methodological quality of these trials. The effects of all probiotics and individual strains were analysed. The risk of diarrhoea lasting three or more days was reduced by 0.40 in the probiotic compared with the placebo group (95% confidence interval (CI) 0.28 to 0.57, random-effects model, eight trials including 731 children), and probiotics reduced the duration of diarrhoea by 18.2 hours (95% CI 9.5 to 26.9 hours, random-effects model, eight trials including 773 children). The statistically significant heterogeneity in this result was resolved when one study, which employed a mixture of three probiotic organisms, was excluded. Lactobacillus GG was thought to be particularly effective in rotavirus diarrhoea.
A meta-analysis undertaken by Van Niel 2002 was restricted to adequately randomized and blinded studies of several strains of lactobacilli in children. Children who had received recent antibiotics were excluded from the study. Probiotics reduced the duration of diarrhoea by 0.7 days (95% CI 0.3 to 1.2 days, seven studies including 675 children) and diarrhoea frequency on day 2 by 1.6 (95% CI 0.7 to 2.6, three studies including 122 children). The heterogeneity of results among the studies prevented an analysis of the effects of individual strains of lactobacilli.
Three meta-analyses have focused on randomized controlled trials of specific probiotics in acute infectious diarrhoea in children. Szajewska 2007a analysed trials of Lactobacillus casei strain GG where a > 80% follow up was achieved. Trial results published as letters to the editor, abstracts, and proceedings from scientific meetings were not included. L.casei GG reduced the duration of diarrhoea by 1.1 days (95% CI 0.3 to 1.9, seven trials, 876 infants) and was particularly effective in rotavirus diarrhoea (duration reduced by 2.1 days, 95% CI 0.6 to 3.6). However, the authors urged caution in the interpretation of the results in view of methodological limitations in the trials and the heterogeneity of the results in the studies. Chmielewska 2008 identified two trials of Lactobacillus reuteri strain ATTCC 55730. This probiotic reduced the duration of diarrhoea by 22 hours (95% CI 6 to 38, 106 participants). In an update of a previous review (Szajewska 2007b), Szajewska 2009 pooled data from seven randomized controlled trials of Saccharomyces boulardii in 944 otherwise healthy children with acute gastroenteritis. The duration of diarrhoea was reduced by 1.08 days (95% CI 0.53 to 1.64) in children who received the yeast compared with the placebo although there was marked heterogeneity in results among the studies.
A recent review concluded that the beneficial effects of probiotics in acute infectious diarrhoea were dependent on the strain of bacteria and the dose (a greater effect with doses >10
Our review aims to assess the evidence base to inform the use of probiotics in acute infectious diarrhoea. To maximize use of available data, we included participants of all ages, unpublished studies, and non-blinded (open) studies. We assessed the relevant methodological aspects of trials individually (Juni 1999). These were the generation of allocation sequence, allocation concealment, blinding, and loss to follow up. To maximize the relevance of our findings for clinical practice we included studies in which participants with infectious diarrhoea had received antibiotics prior to recruitment.
For primary outcomes, we chose the duration of diarrhoea and diarrhoea lasting ≥ 4 days, as these are directly relevant to the development of persistent diarrhoea, and stool frequency on day 2 after intervention as a marker of diarrhoea severity.
This review is a substantial update of the original version, first published in 2003 (Allen 2003).
To assess the effects of probiotics in proven or presumed acute infectious diarrhoea.
Criteria for considering studies for this review
Types of studies
Randomized and quasi-randomized controlled trials reporting the effect of probiotic(s) on acute infectious diarrhoea. Studies of probiotics in acute diarrhoea that reported other outcomes (eg their effect on rotavirus shedding in stools) but no diarrhoea outcomes were not included.
Types of participants
Adults and children with acute diarrhoea (duration < 14 days) that was proven or presumed to be caused by an infectious agent.
Excluded: studies of diarrhoea known or thought to have other causes (eg antibiotic-associated diarrhoea and studies of persistent diarrhoea).
Types of interventions
Specific, identified probiotic.
Excluded: yogurt or other fermented foods in which specific probiotic organisms were not identified.
Placebo or no probiotic.
Intervention and control arm to be otherwise treated identically in relation to other treatments and drugs.
Types of outcome measures
Duration of diarrhoea
Diarrhoea lasting ≥ 4 days
Stool frequency on day 2 after intervention
Diarrhoea lasting ≥ 3 days
Stool frequency on day 3 after intervention
Search methods for identification of studies
We have attempted to identify all relevant studies regardless of language or publication status (published, unpublished, in press, and in progress). Searches of all databases was done on 1 July 2010.
We searched the Cochrane Infectious Diseases Group's trials register using the search terms: diarrhea/; diarr$(tw); diarhea(tw); probiotic(tw); Lactobacill$(tw); Lactococc$(tw); Bifidobacter$(tw); Enterococc$(tw); Streptococc$(tw); Saccharomyces(tw). Full details of the Cochrane Infectious Diseases Group's methods and the journals handsearched are published in The Cochrane Library in the section on 'Collaborative Review Groups'.
We searched the Cochrane Controlled Trials Register published on The Cochrane Library (Issue 2, 2010) using the search terms: diarrhea/; diarr$(tw); diarhea(tw); probiotic(tw); Lactobacill$(tw); Lactococc$(tw); Bifidobacter$(tw); Enterococc$(tw); Streptococc$(tw); Saccharomyces(tw).
We searched MEDLINE (1966 to 2010) and EMBASE (1988 to 2010 using the search strategy defined by The Cochrane Collaboration (Clarke 2003) and following search terms: diarrhea/; diarr$(tw); diarhea(tw); probiotic(tw); Lactobacill$(tw); Lactococc$(tw); Bifidobacter$(tw); Enterococc$(tw); Streptococc$(tw); Saccharomyces(tw).
The detailed search strategy is shown in Appendix 1.
In preparation for the original review (Allen 2003), we contacted organizations and individuals working in the field, and the following pharmaceutical companies that manufacture probiotic agents to help identify additional published trials and unpublished data: Biogaia Biologicals, Lund, Sweden; Nestle Foundation, Lausanne, Switzerland; Probiotics International Ltd, Somerset, UK; Ross Products Division of Abbott Laboratories, Ohio, USA, and Yakult, London, UK. We have not re-contacted individuals or companies for this update.
We also drew on existing reviews of this topic and checked the citations of all trials identified by the above methods.
Data collection and analysis
SA and LD independently reviewed the titles of articles and, where available, abstracts generated by the search to identify potentially relevant studies. All articles that could meet the inclusion criteria as identified by either of the reviewers were selected and the full article reviewed. Eligibility was assessed independently by SA and LD using a form based on the information presented in the article. We planned to contact trial authors if eligibility was unclear. Discrepancies among reviewers' eligibility assessments were resolved by discussion. Trial reports were scrutinized to ensure that multiple publications from the same trial were included only once. Excluded studies and the reasons for their exclusion were listed.
Assessment of methodological quality
Two reviewers (EM, GG), blinded to the origin of the articles, independently assessed the methodological quality of identified studies using generation of allocation sequence, allocation concealment, blinding, and loss to follow up, and we recorded this information on a standard form.
We considered the generation of allocation sequence to be adequate if the study authors stated that they used a method resulting in unpredictable sequences (such as a random number table or list or computer-generated random numbers), unclear if a trial was stated to be randomized but no further information was provided and inadequate where allocation could be related to prognosis and therefore introduced selection bias (for example, the date of birth or date of admission to hospital).
We considered allocation concealment to be adequate if the assignment to arms of the study could not be predicted by the investigators or participants (for example, central randomization or numbered, identical drug containers), unclear if the method used to achieve concealment was not described or inadequate if they used a method such as alternation where the allocation of participants could be predicted.
We considered blinding to be adequate when studies were double blind (when an identical placebo was used and recruitment to intervention or control arms was not known by either the investigator or the participants), unclear if methods of blinding were not described adequately, and inadequate when blinding was not used or where the authors stated that unblinding had occurred.
We considered loss to follow up to be adequate when study endpoints were presented for 90% or more of the participants enrolled at the beginning, inadequate when follow up was less than this and unclear when either or both the number of participants recruited at the beginning of the study and the number of participants who completed the study were not clear.
LD resolved disagreements regarding the assessment of methodological quality.
SA, BO, SP, and SA independently extracted data using standard forms. Key data items were the aetiology and duration of diarrhoea, details of probiotic organism, participants' characteristics (nutritional and human immunodeficiency virus (HIV) status), location (countries classified according to mortality stratum; WHO 2001), and the outcome measures listed above. The number of participants recruited and the number for whom outcome data was reported were extracted and included in the Characteristics of included studies table.
For dichotomous outcomes, the number of participants experiencing the event, and the total number of participants in each intervention group was extracted. For continuous outcomes, arithmetic means, standard deviations (SD), and the numbers of participants in each intervention group was extracted. SDs were calculated from 95% CI and standard errors, where these were reported. The findings of trials that presented data that could not be included in pooled analyses (eg median and inter-quartile range (IQR)), or reported outcomes other than the primary and secondary outcomes employed in this review were reported in the text.
We pooled data from studies that used comparable outcome measures. For the duration of diarrhoea and number of stools per day of intervention, we achieved a pooled estimate of treatment effect by calculating the weighted mean difference. For the number of participants with diarrhoea lasting 3 days or more, or 4 days or more after starting the intervention, we calculated a pooled estimate of the relative risk (RR) among probiotic and non-probiotic groups.
We reported the proportion of participants for whom outcome data were available in a 'Risk of bias' table for each study. We performed analyses according to the intention-to-treat principle using an available case analysis approach.
Where there was significant heterogeneity (P < 0.1) in outcomes across studies assessed by the Chi
We inspected the forest plots to detect non-overlapping CIs, applied the Chi
We proceeded to pool data for meta-analysis to provide a qualitative assessment of probiotic effect as a guide to clinical practice.
We expected that heterogeneity in results among studies would result from clinical diversity, including differences in probiotic(s) used, dose of organisms, types of participants, causes and severity of diarrhoea and the socioeconomic status of countries where the studies were undertaken (Wolvers 2010). Therefore, where there were results for a diarrhoea outcome available from three or more studies we conducted subgroup analyses according to the:
- probiotic strain; single probiotic organisms versus combinations of two or more organisms, dose of live organisms (high dose [> 10
10CFU/day] versus lower dose [≤10 10CFU/day]); killed organisms;
- age of participants;
- identified diarrhoeal pathogens (rotavirus, bacterial diarrhoea);
- severity of diarrhoea according to whether the participants were likely to have had mild diarrhoea and were, therefore, managed as outpatients;
- mortality stratum for children and adults in the country or countries where the trial was undertaken (WHO 2001) to account for regional differences in major diarrhoeal pathogens and diarrhoea severity related to the availability of clean water and level of sanitation. To facilitate meta-analysis, countries were divided into two groups according to whether either child or adult mortality, or both, was classified as high.
Finally, we inspected funnel plots for the primary outcomes to assess publication bias.
Description of studies
Our search identified 120 potentially relevant studies. Of these, 63 met the inclusion criteria. Overall, 57 studies were excluded, including five that were preliminary or duplicate reports of other included studies (Characteristics of excluded studies). Eligibility regarding inclusion in this review was clear for all studies and clarification from trial authors was not required. We have not been able to locate the reports of two studies (Contreras 1983; Salgado) and one study is ongoing (Freedman 2010). None of the 56 included trials were cluster randomized.
Of the 63 included studies, 23 were published in the 1980s-1990s, 37 between 2000-2009 and two in 2010; one study was unpublished.
According to country mortality strata for children/adults (WHO 2001), 41 trials were undertaken in countries where both child and adult mortality was classified as low or very low and 19 where either child or adult mortality was high. Two international studies recruited participants from countries crossing the mortality strata (Guandalini 2000; Jasinski 2002). Finally, the study by Ritchie 2010 was undertaken in Australia (very low child and low adult mortality) but recruited Aboriginal children who commonly had co-morbidities such as pneumonia and malnutrition related to poverty and social disadvantage in the top end of the Northern Territory. Therefore, data from this study were not included in analysis according to country mortality strata.
A total of 47 studies were conducted in a single centre; 15 recruited participants from two to 150 centres. The number of recruitment centres was unclear in one study (D'Apuzzo 1982).
The 63 selected studies recruited a total of 8014 participants. There were 6489 infants and children (age < 18 years) and 352 adults. In three studies (1173 participants) the exact ages of participants was not clear: Bruno 1983 studied participants aged 14 years and above, participants in Wunderlich 1989 had a mean age of 33 years (age range not stated) and the age of the participants in Frigerio 1986 was not stated.
Forty-four studies recruited inpatients, seven recruited outpatients and seven recruited both inpatients and outpatients. It was unclear in five studies whether the participants were inpatients or outpatients.
Although all studies recruited participants with acute diarrhoea, the criteria for acute diarrhoea varied considerably among studies (see Characteristics of included studies). Descriptions of stool consistency included watery, loose or liquid stools, or both, semi-liquid, increased fluidity, pasty, mucousy or non-formed in 46 studies but no description was stated in 17 studies. The minimum number of stools/day was specified in 36 studies; this ranged from ≥one to ≥ five stools with the most commonly used criteria being ≥ three (16 studies) and ≥four stools in 24 hours (13 studies). One study specified stool frequency as at least twice normal frequency, one as increased frequency and in one study stool consistency was taken into account. The minimum number of stools was not specified in 24 studies. The maximum duration of diarrhoea at recruitment was specified in 40 studies and varied between one and 14 days. The maximum diarrhoea duration was not specified in 23 studies.
Similarly, criteria used for the end of the diarrhoeal episode varied markedly among studies. The last liquid or watery stool (nine studies) and first normal stool (seven studies) were the most common. Twenty-one studies used a variety of criteria based on stool frequency and consistency in a specified period (eg first formed stool if followed by two consecutive non-watery stools or 12 hours without evacuation; Mao 2008). Four studies also included the resolution of associated symptoms (eg < two stools/day, formed, yellow/brown stools without mucus and no abdominal pains, vomiting, or fever for the whole day; D'Apuzzo 1982). Criteria were not stated in 17 studies.
Eighteen studies were either restricted to children with rotavirus diarrhoea or reported outcomes for a subgroup of children with rotavirus diarrhoea. Children with rotavirus diarrhoea were excluded in one study (Lievin Le-Maol 2007). Ten studies stated that participants with bloody diarrhoea were included whereas these were excluded in 32 studies. It was unclear whether participants with bloody diarrhoea were included in 21 studies. No study specifically recruited or excluded travellers, and none identified any of the participants as suffering from travellers' diarrhoea.
No study specifically recruited participants known to have HIV infection and no study stated HIV positivity as an exclusion criterion, but many excluded participants with chronic illness or immunosuppression, or both.
Nutritional status was reported in 35 studies, all undertaken in children. Ten studies either recruited malnourished children only or included malnourished children; 20 studies excluded severe malnutrition; five studies recruited well-nourished children only or excluded those with a chronic illness.
Twenty-six studies excluded participants who had received antibiotic treatment before recruitment, eight included participants who had received antibiotic treatment before recruitment and this information was unclear in 29 studies.
The hydration status of the participants was reported in 35 studies; 22 studies included participants with severe dehydration whereas 10 studies recruited only children with mild or moderate dehydration.
Many different probiotics were tested. Most studies tested live preparations of lactic acid bacteria and bifidobacteria. Several studies identified the probiotic organisms only by the species name without specific identification details such as a culture collection number. Few studies undertook analyses to confirm the identity or viability of the organism(s).
Forty-six studies tested a single organism and 17 tested combinations of between two to eight organisms. The most common organisms evaluated were L. casei strain GG (13 studies), S. boulardii (10 studies) and Enterococcus lactic acid bacteria (LAB) SF68 (five studies). All other organisms and all combinations were tested in three or fewer studies. Canani 2007 allocated children to one of five different probiotic regimens and compared outcomes with a single control group. For the purposes of this review, we selected the L. casei GG group for inclusion because several other studies tested this probiotic and we wanted to maximize the data available for meta-analysis. Grandi 2009 allocated children either to a single organism or a four-organism group and compared outcomes with a single control group. No data extractable for meta-analysis were reported in this study.
Forty-seven studies tested live organisms, five studies tested a killed probiotic preparation (Billoo 2006; Boulloche 1994; Lievin Le-Maol 2007; Simakachorn 2000; Khanna 2005), and one a pasteurized yogurt (Pashapour 2006). The viability of the organisms was unclear in 10 studies.
Three studies compared different dosages (number of organisms) of the same probiotic (Basu 2009, Mao 2008, Shornikova 1997b) with a single control group. We selected the higher probiotic dose group for inclusion in the review but have included results from the lower dose group in the text. Overall, 15 studies used a high dose of organisms (> 10
As well as differences in dose or organisms, there was a wide variation in the treatment regimens according to timing of intervention, means of administration and duration of treatment. Probiotics were administered directly to the participants or mixed with a variety of fluids and foods. Although expressed breast milk was used to administer probiotics in some studies, some studies excluded exclusively breast-fed infants to minimize the interruption of normal feeding.
Forty-three studies used a placebo in the no probiotic control group; the remaining studies managed participants according to usual clinical practice.
Risk of bias in included studies
Methodological quality varied considerably (see Characteristics of included studies). Twenty-three studies were considered adequate for generation of the allocation sequence, 15 for concealment of allocation, 35 for blinding and 45 for loss to follow up. Ten studies were adequate for all of the four methodological quality assessment parameters and five studies were inadequate for all four parameters.
Effects of interventions
The forest plots demonstrate that probiotics reduce the duration of diarrhoea. Values for duration of diarrhoea in the control arm varied widely, from 39.1 to 173.5 hours, and the difference between the intervention groups ranged from -79.2 to 7.0 hours ( Analysis 1.1). Similar variability was evident in the other outcomes. Despite the high level of quantitative heterogeneity, the pattern was striking, and meta-analysis shows an important effect which is statistically significant. Using a random effects approach, probiotics reduced the mean duration of diarrhoea (mean difference 24.76 hours; 95% confidence interval 15.9 to 33.6 hours; n=4555, trials=35; Analysis 1.1), diarrhoea lasting ≥4 days (risk ratio 0.41; 0.32 to 0.53; n=2853, trials=29; Analysis 1.2) and stool frequency on day 2 (mean difference 0.80; 0.45 to 1.14; n=2751, trials=20; Analysis 1.3). The differences in these analyses are an average across all studies with quantitative heterogeneity, demonstrating that probiotics have a substantive and significant effect, rather than being a precise estimate of the size of the effect.
The funnel plots for the primary outcomes (Figure 1, Figure 2, Figure 3) did not indicate publication bias as the largest intervention effects were observed in studies with a large number of participants as well as smaller studies.
|Figure 1. Funnel plot of comparison: 1 Primary diarrhoea outcomes, outcome: 1.1 Mean duration of diarrhoea.|
|Figure 2. Funnel plot of comparison: 1 Primary diarrhoea outcomes, outcome: 1.2 Diarrhoea lasting ≥ 4 days.|
|Figure 3. Funnel plot of comparison: 1 Primary diarrhoea outcomes, outcome: 1.3 Mean stool frequency on day 2.|
The findings for diarrhoea lasting ≥ 3days ( Analysis 2.1) and stool frequency on day 3 after intervention ( Analysis 2.2) were broadly similar to the primary outcomes and there was also marked statistical heterogeneity among studies.
Seven studies reported diarrhoea outcomes data that could not be included in analyses. Billoo 2006 evaluated S. boulardii in infants and children admitted with acute watery diarrhoea of mild to moderate severity in Pakistan. The mean duration of diarrhoea was reduced in the probiotic compared with the control group (n = 50, 86.4 hours versus n = 50, 115.2 hours, respectively; P = 0.001). Stool frequency on days 3 (P = 0.01) and 6 (P = 0.001) was also reduced in the probiotic group. Czerwionka 2009 evaluated Lactobacillus rhamnosus in children with acute diarrhoea in Poland. The total number of stools per child was statistically significantly lower in the probiotic group than the controls. Misra 2009 evaluated L. rhamnosus GG in children in India. The mean duration of diarrhoea was 70.6 hours in the probiotic and 78.0 hours in the control group (P = 0.20).
Grandi 2009 allocated young children admitted with acute rotavirus diarrhoea to receive either oral rehydration fluid (ORF) + S. boulardii, ORF + Lactobacillus acidophilus, L. rhamnosus, Bifidobacterium longum and S. boulardii, or ORF alone (control group). The median duration of diarrhoea was shorter in both of the probiotic groups compared with the controls but this was statistically significant only for S. boulardii (58 hours versus 84.5 hours, respectively; P = 0.04).
In a short abstract, Frigerio 1986 reported that the duration of diarrhoea was significantly reduced (P < 0.01) among 540 patients with an acute diarrhoeal disorder attending hospitals in Italy who received Enterococcus LAB SF 68 compared with 534 who received a placebo.
In a further study, Sepp 1995 evaluated adding L. casei GG to trimethoprim-sulfamethoxazole, compared to trimethoprim-sulfamethoxazole alone, in children with acute diarrhoea caused by shigellosis in Estonia. The duration of diarrhoea was similar in the probiotic (median 0.5 days) and the control group (1 day; not statistically significant). Also, the proportion of children with ongoing diarrhoea on day 5 was similar in the probiotic and control groups (6/13 (46.3%) versus 9/12 (75.0%); not statistically significant). However, a greater proportion was cured in the probiotic than the control group on day 10 (P < 0.05). Finally, in an open study, Táborská 1997 evaluated live L. acidophilus ND in infants and children admitted with acute gastroenteritis in the Czech Republic. The resolution of enteric symptoms during days 1 to 5 of the intervention was similar in the two groups.
Exploration of heterogeneity
Sensitivity analysis for primary outcomes
When analysis was restricted to trials assessed to be adequate for the four criteria of study quality (Characteristics of included studies), highly statistically significant between-study heterogeneity persisted (forest plots not shown ( Table 1). This suggests that differences in outcomes between studies were caused by factors other than differences in methodological quality.
In addition to the methodological quality of studies as a potential source of heterogeneity in the primary outcomes, we explored other prespecified factors in subgroup analyses where outcomes were reported in three or more studies (probiotic strain: Analysis 3.1, Analysis 3.2, Analysis 3.3; single organism versus combinations: Analysis 4.1, Analysis 4.2, Analysis 4.3; live versus killed organisms: Analysis 5.1; dose or organisms: Analysis 6.1, Analysis 6.2, Analysis 6.3; children with rotavirus diarrhoea: Analysis 7.1, Analysis 7.2; severity of diarrhoea: Analysis 8.1 and, finally, mortality stratum for children and adults in the countries where the studies were undertaken: Analysis 9.1, Analysis 9.2, Analysis 9.3). With few exceptions, the magnitude of probiotic effect on diarrhoea outcomes was similar to that for all trials and marked heterogeneity in results persisted in the sub-group analyses.
In three of the sub-group analyses of trials that reported mean stool frequency on day 2, the magnitude of the effect in the intervention group was similar to that for all trials but there was greater consistency in the findings. This occurred in six trials (1335 participants) that assessed L. casei strain GG ( Analysis 3.3), eight trials (861 participants) that used a high dose of live organisms (> 10
The sub-group analysis according to diarrhoea severity suggested that probiotics resulted in a greater reduction in mean duration of diarrhoea in mild diarrhoea (studies of out-patients) than in more severe diarrhoea (inpatients; Analysis 8.1). However, marked heterogeneity in findings persisted and, therefore, the significance of this finding is unclear.
Finally, probiotics appeared to be less effective in reducing mean stool frequency on day 2 in countries with high child and adult mortality rates compared with those with low or very low mortality rates ( Analysis 9.3). However, marked heterogeneity persisted and probiotic effects were similar in both settings for the other diarrhoea outcomes ( Analysis 9.1; Analysis 9.2).
On balance, we found no clear evidence that stratification according to the sub-groups modified probiotic effect.
Several studies reported findings relevant to the subgroup analyses that could not be included in the analyses.
Probiotic organisms; strain, single organisms versus combinations and dose
Canani 2007 reported a statistically significantly reduced mean duration of diarrhoea for three different probiotics (live L. casei strain GG ( Analysis 1.1), a combination of live Lactobacillus delbrueckii, L. acidophilus, Streptococcus thermophilus and Bacillus bifidum, and S. boulardii) compared with controls but there was no effect of live Enterococcus faecium SF68 or live Bacillus clausii strains O/C84, N/R84, T84, SIN84. These findings were generally supported by effectiveness in reducing stool frequency on d 2 and 3 reported in this study, except that the live L. casei strain GG did not reduce stool frequency on day 3 ( Analysis 2.2) and S. boulardii did not reduce stool frequency on day 2.
Grandi 2009 allocated children with rotavirus diarrhoea to either an S. boulardii group or a group treated with a combination of four organisms (L. acidophilus, L. rhamnosus, B. longum and S. boulardii). The median duration of diarrhoea was shorter in both of the probiotic groups compared with the controls, but this was statistically significant only for S. boulardii (58 hours versus 84.5 hours, respectively; P = 0.04).
Three studies directly compared different doses of the same probiotic preparation in infants and children, most of whom had rotavirus diarrhoea. Mao 2008 evaluated two dose levels of a combination of Bifidobacterium lactis B12 and S. thermophilus TH4. Probiotics were administered in milk powder but the number of organisms administered in each group was not clear. The mean duration of diarrhoea and number of liquid stools/day were similar in the low dose and high dose groups.
Shornikova 1997b evaluated L. reuteri 10
Finally, on the basis of their previous study that did not show an effect of a low dose of L. rhamnosus GG on acute diarrhoea in a dose of 120 x 10
Age of participants
It was not possible to assess the effects of probiotics in adults as < three studies reported the same diarrhoea outcomes. The primary analysis of mean duration of diarrhoea did not include studies undertaken in adults ( Analysis 1.1). Removing studies of adults from the other primary analyses did not reduce heterogeneity (forest plots not shown). Overall, there was insufficient evidence regarding the efficacy of probiotics according to participants' age.
Children with rotavirus diarrhoea
In keeping with the findings for all children in their study, Simakachorn 2000 reported that fewer children with rotavirus diarrhoea in the probiotic than the control group had watery diarrhoea after 24 hours (3/19 versus 9/16; P = 0.012). Similarly, Boulloche 1994 reported that the resolution of diarrhoea in the probiotic group was similar for rotavirus positive and rotavirus negative participants.
Guandalini 2000 reported that mean stool frequency on day 3 of intervention was lower in the probiotic group (0.4, n = 56) than in the controls (2.0, n = 45; P < 0.05) and this was a greater reduction than that seen in all-cause diarrhoea in this study. In contrast, Costa-Ribeiro 2003 reported that there was no significant difference in the stool output or duration of diarrhoea between children allocated to probiotics versus placebo.
Only four trials reported outcomes for participants confirmed to have bacterial diarrhoea. Two studies assessed L. casei strain GG. Shornikova 1997a reported that the stool frequency was similar in the probiotic (n = 11) and placebo (n = 15) groups (P = 0.42). Guandalini 2000 reported that the mean (SD) duration of diarrhoea was similar in the probiotic and control groups (n = 35, 73.3 (29.3) versus n = 34, 72.0 (32.4) hours, respectively). The mean stool frequency on day 2 was also similar in the probiotic and control groups (5.0 and 5.5, respectively). Chen 2010 evaluated a combination of three organisms and reported that the mean (SD) duration of diarrhoea was not reduced significantly in children receiving probiotics (n = 27, 71.6 (32.8) hours) compared with controls (n = 30, 101.5 (46.8) hours; P = 0.082). In contrast, Htwe 2008 reported that in 21 children with pathogenic E. coli in stools, S. boulardii significantly improved stool consistency on d 3 (P = 0.004) and 4 (P = 0.025) compared with controls.
Of all 63 selected studies, 43 studies reported no adverse events and 20 gave no information on adverse events. Henker 2008 reported that one participant in the probiotic group had a mild hypersensitivity reaction that was assessed as being possibly related to the intervention. However, these authors commented that the probiotic was safe and well tolerated. With this exception, no authors reported an adverse effect that they considered to be attributable to the probiotic.
Many studies reported on vomiting. Boudraa 2001 reported a similar frequency of vomiting in the probiotic and control groups. Pant 1996 reported that 1/19 children in the control group vomited one dose of the medication, but no vomiting occurred in the 20 children in the probiotic group. Raza 1995 reported that the frequency of vomiting on the second day of intervention was statistically significantly less in children in the probiotic than the placebo group. Shornikova 1997c reported that fewer children in the probiotic than the placebo group vomited from the second day of treatment and this was statistically significant on day 2 and day 4. No child in the probiotic group vomited after the third day of treatment whereas vomiting persisted to the sixth day in 2/21 children in the placebo group. Kurugol 2005 reported that one child had meteorism but the group allocation was not stated.
A striking finding of this review is that most trials reported that probiotics improved diarrhoea. A beneficial effect of probiotics was consistent across the different diarrhoea outcomes and was statistically significant in many trials.
With the exception of possible mild hypersensitivity to E. coli strain Nissle reported in one participant (Henker 2008), no authors reported adverse events that they attributed to probiotics. Vomiting is common in acute diarrhoea and was the most frequently reported adverse event. Vomiting occurred less frequently in the probiotic than the control groups and, therefore, would appear to be a symptom of the illness rather than an adverse effect of probiotics. The reasons for non-compliance with protocol in some studies were not stated, but were unlikely to be related to the adverse events of probiotics since similar numbers of participants in the probiotic and control groups failed to comply. The causes of the withdrawal of participants from trials were related mostly to their primary illness rather than the interventions. Although this review supports the excellent safety record of probiotics, most of the studies recruited previously healthy people and more studies of susceptible individuals, for example, malnourished children and people with human immunodeficiency virus infection, are required to further evaluate safety.
The marked statistical heterogeneity between studies was expected given the marked clinical diversity in the definitions of diarrhoea and end of the diarrhoeal episode, the probiotic(s) tested, the treatment regimens, the diarrhoeal pathogens identified, the types of participants and the settings in which the trials were undertaken. Although these factors varied greatly among studies, individual studies used the same criteria and outcomes for both the probiotic and control groups. Although there was great variability in the methodological quality of the trials, there was no evidence that poor study design had led to an overestimate of the effects of probiotics.
Few studies reported outcomes for participants with bacterial diarrhoea and it was not possible to extract data for meta-analysis from any of these studies. Many of the other studies that reported a beneficial effect of probiotics included a significant proportion of participants with bacterial diarrhoea or bloody stools, or both. Although this suggests that probiotics are efficacious, more research is needed to assess probiotics in bacterial diarrhoea.
The subgroup analyses did not explain between-study statistical heterogeneity. Therefore, this review does not find important differences in probiotic effect according to probiotic strain, the number of different strains, the viability of the organisms, low versus high dose preparations, the causes or severity of diarrhoea or whether the studies were done in developed or developing countries. These findings are encouraging as effective interventions to prevent the progression from acute to persistent diarrhoea (> 14 days; closely associated with malnutrition in children in developing countries [Walker-Smith 1993]), are a priority.
The persistence of statistical heterogeneity in subgroup analyses is perhaps not surprising given the marked clinical variability among studies. This was demonstrated clearly by the wide range of values for primary outcomes reported in participants allocated to the control groups. There is general consensus that effects of probiotics are strain-specific and that results obtained with one probiotic cannot be extrapolated to other organisms, including closely related strains (Rijkers 2010). However, this review found that studies tested many different probiotics in many different settings yet nearly all reported beneficial outcomes. This suggests that a mechanism common to most probiotics, for example, colonization resistance, is effective against a wide range of gut pathogens. Probiotics are likely to have multiple mechanisms of action in the gut that may include effects on host immunity and gut mucosal barrier integrity as well as effects against diarrhoeal pathogens. Variations in several host and environmental factors that may determine the commensal gut flora may modify probiotic efficacy (Wolvers 2010). These include age, diet and eating practices, level of sanitation and exposure to antibiotics. It is likely that other factors, not considered in this review, underlie the marked among-studies heterogeneity.
The marked clinical variability among studies complicates meta-analysis and, therefore, weakens the evidence base to inform clinical practice. In particular, variability in the definition of diarrhoeal episodes results in misclassification and impairs the comparability of the findings from different studies (Baqui 1991). More large, well-designed studies are needed of specific probiotic regimens in specific settings. In future research, the standardization of definitions of acute diarrhoea, treatment regimens, inclusion criteria and outcome measures are needed to facilitate comparison of results across studies. All studies should try to present data separately for important subgroups, for example, according to participant nutritional status and identified causes of diarrhoea, such as rotavirus or bacterial causes. Guidance on undertaking trials with probiotics, such as reliably identifying the agent used, testing the viability of organisms and confirming their quantity, is readily available (Rijkers 2010; Wolvers 2010). Since most episodes of acute diarrhoea are uncomplicated, self-limiting, and require no specific treatment, cost-effect analyses need to determine whether probiotics should be used in particular groups of people.
Implications for practice
Probiotics administered in addition to rehydration therapy resulted in clear reductions in the duration and severity of diarrhoea, and were not associated with adverse effects. This review supports the use of probiotics in acute, infectious diarrhoea. However, marked clinical variability between studies resulted in insufficient studies of specific probiotic regimens in defined groups of children or adults to inform the development of evidence-based treatment guidelines.
Implications for research
Although many different probiotics were effective in reducing diarrhoea, to better inform clinical practice studies of specific probiotic regimens in large numbers of participants with well-defined diarrhoeal illness are needed. Trials need to use standardized definitions for acute diarrhoea and the resolution of the illness. They need to identify infectious causes of diarrhoea and present data separately for important participant subgroups, such as viral and bacterial causes of diarrhoea. All studies should include a reliable identification of the probiotic being tested, and confirm the viability and number of organisms for live probiotics. More research is needed to assess the role of probiotics in developing countries, especially in preventing the progression from acute to persistent diarrhoea and associated malnutrition.
Basic research is needed to identify generic and strain-specific mechanisms underlying the apparent beneficial effects of probiotics in acute diarrhoea.
Thanks to Dr Brown Okoko, Sam Parker and Stephanie Allen for help with data extraction.
The authors would like to dedicate this review to the memory of Dr Brown Okoko, an author on the previous version of this review, who died unexpectedly in 2008.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
Appendix 1. Detailed search strategy
Last assessed as up-to-date: 10 August 2010.
Protocol first published: Issue 2, 2001
Review first published: Issue 2, 2004
Contributions of authors
Stephen Allen and Leonila Dans identified articles for inclusion in the review. Leonila Dans, Elizabeth Martinez, and Germana Gregorio assessed study quality, and Leonila Dans settled any disagreements. Stephen Allen extracted data. Stephen Allen took the main responsibility for analysis and writing the review. All reviewers contributed to the final version.
Declarations of interest
Stephen Allen is participating in ongoing research studies of lactobacilli and bifidobacteria provided by Cultech Ltd, UK, in the prevention of atopic disorders in infants and antibiotic-associated diarrhoea in older people. In previous research, Scientific Hospital Supplies, UK, and Valio Ltd, Finland, have provided L. casei strain GG and also supported his attendance at a training workshop. Elizabeth Martinez is a Medical Manager for United Laboratories Inc., Philippines.
Sources of support
- Swansea School of Medicine, UK.
- Cochrane Infectious Disease Group, Liverpool School of Tropical Medicine, UK.
Differences between protocol and review
The following secondary outcomes have been removed as they were either uncommon or not reported: need for unscheduled intravenous (IV) rehydration after randomization; deaths; adverse events, such as vomiting; withdrawal from study.
This review is a substantial update of the original version, first published in 2003 (Allen 2003).
Medical Subject Headings (MeSH)
MeSH check words
Adult; Child; Child, Preschool; Humans; Infant
* Indicates the major publication for the study